Tag for finding my other articles on this stock: MRP

Key takeaways

Strong customer demand but still no answer to fundamental questions around business model and returns. Current discount to book value is 18% which feels about right to me.

Subscribe now

Thesis and valuation update

No change to thesis. The risk is that this works until it doesn’t.

• Millrose clearly have strong customer demand, but I worry the business model is flawed.

• The first problem is that if homebuilders refuse to exercise options, Millrose’s cash flow stops abruptly. So far every option contract has been honoured, but I worry what happens in a recession.

• The second problem is that Millrose cannot retain profits but it does retain losses. Mathematically, I think book value has to fall over time.

• The third problem, and the biggest one for now, is that Millrose cannot grow without either lifting the debt ceiling, issuing shares, or coming up with a funky new capital structure. If Millrose can trade above book value it could issue shares and grow, but why would it? The ROE does not justify a p/bv >1x. Commentary this quarter suggests they will lift the debt ceiling, which will drive growth but also increase risk.

Share

Notes

• Net income 0.76/share.

  • AFFO/share $0.77. Run rate AFFO $0.80/share => $1.60 per year.

  • Dividend up to $0.77 - 100% AFFO payout.

  • BVPS $35.24, down 2c.

  • Invested capital now $8.8bn.

  • Now have 19 counterparties.

  • 32% of invested capital is not with Lennar and new issuances in this bucket yielded 10.6% in the q.

  • Received $1bn in the q as builders drew down lots, and reinvested $1.1bn, with the balance funded with debt.

  • No change in underwriting conditions. They mentioned that deposits have come down from 20-25% to 10% but that their credit policy is agnostic between bigger deposits and better terms. Need to understand this trade-off better.

• Zero option terminations since inception.

  • Every counterparty has honoured every option, at a time when several public builders have recorded walkaway charges on abandoned parcels.

  • But “because it hasn’t happened, doesn’t mean it won’t happen”.

  • Part of their underwriting is: who else could take this if the counterparty walks?

  • Emphasis throughout the call on scale, datasets, and disciplined underwriting as competitive advantages.

• “Demand for what we do has never been higher….We believe this is more than a cyclical response to today’s market. It reflects a structural evolution in how builders think about capital allocation.”

• Commentary suggests they are considering raising the leverage cap based on the strong performance of the operating platform.

  • “We know that if we pause our purchases, we’ll be able to generate cash rather quickly to pay down debt.”

  • Believe they are in a good position to be able to argue for investment grade ratings from the agencies.

• Industry inventory is normalising which is good news, cyclically.

Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Tag for finding my other articles on this stock: MSFT

Key takeaways

• 17% constant currency revenue growth, with operating leverage despite pressure on gross margins from the infrastructure buildout. Ebit grew 18%.

• Cloud and Azure grew 31% and 43% respectively, and they’re now roughly two-thirds and one-third of Microsoft’s overall revenue, so overall revenue may accelerate as they increasingly dominate the mix.

• Copilot paid seats now 30m, up 50% q/q. This is perhaps the simplest single measure of whether Microsoft is succeeding as the distributor of commoditised intelligence to enterprise.

Subscribe now

Additional thoughts

• Excellent results, especially given the More Personal Computing segment is struggling on weak PC sales and Xbox results.

• In Part 2 of Can Microsoft Compete? (see links above) I discuss my view that intelligence will commoditise, with multiple models competing for workloads behind the frontier. I think the evidence for this is hardening. Microsoft is building a model-agnostic harness that will automatically route demand to the cheapest model for each particular use case, and will protect an enterprise’s knowledge and data in the process. That’s obviously more appealing for a customer than being locked into one frontier LLM provider that might not remain at the frontier, will always want higher pricing, and might use customer data for its own ends.

• The change in depreciation schedules is noise. Microsoft have extended useful lives for data centre and office buildings from 15 years to 25 years. They mostly lease these buildings. When a lease accounts for a large portion of an asset’s life it is a finance life. When it does not, it is an operating lease. This change therefore shifts building leases from finance to operating. This does not change actual cash flows one iota, but operating costs will rise marginally and financing costs will fall marginally. In addition, finance leases are considered capex and operating leases are not, the logic being that if you lease a building for most of its useful life you are effectively borrowing money to buy the asset. Therefore, as buildings shift from finance to operating leases, reported capex will fall. Some analysts are suggesting this is a red flag, implying that Microsoft chose to change depreciation schedules in order to massage its capex number down. I disagree. A 15 year asset life is too short for buildings and I have never understood it. 25 years is more reasonable, especially in light of Andy Jassy’s disclosure that Amazon expect to be monetise their data centre buildings over “30 plus” years. If anything, the implication is that Microsoft has overstated past capex.

• ROI on capex. I see a lot of negative commentary on this. Many investors seem to have a huge issue with the fact that we do not know that the ROI will be strong. But when do you ever know? All growth projects involve risk. I think the risks here are manageable for a number of reasons. AI is a transformational technology and I think underestimating long term demand is easier than overestimating it. Certainly demand is well ahead of supply today - this is clear from the comments of all the hyperscalers, the fact that Azure revenue accelerates whenever Microsoft has capacity to meet it, and spot pricing being well ahead of contract pricing. There are also big braking factors on supply, including power and the capacity to manufacture chips. As with all major buildouts, capacity may exceed demand occasionally, and when it does the bears will dance a little jig, but it is fairly easy to slow capex and I expect demand to keep growing, so I would not expect those periods to last long. There are concerns over the asset lives of GPUs and CPUs, but these are mainly driven by obsolescence risk, and chip depreciation schedules are far shorter than time until burnout, so significant obsolescence risk is already baked into the accounting. (In fact the real risk may be to the upside, if chip lives can be extended by disaggregating prefill and inference - Gavin Baker discusses this on a recent Invest Like the Best podcast.) Finally, as I discuss in my review, the hyperscalers are effectively building a new capex-heavy business alongside their existing and growing capital light one. If infrastructure commoditises and there is no ROI, then they don’t need to own it - they will have created a new, separate, and huge utility industry focused on delivering cheap compute. That’s not a bad outcome for the capital-light orchestration and agentic layers, which will enjoy massive TAM expansion as intelligence gets cheaper.

• In short I think this was a strongly thesis-confirming quarter and Microsoft remains good value, trading at 23-24x P/E for a company likely to grow in the teens with deep moats.

Share

Notes

• FY revenue $331bn, +18%, with cloud $214bn + 27% and Azure $100bn +41%. Ebit +21%.

• 4q revenue +17% c/c, ebit +18%. EPS +23% - this excludes the mark-to-market for OpenAI, but EBIT better represents underlying growth.

  • CFOPS +30% to $55bn.

  • $41bn capex split 1/3 long term and 2/3 short term.

  • $20bn FCF on a cash basis i.e. excluding finance leases from capex.

  • Commercial RPO +84%, with all q/q growth coming from non-frontier customers.

• Extending useful life of data center and office buildings from 15 to 25 years, “reflecting our operating history and expected use of these assets”. This has a minimal impact on 2027 ebit but reduces reported capex by $15bn from $190bn to $175bn without actually impacting money spent.

• Productivity and Business Processes revenue +14%, ebit +14%, with margins depressed by investments in Copilot.

  • Paid M365 Commercial seats grew 6% year-over-year, with ARPU growth from Copilot and premium tiers.

  • M365 consumer cloud +22%, 7% seats and rest ARPU.

  • LinkedIn +10% in constant currency.

  • >30m paid Copilot seats, with net seat adds more than doubling q/q and the number of customers with >50k seats up 7x y/y.

  • Copilot Cowork is now generally available and they have introduced Autopilots: autonomous, long-running agents with full enterprise compliance.

  • A Copilot superapp is coming, bringing chat, Cowork, Autopilot, and Code into one place, alongside skills for the other applications in the enterprise like the CRM and ERP, giving an enterprise-wide AI tool that is pre-wired into existing IT Ops and Security setups - this will “change what people think of M365 capabilities” and massively expand the TAM.

  • Copilot performance is improving, with satisfaction rates 2x over 3 quarters, conversations per user up 2x y/y, weekly engagement on par with Outlook and Teams, and the time from deployment 80% MAU across a customers’ user base down from months to days.

  • New E7 suite brings together Copilot, E5, Entra, and Agent 365 and has sold millions of seats in 2 months since launch.

  • Seat + consumption models expand TAM and are driving “significant revenue”.

• Intelligent Cloud revenue +31% (Azure +43%), ebit +31%, with gross margins down on the mix shift to Azure, the infrastructure buildout, and higher Copilot usage, partially offset by strong operating leverage.

  • Revenue acceleration driven by efficiency gains across the CPU and GPU fleet, earlier delivery of new capacity which was quickly monetized, and GitHub Copilot’s shift to consumption-based pricing (which also started improving gross margin on this product).

  • Azure has “an incredibly diverse book of business by geo, by segment, by industry” plus a big first party app business which allows them to manage a slowdown in demand.

  • Customers building with multiple model providers up 5x y/y. Microsoft is building “a new model system where the harness, context, memory, and action space are separate from any one model family”, reducing cost per outcome and improving business continuity and resilience because every model is substitutable.

  • “Every firm is going to evaluate who are the providers who are helping them with their outcomes and their knowledge creation… This is not going to be about come in and take all my knowledge and benefit yourself…[so] you’ve got to keep your harness separate from the model…the harness will ensure that your memory, your context, all of that is external. That means any given model at any given time is swappable… We ourselves are using [this architecture]. Copilot is built that way. GitHub Copilot is built that way. Our Security Copilot is built that way. And we want to democratize that design pattern so that every enterprise can use it.”

  • >12 new proprietary models this year across image, voice, transcription, coding, security, and reasoning all targeting cost-efficient inference for enterprise use cases and all designed for proprietary silicon, giving 40% better performance per watt when running MAI models on Maia 200. E.g. in Excel, MAI-Code-1-Flash is delivering comparable quality to GPT-5.6 for the most common task while operating at significantly lower costs and in security, MAI-Cyber-1-Flash achieves better performance than Mythos at half the cost.

  • “The data estate is evolving from primarily supporting apps used by people to supporting agents”. AI-optimized databases like Cosmos DB and PostgreSQL give agents fast, secure access to real-time data and context. PostgreSQL revenue +55% and accelerating, paid Fabric customers +60%, Foundry revenue more than doubled.

  • Azure demand continues to exceed supply. If this changes they can slow the purchases of shorter-lived assets like GPUs and CPUs immediately. These are now the majority of capex and the driver of COGs.

  • On track to 2x compute in 2 years, and efficiency rising, with Copilot throughput up 4x. Maia 200 scaling and delivers 30% better performance per dollar than the latest generation hardware in the fleet. Cobalt CPUs also scaling.

  • Component pricing is spiking, hurting capex, but efficiency helps offset this and on-prem pricing is rising even faster, so cloud is more competitive. Contract pricing reflects input pricing, so there is a direct transfer to revenue growth, and I infer that when component pricing comes back down there might be margin and FCF improvements.

• More Personal Computing revenue -5% on lower PC market demand on rising component and device prices against a high prior-year comp for Windows, and weak revenue for XBox, offset by growth in Search.

  • Ebit decreased 15% including impairment charges in Xbox.

  • Xbox and gaming: aim to return to growth in 2027.

  • Windows: making it the best place to run secure edge AI. “We see significant opportunity for Windows to become the offload for unmetered intelligence, combining powerful on-device compute with enterprise-grade security.”

Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

This email is a summary of my activity this month.

June was quieter than May! I advise startups as a sideline, and was busy with that this month. Highlight for me was my review of Helios Fairfax, which I loved researching.

Trades

• I struggle to admit this, but I bought a tiny tracker in SpaceX. I can’t stand the valuation, but it is a number of companies pursuing transformational technologies that I think it is important to track and understand.

  Subscribe now

Articles I wrote

• Helios Fairfax Partners review: African growth companies at a 50% discount with a free option on top.

What I found interesting this month

• on Microsoft’s valuation.

• on Hikari Tsushin.

• on insurance brokers.

  Subscribe now

Thanks for reading - and please get in touch if you have questions.

Pete

What it does: invests and manages third party investments in Africa.

Elevator pitch: HFP trades at a 50% discount to tangible book value. This is too cheap for a diversified collection of interesting African opportunities plus exposure to the fee streams from managing third party capital. Exposure to the emerging markets fundraising cycle, which is at a 10-year low, is an added bonus.

Mental model: value, potential scaler (read about my mental models here).

Valuation and potential returns: 50% of TBV allows for potential returns of 15-30%.

Exchange and ticker: TSX, HFPC.U

Stock price and market cap: $1.70, $190m.

Do I own it? Yes.

IR website: here.

Tag for finding my other articles on this stock: HFP

About this blog: I have been investing for 25 years, professionally and personally. I look for stocks that have a high probability of compounding at 15% for at least 5 years with limited downside. I write these stocks up on my blog. You can find more about me, my philosophy, my mental models, and my portfolio structure on my site.

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Introduction

Helios Fairfax Partners (HFP) is the product of the merger of Helios and Fairfax Africa Holdings - one a leading alternative asset manager, the other a permanent capital vehicle, both focused on Africa. The company combines alternative asset manager fee earnings and carry with permanent capital for investing in some of the youngest, fastest-growing, and capital-constrained economies on earth. As a result of liquidating legacy investments and a downcycle in emerging market fundraising, the stock screens poorly and trades at less than half of book value. But book value has now grown for 5 quarters in a row, newer investments seem to be performing, and the asset manager fee stream may be turning a corner. The company aims to compound book value at 15% and pay asset management profits out as dividends. If it can do this, the returns from today’s discounted price will be excellent.

Subscribe now

History

In 2004, Tope Lawani and Babatunde Soyoye founded Helios as an African private equity fund manager. Helios raised its first African PE fund in 2006, and followed this with larger funds in 2009 and 2015.

In 2017, Fairfax Financial Holdings (FFH) launched a permanent capital vehicle for investing in Africa called Fairfax Africa Holdings (FAH). It didn’t work and by 2020 FAH was looking for strategic alternatives. They landed on the idea of merging with Helios. Helios got a permanent capital vehicle and FAH got a share of Helios’ economics.

This looked great on paper but the period from 2020-2024 was a nightmare. Helios liquidated the old FAH portfolio as best they could, but took losses along the way. In addition, covid and its consequences (inflation and rising rates) made it hard to exit investments in legacy Helios funds and raise funds for new ones, meaning significant amounts of carry was lost and management fees went down. Book value per share (BVPS) fell for 7 years in a row, so HFP screens horribly.

However, under the surface, progress was made. By 2024/5 pretty much all of the legacy portfolio had been sold and reinvested in much more promising businesses. And while emerging markets fundraising is still in a downcycle, that cycle will turn, and investments in the fundraising team and in diversifying the product offering show early signs of bearing fruit.

Share

Alignment

FFH owns 30m multiple voting shares and 7.3m subordinate voting shares. That gives them a 35% economic interest and a 53% voting interest. Principal Holdco, which is owned by Lawani and Soyoye, owns 25.5m multiple voters and 24.6m subordinate voters representing a 46% voting and economic interest. The remaining subordinate voters trade on the TSX. These are what we can buy; they represent 19% of the economic interest and virtually no voting power.

Some will view the lack of voting power as a red flag. I don’t, for at least three reasons.

1. Having controlling shareholders focused on the long term prevents shareholders doing something value-destructive for short term gains (such as winding up the company when it trades at a big discount).

2. I think having management with skin in the game is more important than having voting control - especially in alternative asset management. Lawani and Soyoye both have the substantial majority of their net worth in HFP stock, per AGM Q&A.

3. Fairfax has voting control, which creates a balance of power.

Share

Why Africa

Africa has incredible potential, and a reputation for squandering it. This reputation is deserved, but progress is being made. African countries account for 12 of the 20 fastest growing countries in the world in 2025-6. Over the next few decades Africa should be transformed. The investment case is underpinned by robust growth, a rising working-age population, rapid urbanisation, and improving business environments, despite enduring infrastructure and governance gaps.

HFP targets two megatrends. The first is demographics and urbanisation. Africa is staggeringly young, with a median age of about 20. Its working-age population is growing around 3% a year: by 2035 it will be bigger than China’s; by 2050 it will have doubled; and in 2100 Africa will be the only region on earth that has a bigger working-age population than today. In addition, Africa is urbanising. In 2009 Africa had 52 cities with 1m people or more; today, it has over 100; by 2050 it will have 160. The total urban population will double by 2050. This matters economically: urban populations are dramatically more productive than rural ones. Every urban resident needs housing, transport, healthcare, education, food retail, financial services, and connectivity. Urban consumption is also fundamentally different from rural subsistence: it is monetised, formal, and financeable. Africa is creating a large consumer class in economies that remain extraordinarily underserved.

HFP’s second megatrend is technology and innovation. Africa is poor and suffers chronic infrastructure deficits. As a result, customer value is low and the cost to serve is high. In addition Africa has cheap labour but a very high cost of production when you take skills gaps and logistics costs into account. Technology and innovation help solve these problems. Africa is leapfrogging expensive legacy technologies - mobile phones and internet rather than landlines and broadband, household solar rather than vast power generation and transmission networks, e-banking rather than inefficient branches, online skills development rather than poor government schools. Mobile money, pioneered in Kenya with M-PESA and now ubiquitous across East and West Africa, has given hundreds of millions of African adults access to formal financial services for the first time with just a basic mobile phone. And if exports are increasingly digital, Africa’s lack of physical infrastructure will be less important. Helios: “the good thing is that [in just the last 7-10 years] more and more young people are getting into technology and building platforms and coming up with great ideas.” Technology is both driving growth by solving inefficiencies and creating great investment opportunities.

The pushback is governance. Africa is poorly governed. Institutions are weak, economies are overregulated, currencies are unstable. But remember:

1. Africa is diverse. There are 54 countries, and some are much better than others.

2. Helios invest in companies, not countries. Companies with reliable counterparties, hard currency revenues, and exposure to multiple countries are more resilient than others.

3. Things are improving. This is important: money is made when things go from bad to ok. What are the signs of progress? Here are two: a growing number of countries in Africa have regular elections and peaceful transitions of power, and the World Bank says that Sub-Saharan Africa undertook more business environment reforms than any other region from 2010-2020. (Rwanda rose from near the bottom of the Ease of Doing Business ranking to the top 30 over 15 years via a systematic programme of regulatory simplification, e-government, and anti-corruption enforcement, and a landmark long term initiative is the African Continental Free Trade Area, which aims to eliminate intra-Africa tariffs on most goods and services and eventually create a $3.4tn African single market.)

4. One of Helios’ most successful investments was a telecom tower business in one of Africa’s worst-governed countries, the Democratic Republic of the Congo.

I do not count Africa’s abundance of natural resources as an advantage. In areas with poor governance, natural resources are a curse, encouraging corruption and driving boom-bust cycles. One possible exception is agriculture, which is labour intensive and hard to monopolise, so compared with (say) mining it is better for people and less tempting for corrupt politicians. Africa is blessed with an abundance of farmland and cheap labour. Slowly, slowly it will build the infrastructure to exploit these advantages.

Because Africa’s GDP is tiny and its governance is poor, it receives very little investment. Even compared to other emerging markets there are massive funding gaps for venture capital and private equity. As a result, valuations are low. This will slowly change. As GDP rises, inflation comes under control, currencies stabilise, and local savings grow, financial markets will deepen. This trend has been clear in Latin America over the last 40 years and I think it will happen in Africa over the next 40.

The point of this section is not to persuade you that Africa is the next China. It isn’t. But it has compounded GDP at 4-5% since the late 1990s and I see no reason why it cannot keep that up. If it can, then the living conditions of 2 billion people will be absolutely transformed. Africa’s GDP per capita is $2,000, or $7,000 in purchasing power parity terms. That is roughly the threshold between subsistence and the emerging middle class. In other words, the next decade or two of growth will see more and more Africans able to buy aspirational goods, pay for healthcare, save for the future, and spend on experiences. Nascent markets will become vast and significant companies will be built.

Subscribe now

Balance sheet

HFP’s balance sheet splits into 3 broad categories:

• Investments in the Helios funds.

• Co-investments alongside the Helios funds.

• Helios, the asset manager, which is now consolidated on the balance sheet (appearing in various line items) but appears in the investment list as TopCo.

Share

Investments in funds

Before we dive in, a word on valuations. The funds are carried at NAV. There isn’t much disclosure on the valuation of the assets inside the funds, which is irritating. We do know that several assets are valued using comparable market multiples, which is arguably less open to manipulation than a DCF but introduces some volatility. Two things give me some confidence: several of the funds (or their successors) are raising capital, which suggests third party investors find the valuations palatable; and where valuations are disclosed, which is mainly for the co-investments, they seem reasonable.

Helios IV and V are Helios’ latest private equity funds. The LP (Limited Partner) interests arise because HFP seeds the funds to help attract third party capital. The SLP (Special Limited Partner) interests arise because, like most managers, Helios makes a GP commitment to its funds which it calls the Management Team Commitment. HFP funds 50% of this commitment in return for 50% of any carry earned. Unless FFH agrees an exception, HFP’s SLP contributions are capped at the lesser of $7.5m per fund or 2% of aggregate fund. Neither the LP nor SLP Interests incur management fees or carry.

Substantially all of HFP’s PE exposure ($142m) is to Helios IV, which will mature over the next few years. Its IRR through December 2025 was 23%. That is unrealised, but promising: HFP’s investment has doubled on the back of strong investee company performance. HFP lists 7 private companies held by Helios IV in payments, insurance, discount grocery retail, healthcare, and data centres.

Helios CLEAR (Climate, Energy Access & Resilience) targets climate adaptation and mitigation investments in Africa. It has raised $250m and is seeking more. It made its first investment in 2025. HFP’s investment in CLEAR is small, but CLEAR is significant because it represents a new fund family.

Helios Sport and Entertainment Group (HSEG) is a private permanent capital vehicle which HFP seeded by contributing cash, loans, and its NBA Africa stake. HFP’s current exposure is $18m in loans and $111m equity, which is up from $87m at cost. In 1q26 HSEG achieved a first $30m close on a $75m Series B round, of which $28m was third party capital. Africa has vast talent and global success in sport, music, and content generally. HSEG invests in sports rights, content, and ecosystem enablers like NBA Africa, venue management companies, and talent agencies. All of these attract large scale, blue-chip capital in other geographies - but not in Africa, where HSEG has something of a first mover advantage.

HSEG’s marquee asset is its holding in NBA Africa. Africa has exported players to the NBA for decades, and as a result basketball is the second most popular sport on the continent (after football) despite approximately zero investment. Interestingly, and in contrast to football, in some countries 40% of the fan base is female. Historically sport was difficult to monetise in Africa, where most people can’t access or afford all-you-can eat cable subscriptions. However with smartphones content can be monetised in new ways - for example pay per view, highlights only, or with discounts for watching a day late. I really like the fact that HFP can make potentially multi-decade investments like this because it has permanent capital. NBA Africa was not prepared to transact with a time-limited fund.

HSEG’s other investments are the African part of the Professional Fighters League, which is the second largest MMA business globally after UFC; Zaria, which was cofounded by Masai Ujiri and builds marquee sports and entertainment arenas in mixed use developments such as Zaria Court in Kigali and Nairobi’s Railway City (overview at 16:30 but the whole video is interesting); and The Malachite Group, which organises music festivals and events and manages African talent. Several of these investments are loans with equity-like upside.

Helios Digital Ventures (HDV) targets early stage tech companies with potentially exponential outcomes. HDV aims to invest $5-20m in rounds of $20-50m to build pan-African tech businesses of the kind that have emerged in Asia and Latam. It has invested in payments, fintech, biotech, and SAAS. HFP’s stake started as an accruing loan. At yearend 2025 this was converted into an LP interest and in 1q26, its first quarter as equity, the stake was marked up 4% to $47m.

Seven Rivers was a hedge fund investing in public equity and credit in Africa. Its performance seemed strong - 50% in 2025 and another 30% in 1q26 - but the decision was taken to close it in April 2026. The $48m stake will therefore be liquidated and added to the $14m in cash on hand.

Share Building Arks

Co-investments

HFP invests in individual companies when the holding period is likely to be very long and/or when it wants additional exposure to a specific idea. Current direct holdings are:

• $37m in Trone, which distributes and maintains medical imaging and diagnostic equipment and produces and distributes contrast pharmaceuticals for imaging in Morocco and Francophone Africa. HFP has additional indirect exposure through Helios IV - between them, Helios entities and the management team own the whole company. HFP is up 140% on its investment, which is carried at 9.4x ebitda.

• $14m in Taj Holdings, which owns HFP’s direct stake in M2P, an Indian infrastructure API and Banking-as-a-Service provider rapidly expanding across Africa. HFP has additional indirect exposure through Helios IV. The stake is valued at 6.8x revenue and has been marked down from $16m at cost on lower market multiples and lower forecast revenue. HFP has a liquidation preference.

• $9m in Conduit, which is building stablecoin-based B2B cross-border payment infrastructure for emerging markets in Latin America and Africa. HFP has additional exposure though HDV, which invested in 2023, 2024, and again in May 2025 when it participated in a $36m Series A led by Dragonfly and Altos Ventures. HFP then added its direct stake in November 2025 (although NB I think there are multiple share classes with different exposures). The latest operating data I have is that transaction volumes grew 16x in 2024 to an annualised rate of $10bn.

• Pending: up to $75m in CAB Payments. Over the last few months there has been a tussle for control of this company, a London-listed cross-border payments and banking provider. Helios IV has owned 45% since before the IPO. Following share price weakness HFP and Helios V have jointly offered to buy the remainder. StoneX have bid a similar price, which goes some way to validating the valuation, but abandoned their offer because Helios IV did not support it. The bid has over 50% voting support, but the board has rejected it as too cheap. If it closes, it will obviously be significant for HFP.

Subscribe now

Helios

Asset management is a great business model - at scale. The infrastructure required to gather and invest capital, once built, can generally support growth without too much additional cost. This operating leverage drives high margins and economies of scale. Unfortunately Helios is subscale - but that’s why you don’t have to pay for it. It is, in effect, an option.

Helios raises capital, invests it, and charges management and carry (performance fees). For its core funds, the management fee is typically 1.5-2% and the carry is 20% of profits over an 8% compounded hurdle. HFP, through TopCo, gets:

• Any excess of management fee revenues over operating costs - i.e., any profits (although Helios is currently lossmaking).

• 50% of any carry (the remaining 50% goes to the Helios team, which encourages long term thinking and employee retention since it usually takes years to earn carry).

On paper HFP pays Helios a fee for investing its balance sheet, but this is eliminated on consolidation. What shows up on the HFP income statement is just third party fees less the cost of running Helios. This is actually a good representation of the real economics: the fee is irrelevant; what matters is that HFP subsidises Helios’ current losses in return for Helios’ possible future profits. HFP’s subsidy covers both the cost of sourcing investments for the balance sheet and the cost of maintaining the equity option in Helios. Based on the 1q26 P&L, the annual cost is 2.3% of HFP’s tangible book value (TBV). Given the value of the option I think this is a good deal: Helios will be very valuable if it can raise more third party capital and turn consistently profitable.

Can it?

Helios was founded in 2004 and raised a succession of funds in its core PE strategy: Helios I at $300m in 2006, Helios II at $900m in 2009, and Helios III at $1.1bn in 2014. Performance was decent, and by 2019 Helios anticipated receiving significant amounts of carry. However after 2015 the emerging markets fundraising cycle turned down, and then covid hit, depressing business activity, hurting valuations, and slowing exits. This impacted Helios in two ways. First, the anticipated carry evaporated as valuations came down, time dragged on, and the 8% hurdle compounded. And second, the fundraising environment became almost impossible so Helios IV, raised in 2020 at $355m, was only one-third the size of Helios III. What looked like a steady fee grower with carry on top hit a nasty bump in the road.

That bump may now be in the rear-view mirror. Helios IV’s performance has been excellent so far. That’s good for carry and great for fundraising. Helios V’s first close was $338m, almost as big as the whole of Helios IV. Its target is $750m, which would be a significant step back towards the scale of funds II and III. In addition Helios has attracted third party capital into two new strategies: CLEAR is hopefully the first of a new fund family and has raised $250m out of a targeted $400m, while HSEG is a permanent capital vehicle which has raised $30m out of a targeted initial $70m. This is all positive for fee-paying assets. What does it mean for profits?

In 1q26, HFP consolidated Helios. Previously, it carried Helios (TopCo) on the balance sheet at DCF fair value and changes in fair value went through the P&L. There was almost no disclosure of Helios’ actual economics. Now, Helios’ fee streams and costs are reported on the P&L, making analysis much easier. We only have figures for 1q26 when management fees were £5.9m, consulting fees $2.9m, and G&A $10.8m. This nets to fee-related earnings of -$2m in the quarter and -$8m annualised. I think these losses will improve over time, but not necessarily in a straight line. Helios funds earn fees over committed capital. This means each fund generates flat fees for several years and then declining fees as investments get sold. Helios II and III stopped paying fees in 1q26, which could lead to a drop in fees in 2q26. Helios IV, V and CLEAR are generating fees, but IV will start selling investments over the next few years so V and CLEAR are the growth engines. In short, timing differences could drive quarterly volatility in fee related earnings but continued fundraisings at V and CLEAR should drive growth.

Helios’ other important profit stream is carry. The outlook here is also promising. Helios’ performance record is respectable. In total, funds I, II, and III raised $2.3bn. They have so far generated $4.7bn via exits (2026 AGM transcript) but little carry - carry is calculated over an 8% compounding hurdle so it is very sensitive to exit timing, and covid caused delays. Helios IV, however, has produced an unrealised IRR of 23% so far, and HFP’s share of unrealised carry stood at $22m as of yearend 2025.

(While we are on the topic of performance, two points are worthy of note. First, Helios uses less leverage than US PE: in 2023, >20% of all US PE deals had >10x debt-to-EBITDA and ~60% had >6x. For Helios, leverage is typically less than 30% of the capital in a deal. Second, Helios’ exits are high quality. Most of the $4.7bn of exits from Helios I, II and III to date were IPOs or sales to strategic buyers, not pass-the-parcel amongst financial owners.)

There are two other things that could help Helios grow:

• A turn in the emerging markets fundraising cycle. EM was “hot” through the late 2000’s and early 2010s, on the back of rapid GDP growth, strong currencies, and high commodity prices, all of which tend to be correlated in EM, plus a lack of investor enthusiasm for investing in the developed world after the GFC and the Eurozone crisis. This dynamic aggressively reversed after 2015 and EM funds have found it extraordinarily hard to raise money for a decade. But this is a cycle, and it will turn. I actively seek exposure to this.

• Successfully launching a fourth major strategy (after PE, CLEAR, and HSEG). Seven Rivers had good performance but did not attract third party capital and has been closed. Helios Energy Transition Infrastructure (HETI) was going to be a public vehicle investing in long dated, dollar + inflation infrastructure assets, but was shelved in 2024. Either could be resurrected if conditions change. Digital Ventures is nascent but ongoing. Helios DataSphere is in development as a pan-African data-centre development and operating platform. HFP’s balance sheet is an advantage here because Helios can seed funds without third party capital. It is far too early to ascribe value to the fees new strategies might generate, but it is positive that Helios is trying to diversify its fee streams.

Putting this together, it is quite possible that by around 2030 Helios could have:

1. Two recurring fund families (PE and CLEAR) capable of raising funds in the +/- $1bn range.

2. A growing permanent capital vehicle in HSEG which I guess might have $2-300m of third party capital.

3. Small but growing third party assets in a fourth major strategy.

4. Realised carry income from Helios IV.

5. Unrealised carry building up from Helios V and CLEAR I.

6. An EM fundraising cycle turning in its favour.

Almost any combination of these things would make Helios a very different animal - bigger, more profitable, and more diversified than it has ever been.

Share

Other balance sheet notes

• Total liabilities are 23% of tangible assets and borrowings are 7% of tangible assets.

• HFP has $33m in debt including a small amount of fund level leverage which is secured on fund assets but ultimately is recourse to HFP.

• HFP has drawn $10m of a $100m facility maturing Feb 2031, at SOFR + 6%. Including $14m in cash and $48m in Seven Rivers which is being liquidated, HFP has significant liquidity.

• Borrowings would be 22% of tangible assets if HFP maxed out its loan facility and invested the proceeds in tangible assets.

• HFP has unrecognised deferred tax assets ($41 million at year-end 2025), mostly from net capital losses and investment differences.

• Limited partner distributions payable is money owed to fund II and III LPs as they wind down.

• 100% of the SLP Interests is consolidated onto balance sheet and the 50% that is funded by management appears as an offsetting liability (amounts attributable to other limited partners).

• Until 1q26, HFP carried its interest in Helios at fair value using a DCF. In 1q26, HFP determined that it should consolidate Helios. No consideration was exchanged but for accounting purposes HFP “bought” Helios at the yearend 2025 carrying value. The DCF inputs for this were disclosed. For management fees, they were 9% growth for 7 years, 4.5% thereafter, 34% pretax profit margins, and a 17% discount rate. For carry, they were exit multiples of 1.5-3.7x, exit dates from 2026-2031, and discount rates of 24-28%. This transaction created $52m of identified intangibles (the present value of expected profits from existing management contracts, of which $2m was amortised in 1q26) and $60m of goodwill (the present value of expected profits from future management contracts, which is subject to impairment testing).

Subscribe now

Valuation and expected returns

The share price is bouncing around $1.70. This is a 50% discount to TBVPS of $3.26. TBV is made up of holdings in the Helios funds, direct holdings in African companies, and cash. It does not include anything for Helios, the asset manager. What little disclosure we have suggests carrying values are reasonable, as does third party investor validation. HFP’s balance sheet is only lightly levered, limiting downside. On a balance of probabilities, I am prepared to trust that HFP could realise tangible book value in an orderly liquidation and that TBV will grow as the investments perform.

The discount to BVPS ($4.27) is even greater at 60%. The difference between TBV and BV is the intangibles and goodwill created when HFP consolidated Helios (see above). While I think the assumptions underlying the “acquisition price” are aggressive, they are not insane, and it is interesting that $0.46 per share of identified intangibles represent the expected profits from existing contracts. If you include this the “adjusted TBVPS” is $3.72 and the discount is 54%.

With such large discounts, you don’t need heroic assumptions:

• If TBVPS can compound at 10% and the discount moves to 30% over 5 years, the stock could return 17% per year. This requires the underlying investments to do ok, but does not require Helios to work.

• If BVPS can compound at 15% and HFP can pay a dividend from Helios profits (which are the goals stated at the 2026 AGM), I would expect the stock to trade at 1x book value in 5 years and it could return 38% per year plus dividends.

Obviously worse scenarios are imaginable. Worse scenarios are always imaginable. But HFP is a diverse collection of carefully-chosen assets in a fast growing, capital starved part of the world. I think an outcome within the range of returns shown above is realistic.

Subscribe now

Conclusion

HFP is a diverse collection of seemingly decent investments trading at an extreme discount. It has optionality in Helios, which appears to be heading in the right direction, and also in the underlying investments, some of which could be home runs. It is run by motivated, experienced, and aligned managers. It offers exposure to a turn in the EM fundraising cycle, which I actively seek. It is illiquid, so if interest picks up the stock could move quickly. It is starting to screen better - book value has grown for the last 5 quarters, and its recent consolidation of Helios opens the door to industry-standard valuation metrics. It is one of the more speculative holdings in my portfolio, and one of the smaller ones. But I think it has the potential to deliver excellent returns.

Links to previous Reviews

1. IRSA

2. Brookfield

3. Cheniere

4. Uber

5. Howard Hughes Holdings

6. Millrose Properties

7. Microsoft Part 1, Part 2, Part 3.

8. My notes

Thanks for reading - if you enjoyed reading this please subscribe, like, and restack, and do get in touch if you have questions.

Pete

This email is a summary of my activity this month.

First, a big thank you. Building Arks has tripled its subscribers this month. Quite incredible, and very pleasing.

Trades

• Added to Exor, which is at a silly discount (as usual).

• Bought a tracker in Derwent London. There is value in the UK.

• Added a little to my largest position, Fairfax Financial. First time I have added since June, 2020.

• Added a little IRSA when it sold off.

• Bought a tracker in The Hong Kong Shanghai Hotel.

• Several reductions to Ensign Energy Services, Borr Drilling, and Transocean to add to Howard Hughes. Notes and reasoning here, here, here, here, and here. As you can tell I tend to average in and out!

• Reduced Grupo Aval and Grupo Cibest into the Colombian election. Not a smart move so far but I had good gains I wanted to protect.

  Subscribe now

Articles I wrote

• Brookfield results. Going through a slow patch.

• Millrose results. Fine…for now.

• Microsoft results. Copilot becoming a growth engine.

• Howard Hughes results. Strong MPC cash flows.

• IRSA results. Very promising as Argentina changes.

• Uber results. Thesis developing nicely.

• Cheniere results - strong, and a masterclass in how to execute a buyback.

• Can Microsoft Compete? Part 1, Part 2, Part 3. Tl;dr: yes, it has an exciting future as the distributor of commoditised intelligence to enterprise.

What I found interesting this month

• on Hokkaido Electric Power.

• on why SAAS companies have time to respond.

• on Kingsway, which I have my eye on.

• on AI harnesses as moats. Interesting, but I had a question.

• on Anthropic business adoption.

• on YPF. Highly relevant to Argentina and therefore IRSA.

• on Derwent London.

• on Apollo’s strong results and Howard Hughes.

• on Amazon’s robotics growth.

• on deep value in The Hong Kong Shanghai Hotel.

• on Japan’s collapsing car industry.

• on moats in the codebase.

• on Uber Freight. Is this an emerging growth driver?

• on Hermes. Persuasive piece on moats.

  Subscribe now

Thanks for reading - and please get in touch if you have questions.

Pete

Tag for finding my other articles on this stock: BN

Key takeaways

A slower growth quarter with acceleration expected through the year. Private credit worries are not systemic and Brookfield’s areas of exposure are fine. Momentum is accelerating in real estate. Merging BNT (insurance) back into BN gives the insurance operation a huge capital base. Both BN and BAM have bought back significant amounts of stock recently.

Subscribe now

Thesis and valuation update

No major change.

• Probably the single thing that stood out to me is that office construction costs have risen so significantly that the rents required to justify newbuilds are sometimes double current market rents, and demand is growing.

• 2026 will be a record year for fundraising.

• There is useful discussion of insurance capital and use of Bermuda reinsurers in the BN call.

• Most of Brookfield’s insurance business is life/annuity, but they have a small P&C insurer which they are increasingly talking about scaling.

• BEP have commented for several quarters now that battery costs have come down so much that they are economic for balancing grids an evening out renewable generation.

• I recommend reading letter too - good discussion about how little macro matters, and how they observe, test, and perfect businesses before scaling them.

Share

Notes (please note these are call summaries. I don’t dive into the detailed reporting every quarter).

BN

• DE $1.6bn, $6bn LTM, and will strengthen through the year. DEBR $0.59/share, up 7%, LTM $2.32/share.

  • Advanced $17bn of sales “substantially all” at or above marks.

  • Realised $157 of carry into income and have $11.8bn unrealised.

  • Think carry inflects in 2h as sales accelerate.

  • Partially monetised a tech investment at a gain. Growing relationships with tech companies are leading to interesting investment opportunities for the BN balance sheet and clients - they have $2bn in tech of which $1bn is SpaceX at its pre-IPO mark.

  • YTD have bought back $470m of BN shares plus $575m at BAM.

• Macro developments “often receive far, far more attention than their long-term impact warrants. Bottom line, we largely try to ignore them when building our business…Our role as investors is to capitalize on attractive entry points to acquire good businesses for value, operate them well and allow compounding to work over time…We take the time to watch an industry, learn how it works, invest in a measured way, refine a business model and only then scale a platform. This allows us to make small mistakes while avoiding large ones. In our experience, successful businesses are not built quickly. They are built deliberately.”

• Private credit: “there are issues that are grabbing a lot of headlines, but in the scale of the broader investment markets, their materiality is low [and] as it relates to Brookfield, these are very immaterial asset classes to us based on our deliberate posture…our credit portfolio is performing incredibly well.”

• Real estate.

  • Sentiment is now catching up with fundamentals. “Buyers looking for solid assets are moving back from software to real assets like these.”

  • In office, signed 2.6m ft2 of leases, 15% above the expiring levels.

  • Office replacement costs have risen significantly. Rents required to justify new construction are double current market rents in many markets. This makes new supply very difficult to deliver.

  • Manhattan West was started in 2020. It cost $1000/ft2; would be $2,500/ft2 now. Most recent lease was signed at nearly 3x the rent of the first lease, and is still not high enough to justify new construction. Recently refinanced this: $1.9bn 10-year nonrecourse mortgage, including $400m cash extracted, at 5.5% - a 107bp spread to treasury.

  • One Leadenhall was fully leased within 6 months of completion and achieved the highest rents ever in the city of London.

  • In retail, tenant consolidation into top-tier malls continues to drive demand. Commenced 1.6m ft2 of leases, 11% above prior levels.

• Brookfield Wealth Solutions

  • Insurance has $180bn in assets, $20bn of regulatory capital, and over $2bn of annualised earnings.

  • Origination

    • “Our priority is not maximizing volume but generating high-quality, durable earnings.” Goal is to compound equity capital at 15% with low risk and generate stable earnings.

    • Aging and the decline of defined benefit pensions creates a long runway for growth.

    • Annuity demand in the US is down 9-10% y/y but Brookfield has picked up 4pp of share.

    • Recently launched on 2 major bank platforms with more to come. Brookfield sell 1/3 of annuities through the bank channel vs. peers at 2/3.

    • Will write $25bn of new policies in 2026.

    • Outflows will be $10-12bn - average liability duration is 8-9 years.

    • Asia - early stages of building a presence in a very significant market, with growing interest from counterparties.

  • Spreads

    • LTM deployed $15bn into Brookfield strategies at an average total return exceeding 10%.

    • Spreads compressed slightly this q. Annuity rates peg to the back end of the curve but cash from newly sold annuities earns the short end until it can be deployed. When the yield curve steepens, therefore, spreads compress slightly. What matters in the long run, however, is the total return over the life of the annuity.

    • Brookfield are well positioned to invest globally for best total return. “We’re trying to build a business where at the top of the house, we can move our capital around to geographies and products, and we can do that without any conflicts or clients or other invested capital partners sitting in any parts of the business.”

  • Capital

    • Each policy-writing company is rated A or A- by the 3 major agencies, and 2 have received upgrades over the last few years.

    • Brookfield generally operate at about 4x the regulatory minimum capital requirement, not including the excess capital at BN.

    • UK is moving against use of Bermudan companies to reinsure pension risk transfer deals. No impact: Just doesn’t use Bermuda as a reinsurance jurisdiction, and anyway Bermudan capital rules align with the UK and Europe so repatriating reinsurance won’t affect competitors either. (My notes: the Bermuda regulatory regime is Solvency II equivalent and for US purposes Bermuda is a NAIC reciprocal jurisdiction. In 2024 Bermuda overhauled its rules to remain Solvency II equivalent, specifically targeting the life reinsurance and annuity block transfer markets where PE-backed reinsurers were taking advantage of Bermuda’s more lenient capital eligibility rules for private credit. Bermuda’s approach to assumed default and downgrade costs on private assets now reportedly results in higher costs than those applied to a reinsurer authorised by the UK PRA, so the competitive advantage of Bermuda domicile now rests primarily on the Pillar Two corporate tax rate of 15%).

    • The planned merger of BN and BWS creates a fully integrated insurance/investment operation and gives insurance a vast permanent capital base.

  • Closed acquisition of Just Group.

    • Leading pension risk transfer and individual annuity provider in the UK. Serves 700k UK pensioners, has $40m of assets. Going-in valuation gives a 10-12% return.

    • Brookfield can improve Just’s investing so that it can grow its £5bn of annual originations.

    • Just is very good at operating small pension schemes, where there is not much competition.

    • With BN’s capital and asset origination, they can also move into large deals, where there is also not much competition. (Competition is in the middle.)

  • Clearbrook (P&C) achieved a 99% CR. Important diversifier into specialist insurance. Have worked hard to exit some lines, derisk liabilities, and grow profitably. Expect opportunities to grow organically and via M&A as the P&C market softens.

Subscribe now

BAM

• FRE +11%, DE 7%.

  • FBC +12%.

  • LTM FRE 18%.

  • Sold $8bn, invested $34bn,

  • Will exceed long term growth targets in 2026.

  • Have bought back $800m over the last 7 months.

  • After the q, they issued $550m of 5y notes at 4.83% and $450m of 10y notes at 5.3%.

• 2026 will be a record fundraising year by a significant amount.

  • Including Just Group and flagship PE, YTD fundraising is $67bn, over half of total 2025 fundraising.

  • Each of Primary Wave, 17Capital and Pinegrove recently recently closed funds that were their largest ever and the largest of their kind.

• Oaktree closes in 2q. Benefits: easier to create customised solutions combining both companies’ products, and potential to optimise 2 balance sheets.

• “One of the clearest ways our platform is evolving is in how we engage with our largest clients.”

  • Investors are consolidating more of their business with fewer managers that can invest at scale across asset classes and geographies and up and down the capital structure.

  • Conversations are leading to broad strategic relationships and customized solutions using insights from across the Brookfield ecosystem.

• Real estate recovery is accelerating.

  • “What we’re seeing on the ground is far ahead of what you’re reading in the headlines”.

  • Significant increases in transaction activity and valuations. Primarily hospitality, logistics, housing, so far - less in office and retail but that will follow - “the fundamentals for office are absolutely flying…in Tier 1 markets, we’re seeing [rents] 50%, 70%, 80% higher than they were 5 years ago”.

• Credit

  • Spreads post-covid were excellent. That attracted capital and spreads compressed. They didn’t maximise growth in this period and “it’s important to separate the fundamentals of private credit from the excesses in select parts of direct lending.”

  • “We have always preferred areas where underwriting matters, where structure matters and where there is real downside protection, notably real asset credit, asset-backed finance and opportunistic credit”.

  • If there is a broader dislocation, Oaktree benefits.

    • “When liquidity becomes scarce and capital is repriced, that is when disciplined investors with flexible capital and deep experience have historically generated some of their best returns.”

    • “We are already tracking dozens of emerging opportunities in real time”.

    • Could deploy tens of billions in a proper credit dislocation but “today, we don’t see a broad-based macro condition that would result in meaningfully higher deployment patterns than what we’ve seen over the last 5 years. But we always see sector-specific distress. Today, we see distress in software, building products, chemicals, autos, packaging.”

• AI: “While there is a significant amount of capital flowing into the sector, the investment opportunity set is incredibly vast. And as a result of that, we can be incredibly selective. We can focus on the best assets in the best markets with the best revenue constructs and the best corporate credit counterparties.”

• Penetration of the individual wealth market is accelerating fast. With respect to 401ks, they are in advanced discussions with some of the largest target date fund providers to put real asset products into default portfolios.

Share

BIP

• FFOPU $0.90, up 10%.

  • Sold or agreed $1bn in sales.

  • Refinanced $1.5bn of nonrecourse debt.

  • Will exceed 10% FFOPU long term growth target this year.

• Assessing merging BIPC and BIP to form a single entity.

• New framework with a leading global OEM - exclusive long term leasing platform for industrial equipment, including for data centres, without residual value interest rate or refinancing risk. $375m (BIP share) over 24 months.

• Intel JV on schedule with first earnings in 3q26.

• Canadian midstream business seeing strong demand - completed $400m of growth projects over the last few months that are now ramping, and have $8bn of bite sized, straightforward, low-multiple growth projects ahead.

• Looking at an IPO for Csquare.

• Tremendous demand across datacentres, compute, fibre, grid stabilisation, behind-the-meter power.

Subscribe now

BBU

• Ebitda $582m, down on sales. LFL ebitda was up 5%. EFO $279m.

• Completed corporate simplification.

• Clarios is $15 per share in NAV at 9-10x ebitda and can 2x in 5y driven by advanced batteries (higher market share and higher margins), cash generation, and tax credits. Clarios received $1bn of tax credit cash this quarter, or $1.50 per share. Expect similar amounts annually through the decade.

• Sold 27% of La Trobe, an Australian asset manager and lender, at an implied 3x MM in 4y.

• Committed to lead a $500m investment ($150m BBU share) alongside OpenAI into the new OpenAI Deployment Company.

  • Primarily an advisory/services business which they know is needed from their own work.

  • Investment is a preferred with >15% upside CAGR but low downside.

  • Also gives the Brookfield ecosystem access to leading technology early.

• Sagen (Canadian residential mortgage insurer) has grown share, reduced expense ratio, and optimised capital efficiency under Brookfield.

  • ROE has gone from low double digits to over 20% and the business can distribute $400m per year over the cycle.

  • House prices are down 20% since early 2022.

  • 80% of the portfolio is fixed rate and most of the remainder have constant payments (so only the mix between interest and variable changes with rates).

  • Loans have full recourse, all insured borrowers in Canada are subject to a stress test that builds in a cushion for affordability in a rising rate environment, and insured borrowers facing financial hardship can extend amortizations.

  • Losses are therefore driven by unemployment (frequency) and home prices (severity). Both are manageable with resilient employment and significant homeowner equity.

  • Loss ratio has risen from 5% to 12% on severity, mainly on 2022/23 vintages with less equity. They price for long run loss ratios of 15-20%. Will be below that this year, but the last few years have been abnormal with strong employment and house price appreciation.

• BRK won a new concession in north east Brazil. Meaningful but will take time to ramp. Interest rates falling which may open an IPO window.

Subscribe now

BEP

• FFOPU +15% to $0.55. LTM FFOPU $2.08, up 12%.

  • Deployed or committed $2.2bn/$550m net.

  • Sold/agreed sale of $2.8bn/$800m net.

  • Closed $4bn of financing including CAD500m of 30-year notes at tightest spread ever.

  • Corporate level debt maturity now 14y.

  • Brought 1.8Gw online (LTM: 9Gw, up 2x in 2y) and contracted 1.7Gw from pipeline.

• Between strong demand and strong bids for mature assets, feel they can grow medium term FFOPU ahead of the 10% long term target.

  • Capital recycling is driven by value - if recycling mature assets generates more value than holding them, they sell. Bids are currently strong. Created Northview Energy in partnership with BCI, Norges Bank, and a Brookfield fund. BEP will sell mature, derisked renewables assets to Northview.

  • “Demand continues to go up. It is higher today than it was last quarter. It’s higher today than it was last year.”

• Good progress on the Westinghouse / US Govt deal including ordering long lead time items.

• Continue to see opportunities in the public markets where companies have projects but not capital. Bought Boralex in partnership with La Caisse. $6.5bn EV. Accretive on close, and can accelerate growth.

• Batteries

  • “Undoubtedly the fastest-growing technology across Brookfield Renewable today is batteries and energy storage. We are seeing that within all of our existing development platforms. We are increasingly looking at stand-alone energy storage opportunities. And the rationale for this is very simple. They remove grid congestion…and they are very quick to deploy.”

  • CapEx for batteries and energy storage has come down 65% over the last 2 years, “making these investments very economic”.

  • It is “absolutely, in no uncertain terms” economic to add batteries to existing renewables generation and offtakers are willing to pay a premium to firm up their power.

• Issued 2.8m BEPC shares to buy 2.8m BEP LP units for a net cash gain of $27m. Continue to explore whether a single structure is better.

Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Tag for finding my other articles on this stock: MRP

Key takeaways

Strong customer demand but no answer to fundamental questions around business model and returns. To really work this needs to trade above book value but there is no reason for it to do so. Current discount is 23%.

Subscribe now

Thesis and valuation update

No change to thesis. The risk is that this works until it doesn’t.

• Millrose clearly have strong customer demand but doing deals that are good for customers isn’t the same as doing deals that are good for shareholders.

• Housing demand isn’t great, so homebuilders are holding onto land they know they can sell relatively soon, and selling longer-dated land to Millrose with an option to buy it back. They pay Millrose 10.7% when they could finance it on their own balance sheet for 6% - they are either dumb or they think they are shifting significant risk to Millrose.

• Millrose pays out all of its option premium profits as dividends, and when it sells land it can only break even or lose money. It cannot retain profits but it does retain losses. Mathematically, I think book value has to fall over time.

• This is a problem, because Millrose is near its debt ceiling and can’t grow without either lifting the debt ceiling, issuing shares, or coming up with a funky new capital structure. They have said they will not issue shares below book value, but if book value can only shrink the shares are unlikely to trade at a premium. I would not be surprised to see Millrose lift its debt ceiling. This obviously increases risks.

• In fairness, a higher debt ceiling could also mean a higher dividend, a higher stock price, and accretive share issuance. That’s the bull case. But the fact is that Millrose’s flywheel only really works if the stock trades above 1x book. And if it does, Millrose will attract competition: their competitive advantages are replicable by any number of asset managers with access to capital.

• Finally: it would reduce risk if the housing market accelerates from here, because homebuilders will be more likely to exercise their options. But if it decelerates… ouch. And one day, it will decelerate.

Share

Notes

• Net income of $122.9 million, or $0.74 per share.

  • AFFO $126m, up 2.5% q/q adjusting for slightly fewer days.

  • 100% AFFO payout.

  • BVPS $35.26, down 2c.

  • Now 143k homesites across 904 communities in 30 states with 17 counterparties, up 2 q/q.

  • Assets now 69/31 Lennar/other. Non-Lennar deals average 10.7% vs an average cost of debt of roughly 6%. These options are typically floating subject to a fixed rate floor, currently around 10%. Millrose mainly fund these deals with floating debt.

  • Amended credit facility to unsecured and added $500m of capacity.

  • Debt/cap 29% vs 33% limit.

• Industry conditions:

  • Housing “demand is choppy, but not collapsing”.

  • Homebuilders are pulling back on starts; maintaining sales through incentives and inventory management; protecting balance sheets amid meaningful margin compression; preserving/growing land pipeline and community count; and restricting direct land ownership. The last two are mutually exclusive without a partner like Millrose.

  • “When builders are earning less on every home, the last thing they want is more capital tied up in land that won’t produce a closing for years, but they can’t afford to lose those future communities either.”

• Competitive advantages

  • “As a publicly traded company [with] a fully unsecured investment-grade…capital structure, Millrose offers counterparties a degree of capital certainty and transparency that private land banking alternatives are structurally unable to replicate.”

  • Scale = sophisticated underwriting, diversification of risks, and certainty of execution.

  • “The software and workflow complexity of managing nearly 144,000 homesites across 904 communities in real time is a barrier that cannot be replicated quickly or cheaply.”

  • Proprietary lot pricing dataset across 30 states “has become a genuine competitive advantage”.

  • Once they have a relationship with a homebuilder it tends to grow as they become operationally integrated with the land finance team and gain trust with the C-suite.

• Confusing commentary on growth

  • They have significant demand but they are close to debt ceiling and don’t have a plan to fund it.

  • They “haven’t yet turned our attention to alternative financing…structures” but they “spend a lot of time thinking…about what could next steps be”. Whatever that means.

  • Hope to trade at levels where they can “unlock the equity markets as a financing vehicle”.

  • “We’re not going to walk away from business, and we’re certainly not going to walk away from our existing clients”. They need capital.

Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

What it does: enterprise software.

Elevator pitch: arguably one of the finest franchises ever built, with deep distribution advantages that let it deliver innovations cheaply to a huge customer base.

Mental model: moat (read about my mental models here).

Valuation and potential returns: 21x June 2027 EPS estimates and growing EPS 20% per year.

Exchange and ticker: Nasdaq, MSFT

Stock price and market cap: $422, $3.1t.

Do I own it? Yes.

IR website: here.

Tag for finding my other articles on this stock: MSFT

About this blog: I have been investing for 25 years, professionally and personally. I look for stocks that have a high probability of compounding at 15% for at least 5 years with limited downside. I write these stocks up on my blog. You can find more about me, my philosophy, my mental models, and my portfolio structure on my site.

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Welcome to the final instalment in my Microsoft saga. The goal of this piece is to determine whether Microsoft can compete over the next 5-10 years. Part 1 introduced the layers of AI the AI stack and looked at two, silicon and infrastructure. Part 2 dived into two more, LLMs and the agentic stack. So far we have concluded that:

• In silicon, Microsoft is behind because it started late, but Maia 200 is promising and Microsoft will close the gap.

• Infrastructure will commoditise slowly, and a new utility compute industry will eventually allow hyperscalers to go (partway) back to being capital light.

• LLMs will commoditise behind the frontier, creating a rolling wave of commoditisation and increasingly cheap intelligence.

• Microsoft is well placed to 1) help enterprises develop and manage their own agents, and 2) develop proprietary agents that can be sold into large verticals. The agentic TAM is huge.

Part 3 starts with distribution and bundling, and then considers whether layers can be combined to create competitive advantage before drawing everything together in conclusion.

Subscribe now

Distribution and bundling

Distribution and bundling have been Microsoft’s deepest moat for years. Through Windows and Office alone, they are in virtually every enterprise IT department and on virtually every enterprise employee’s screen. When a competing product comes along, they clone it and bundle it into an existing suite of products. Suddenly it’s on every desktop, for nothing. It’s not as good as the standalone version, but it is there and it is free. Adoption skyrockets. Over time Microsoft improves the product and then monetises it by re-segmenting the bundle.

On top of that, Microsoft sells trust. They wrap the bundle in all the things IT departments love - permissioning structures, data security. But they don’t charge too much. They’re selling software, so the gross margin on the incremental sale is extremely high. They can afford to keep the price of the bundle well below the total price of all the standalone options you’d have to buy to replace it. (Years ago a colleague of mine looked at this. Everyone at the company lived in Microsoft products all day. Literally all day. And yet the company spent less with Microsoft than it did on milk for coffee and tea.)

The bundle has all the products you want, wrapped in all the compliance and security stuff you need; it requires managing one vendor relationship rather than 10; and it costs peanuts. What IT department wouldn’t use it? Code was never the moat. The cost to replicate the products has been falling for years and is tiny compared to the market opportunity. Distribution, bundling, trust, and price were the moat.

Does AI change this? I don’t think so, but there are wrinkles.

We saw in Part II that Microsoft is building 1) the cloud AI stack for enterprises to build bespoke agents, and 2) first party agents for big use cases. In the AI stack, distribution is a clear advantage but bundling isn’t. Microsoft has few peers when it comes to selling this kind of relationship to enterprise, although it is clear that Google, Amazon, Anthropic, and Open AI will compete. But the sales motion here is less about bundling and more about monolithic contracts, and the economics are about leveraging fixed costs. In proprietary agents, however, distribution and bundling may be a superpower. That is where I want to focus.

Bundling isn’t easy with frontier AI. When Slack launched, Microsoft cloned it, creating Teams. Whatever the cost of doing that was, it vanished into Microsoft’s overall R&D budget. By contrast frontier models cost tens if not hundreds of billions to train, and AI labs may only have a year or two to monetise them before their capabilities are commoditised. Microsoft has chosen not to develop frontier models itself, quite possibly because of that exact dynamic. The result is that it now distributes third party products, not its own. This creates a dependency that was not present before.

However, bundling commoditised AI might be very different. My thesis really hinges on one thing: for any given task, a given level of intelligence is required. That level of intelligence may be impossible one year, frontier the year after, and commoditised the year after that. As the frontier advances, the trailing edge commoditises, the cost of inference falls, and the pricing model switches to seat + consumption, Microsoft can bundle more and more intelligence at less and less cost. And that strikes me as an immensely (and increasingly) powerful proposition for both TAM and margins.

There are about 400m M365 Commercial seats worldwide. Of these, by March 2026, over 20m had adopted the paid M365 Copilot add-on, up 35% q/q. That is with Copilot’s capabilities today. In 5 years Copilot might still not be as good as the frontier models, but in absolute terms its capabilities will be incredible, it will be cheap, and it will be bundled into an enterprise’s existing financial and control arrangements. The chances of increased adoption (and maybe increased growth in M365 Commercial seats, too) are strong.

It’s obviously dangerous to extrapolate from one example, but this post on X captures what I am trying to describe. This user found that Excel in Agent mode can now update financial models from press releases. This might be child’s play for Claude, but that’s not the point. The point is that Copilot is improving and the user 1) I doesn’t care which LLM was used, just that it works; 2) prefers letting Microsoft choose the LLM over getting locked into one frontier lab; 3) might now use Excel more, not less, and for more productive things; and 4) would happily pay Microsoft extra for this capability. And as he says in the comments: why bother using the Claude Excel plugin if Excel Agent mode just works? I think that as commodity models advance, this thought process will play out across millions of use cases in thousands of enterprises.

If this is right it creates a couple of interesting dynamics:

1. Copilot may always be behind the frontier and may never have a great reputation as a result. But so long as it keeps getting better, that might not matter.

2. Frontier models may always dominate frontier workflows, but they will also have to charge frontier prices. As Copilot gets better, workflows developed on Claude might come back to Microsoft as commoditisation and bundling kicks in.

If this plays out as I think, Microsoft will become a critical distribution partner for commoditised LLMs. As such Microsoft can be an agent of model commoditisation, not just a beneficiary of it. It will also be an (unlikely!) ally of regulatory agencies seeking to prevent any potential winner-take-most dynamic in the LLM market.

The economics of bundling AI aren’t as clean as the economics of bundling software. The nth copy of a piece of software can be sold for virtually zero marginal cost, whereas intelligence costs money every time you use it. As discussed in Part II, Microsoft is moving towards a seat + consumption pricing model. This should both increase Microsoft’s revenues and improve Copilot’s performance. But can they make a margin?

• The bear case is that there isn’t a CTO on earth who hasn’t heard of Anthropic and OpenAI. Under pressure to adopt AI and aware that Copilot’s reputation is weak, they will test all options. They will not want to get locked into a single LLM provider, but if they somehow do (despite model commoditisation) then the frontier labs might generate cash flow in excess of the cost of training the next model. In this scenario the frontier labs both disrupt Microsoft’s cloud/productivity/agentic layer and build proprietary infrastructure to put pressure on the hyperscalers.

• The bull case is that with proprietary silicon, commoditised infrastructure, and commoditised trailing edge intelligence, Microsoft can deliver increasingly powerful intelligence cheaply through its unique distribution setup. Meanwhile frontier labs, desperate for cash to train their next model, will have to charge high prices for their older models, holding up a pricing umbrella for Microsoft.

There is clearly some validity to at least the first part of the bear argument. Open AI and Anthropic have bypassed Microsoft’s distribution advantage to some extent - otherwise their revenues would not be growing as fast as they are. Whether the second part of the argument holds - that these revenues are sticky - remains to be seen. It’s probably not binary - both arguments will be true to a degree. But on balance, I find the second argument far more compelling.

Share

Conclusion on distribution and bundling

As the rolling wave of LLM commoditisation advances and inference costs fall, Microsoft will be able to bundle more and more intelligence into its product suites at lower and lower costs. This plays neatly to Microsoft’s time-honoured strategy: be the second mover, clone and bundle the product, drive adoption, make it better, and then slice and dice the bundle to monetise it. If this works Microsoft will always look like it is behind the leading edge, and Copilot may always have a poor reputation compared to frontier models, but it will be good enough for a vast and growing range of uses, it will be cheap, and it will be delivered at strong margins.

Subscribe now

Combining layers

If anything renders Microsoft’s distribution moat obsolete it’s not the insurgency of OpenAI and Anthropic; it’s the full-stack advantage of Google.

In Part I I discussed the layers in the AI stack. Google owns all of them. It is producing its own silicon at scale. It is building infrastructure at pace. It owns its own frontier LLM, Gemini. It owns the full cloud stack needed to build and deploy agents. It owns distribution at vast scale to consumers. And it owns a huge search and advertising business that AI improves.

How does this full stack create advantage? Several ways:

1. Google can optimise margins across the whole stack, not individually for each layer. Said another way, it can use profits in one area to subsidise another while it builds scale and competitive advantage.

2. It can fully integrate and optimise its stack to deliver lower costs.

3. It can move fast, forcing cooperation across layers in ways hard to replicate when you don’t own every layer.

4. It can generate an immediate return on its LLM investments by deploying them to improve its existing businesses.

5. There will never be any question over whether Google has access to frontier intelligence (whereas Zuckerburg argues that OpenAI and Anthropic could claim “safety” reasons for not giving third parties access to their most advanced models).

I have not included the data flywheel here. Google has enormous amounts of proprietary data on which to train models. This data can be used to optimise models for improving Google’s existing businesses, like search. But I’m not convinced it is a huge advantage in enterprise, where Google is subject to the same restrictions as Microsoft (see Part II) and where Microsoft has more data.

The counterargument is that Microsoft also owns a lot of the stack. It doesn’t own a frontier model, but that might be a blessing in disguise given how unclear the ROI is in that layer. Instead, Microsoft owns the layers where ROI is clearer - silicon, infrastructure, cloud/agent stack, and distribution. If it can integrate these well, it should get sufficiently close to Google on cost, margin, and speed. Rather than locking clients into one LLM - which they will not want - it offers a marketplace of thousands of them. It absolutely owns enterprise distribution. Google does have enterprise distribution - over the last decade Google Cloud has become a formidable competitor to Microsoft’s Azure, and it provides an effective stack to customers wanting to build their own agents. But Google does not have an operating system and productivity apps already installed on the desktop of every user. That makes it more difficult to create proprietary agents and roll them out for free, without asking, and within existing security and permissioning arrangements. Finally, Microsoft is building cheap, optimised LLMs for specific, large use cases. In these cases, it will have genuine full stack economics. As cheap models get better and better, I expect Microsoft to address more and more use cases with this full-stack architecture.

I don’t mean to dismiss Google. It is certainly a formidable competitor. Gemini reasons better than Copilot, Google’s cloud business has built strong relationships across enterprise, and there is plenty of anecdotal data to suggest that Google is willing to discount heavily to encourage enterprise to switch. And there are advantages to the full stack. But Microsoft has advantages too. I am not sure Google’s full stack gives it a unique right to win in enterprise.

Share

Conclusion

Here is my Microsoft hypothesis per layer:

• Silicon: Microsoft is behind but catching up with Google and Amazon.

• Infrastructure: Likely to commoditise eventually, especially behind the frontier, creating a new utility compute industry and allowing hyperscalers to return to more capital-light ways.

• LLMs: A rolling wave of commoditisation, with frontier models never more than a year or two ahead of cheaper models. Microsoft will be the distributor of increasingly powerful commoditised AI.

• Agents: A goldmine, with a deep cloud stack allowing customers to develop, orchestrate, monitor, and control agents, plus extremely valuable horizontal and vertical proprietary agents. TAM is every process in every enterprise.

• Distribution: Microsoft’s deepest moat. Microsoft will bundle increasingly powerful intelligence into existing products, wrapping it in trust and pricing it cheaply. It will be hard for IT departments not to buy it.

AI is moving fast and the competitive dynamics are fluid. It is hard to develop confidence in where profits will eventually be made and who will make them. One strategy for dealing with this would be to reformulate investment theses whenever there is new news. I think this leads to high trading costs and chasing the latest hot stock. I prefer to form a 5-10 year high-level thesis based on an understanding of competitive dynamics and moat formation. That thesis can be assessed against short term newsflow, but it is important to give it time to play out. It is also important to diversify exposures.

Microsoft is well positioned to be the distributor of commoditised AI to enterprise. The TAM is huge and the right pricing models combined with Microsoft’s distribution advantages should ensure attractive margins. With revenue growth and operating leverage, I expect that Microsoft can compound earnings at 15-20% for many years despite rising depreciation costs. At 21x June 2027 earnings (less if you give credit for the Open AI equity stake), I think Microsoft is a compelling investment.

To end: as I said at the start, I have not addressed the ultimate bear case - that artificial general intelligence renders all competitive advantage obsolete and reduces returns across the entire tech industry to the cost of capital. I have three things to say on this:

1. I see this as a low-probability outcome for various reasons; in fact it’s equally likely that Microsoft ends up making good margins distributing commoditised AGI.

2. If it does come to pass, affordability and quality of life will improve dramatically regardless of how an investment in Microsoft works out.

3. This risk is easy to hedge. Plenty of assets will do well in a scenario in which humans have more discretionary spending power and more free time.

Links to previous Reviews

1. IRSA

2. Brookfield

3. Cheniere

4. Uber

5. Howard Hughes Holdings

6. Millrose Properties

7. Microsoft Part 1, Part 2.

8. My notes

Thanks for reading - if you enjoyed reading this please subscribe, like, and restack, and do get in touch if you have questions.

Pete

Tag for finding my other articles on this stock: MSFT

Key takeaways

EPS compounding high teens. 35% q/q growth in paid Copilot seats perhaps the most significant individual datapoint.

Subscribe now

Thesis and valuation update

No change to thesis. Intelligence is fragmenting and commoditising. Microsoft is well-placed to distribute commoditised intelligence to enterprise. TAM is vast. Stock trades at 21x June 2027 EPS.

Share

Notes

• Revenue +15% c/c, gross profit 13%, EBIT 16%, EPS 18% excluding OpenAI stake impact.

  • Infrastructure depreciation and agent usage decrease gross margins, offset by infrastructure efficiency gains at the operating level.

  • Cloud revenue +25% c/c at 66% gross margin.

  • Azure revenue +39% c/c, ahead on early delivery of capacity.

  • AI business $37bn ARR, +123%. Includes paid Copilot, Azure AI consumption revenue (OpenAI, Foundry, AI-driven compute/networking/storage, etc.), revenue attributed to AI features embedded in core products.

  • M365 Commercial seats +6%.

  • EBIT margin 46%.

  • Headcount down as they build fast, agile teams.

  • CFOPS +26%. Capex $31.9bn, 2/3rds on short-lived items (which is what correlates with revenue). FCF $15.8bn.

  • FY capex $190bn including $25bn for higher component prices. Demand exceeds supply at least through 2026.

• 2 mutually-supportive priorities as Agents drive TAM:

  • Building the world’s leading cloud and AI infrastructure for third party agents.

  • Building high-value first party agents across core domains like productivity, coding, and security.

• Infrastructure: optimising every layer - DC design, silicon, system software, model architecture.

  • Reducing time to deploy, increasing efficiency. More revenue, sooner, at lower cost.

  • Maia 200 now live in 2 DCs and offers 30% more tokens per dollar than the latest chips in their fleet.

  • 40% improvement in inference throughput for Copilot’s most-used models, on software and hardware optimization.

• Enterprise data and context: “across Fabric, Foundry, Microsoft 365 and our Security Graph, we are building a unified IQ layer for organizational intelligence…And as AI usage grows, so does the context layer, creating a flywheel that continuously improves the grounding, relevance and effectiveness of every agent”.

  • Cosmos DB revenue +50% y/y driven by AI.

  • Fabric has 35k paid customers, up 60% y/y.

  • Amount of data in OneLake (which underlies Fabric) grew 4x.

  • 15k customers use Foundry and Fabric, up 60% y/y, connecting agents to data.

• Third party agents.

  • AI Foundry offers 11,000 models. 10,000 customers have used more than 1. 5,000 have used open source models. Foundry has c.60k users but they vary in sophistication.

  • Agent 365 is the “control plane” - governance, identity, security and management frameworks for agents. “Tens of thousands of companies are already managing tens of millions of agents in Agent 365”.

• First party agents.

  • “Here we are in 2026 and the most exciting things are plug-ins in Word or Excel or CLIs in coding…that means we have a structural position in knowledge work, coding, security, which are the big TAMs”.

  • Copilot is evolving from a chat assistant to a coworker, executing long running tasks across key domains. Innovation is accelerating - M365 saw 625 updates over 12 months, up 50%. Includes intelligent auto-routing to the right model, and use of multiple models to improve responses.

  • Copilot adoption and engagement are exploding. M365 Copilot seat adds up 250% y/y. M365 Copilot paid seats up ~35% q/q to >20m and accelerating. MAUs of first party agents is up 6x ytd. Copilot queries per user are up 20% q/q. Copilot weekly engagement is now at the same level as Outlook (impressive but definition not clear - likely the percentage of licensed users who use it at least once a week).

  • WorkIQ now spans 17 exabytes of data, growing 35% y/y, with billions of e-mails, documents, chats, Teams meetings, and SharePoint sites added each day. “Copilot is uniquely valuable at work where nearly every task depends on organizational context. Work IQ grounds Copilot responses in the full context of an organization, including people, roles, documents and communications, all within the company’s security boundary….this is not some static database. It’s the most important database in any company that is constantly changing every second”. And as Copilot adoption grows, Copilot conversations and artifacts get added, making the data richer.

  • 1st-party models differentiate high-value Copilots and reduce COGS. New voice transcription and image generation models increase GPU efficiency by 67% and 260%, respectively.

  • Examples

    • LinkedIn Talent Solutions helps hirers automate sourcing, screening, and messaging - $450m ARR already.

    • Security Copilot customers up 2x y/y.

    • GitHub “is seeing unprecedented growth driven by proliferation of agentic coding”. Enterprise subscribers to Github Copilot up near 3x y/y, the majority using multiple models.

• Shifting to seats + consumption model (aka seat + agent).

  • Seats include some consumption. Long term consumption commitments will get discounts. Additional consumption charged on top.

  • Customer Service is at the forefront with 60% of customers now buying usage-based credits.

  • Github Copilot is moving to seat + usage too.

  • On a like-for-like basis this reduces bookings, since seats are booked up front and usage is real time, but not revenue.

  • As value-add rises, this accelerates revenue growth in M365 Commercial.

• Windows monthly active devices >1.6bn, “and over time, Windows value will extend to deliver unmetered intelligence at the Edge”. Bing MAUs topped 1bn.

• OpenAI agreement evolution maintains win-win dynamic. Still have royalty-free access to a frontier model through 2032.

Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Tag for finding my other articles on this stock: HHH

Key takeaways

Howard Hughes released a new set of metrics and valuation methodologies. These are useful, especially adjusted maintenance free cash flow for the Operating Assets segment. Management peg per share value at $104 today with potential for $211 by 2030 - their deck is linked below. Share price today is $63.

Subscribe now

Thesis and valuation update

No change to thesis. The real estate business shows momentum and is generating cash which will mainly be redeployed into Vantage, the new insurance subsidiary. That acquisition should close soon. Marc Grandisson, the ex-CEO of Arch, one of the best-performing insurance companies of the last 25 years, has joined the HHH board. This bodes well for Vantage.

The new valuation framework is similar to mine and produces a similar outcome. I’ll update mine for the new metrics and republish it in the next few weeks.

Share

Notes

• Operating summary

  • MPC EBT +33% on higher land sales. “The real estate engine did exactly what we needed to do. It grew cash, it provided pricing power and it converted more land into long-duration income.”

  • Operating Asset TTM same-store NOI +2%, driven by leasing and the burn-off of rate abatements in multifamily and office.

  • Strategic Developments: broke ground on The Launiu which is 70% pre-sold.

  • G&A high this quarter at $25.8m on $3.4m of Vantage acquisition costs.

• New metrics

  • MPCs: residual land value, undiscounted and uninflated.

  • Operating assets: Adjusted Maintenance Free Cash Flow. This is NOI - interest - amortisation of deferred leasing costs - depreciation of tenant improvements. It’s a good proxy for “cash we can take out”. It should grow high single digit organic (mid-single digit same store rental growth + 3-5% from operating leverage) plus 1-2% from building new assets. It is well diversified, with Office, Retail, and Multifamily making up about 30% each and Other the rest.

  • Strategic Developments: gross profit after tax of remaining condo sales from projects in development and advanced predevelopment.

• New valuation methodology

  • Detailed deck here.

  • Their 1q26 valuation is “basically a liquidation value”. NPV is higher if they keep rationing the supply of land, reinvesting in communities, and building the commercial land themselves so they keep the development profit.

• Balance sheet

  • $1bn debt raised at tightest credit spreads in HHH history. Half at 5.875% due 2032, half at 6.125% due 2034.

  • Repaid $750m of 5.375% notes due 2028.

  • $300m 5y mortgage at Downtown Summerlin raised at 5.52%.

  • $1.8bn in cash. Including the PSH Vantage preferred, all plans fully funded.

• General commentary

  • Summerlin land values have compounded at just under 15% over 5 years and the cost of capital should be a modest spread over treasuries given this is an established community and it is a virtual certainty the land will be sold.

  • Real estate overall will generate $2.5-3bn of excess cash by 2030. (This is after building out more Operating Assets. Unclear if it is pretax or after.)

  • Focus for the next few years will be on injecting additional equity into Vantage rather than buying additional operating companies, in the expectation that Vantage will do 15-20% ROE and be valued at 2x equity.

  • Beginning of a meaningful transition in the shareholder base.

  • Shifting from annual guidance to long term objectives per platform, “consistent with how we allocate capital and measure success internally”.

Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Tag for finding my other articles on this stock: IRSA

Key takeaways

The two segments that matter are malls and Ramblas del Plata. Malls is generating consistent ebitda. A positive here is international brands coming to Argentina. RdP is progressing well and will become a significant cashflow generator over the next 3-4 years.

Subscribe now

Thesis and valuation update

No change to thesis. Still priced to compound at +/- 20% in a bull case. I plan to publish a detailed valuation update in the next few weeks.

Share

Notes

Y/Y figures are in inflation-adjusted pesos. Because I did not publish 2q26 results, my notes for those are included at the bottom.

3q26:

• Malls

  • Occupancy 98%. EBITDA -1% y/y for the q and +2% for 9 months.

  • Tenant sales down 9% (9m/9m) on weak consumption and falling prices. However

    • 87% of IRSA revenues are fixed.

    • One reason is clothing inflation has exceeded inflation for several years - this is reversing as the economy reopens and imports compete.

    • No sign of tenant stress - number of transactions is stable, defaults are normal, rate of signing new leases is normal, fixed rents on new leases rising slightly higher than inflation.

  • International brands coming to Argentina as the economy opens. Positive for tenant mix and traffic.

• RdP

  • 2 more plots swapped for $11.3m - over 13,000 sellable m2, 3700 net to IRSA (implied % >27.4%).

  • Total now swapped or sold is 137k sellable m2 for $105m. The swaps account for 97k sellable m2, “nearly 25k” is net to IRSA, implying 25.5% IRSA share. This is inline with expectation that the initial plots would be at the low end of the 25-30% guide, and it is good to see IRSA’s share of the most recent swaps higher.

  • Expect over $300m of revenue from selling units received under swap agreements over the next 3-4 years.

  • 52% of the horizontal infrastructure for Phase 1 is complete and 23% of the total.

• Other

  • Offices fully occupied with small rent increases. Expanding one site with 72% already leased to Mercado Libre. Positive on office with strong demand and fewer new projects.

  • Hotels doing well on inbound tourism.

  • Net debt/ebitda 1.4x, LTV 11%. Will rise as they spend cash to accelerate developments.

  • Unlikely to enter the datacentre business but likely to build a significant logistics business over time.

2q26:

• Malls

  • Over time mall sales should rise with nominal GDP.

  • Recently traffic and volumes have been strong but prices weak, so tenant sales are down 9% y/y.

  • Ebitda +2% for the half though on inflation-linked fixed leases. Distrito

  • Diagonal mall on track for May 2027 opening - adds 22k GLA and first mall in La Plata city.

• Ramblas del Plata

  • Signed 2 more swaps for $12m. Total now 2 plots sold, 13 swapped, $93m, 124k sellable m2.

  • Expect sales prices of over $5000 per square metre for residential over the life of the project. Commercial, over $4000.

• Other

  • Launching coworking at the underutilised Philips building. Going well and will expand.

  • Issued another $180m of the 2035 bond at 8.25% (8% coupon).

  • ND/rental ebitda 1.6x, LTV 13%.

  • Now have enough cash on hand to finance all planned capex and acquisitions; debt ratio will rise as they spend the cash, then fall as they complete the projects.

Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Tag for finding my other articles on this stock: UBER

Key takeaways

Clear evidence of price elasticity - gross bookings is rising as insurance costs come down. This is critical for the long term thesis that Uber’s TAM will explode as AVs bring down costs.

Subscribe now

Thesis and valuation update

Thesis intact. Rapid growth, operating leverage, AV fragmentation, and positive commentary around their relationship with AI personal assistants which I see as the big competitive threat.

23x 2026 normalised EPS (per TIKR) and 16x trailing FCF. Lower if you adjust for equity stakes.

Share

Notes

• Gross bookings +21% (audience 17%, mobility 20%, delivery 23%). Revenue +10% (18% without move of driver payments from COGS to contra-revenue in the UK). Non-GAAP EBIT +42% on segment operating leverage offset by corporate investment in tech headcount. Non-GAAP NI +39% and non-GAAP EPS +44%.

  • TTM FCF $9.8bn.

  • $3bn buyback in the q (40m shares).

  • $8bn of equity stakes on the balance sheet, mostly public.

• Key operating stats/features

  • Insurance coming down, mostly passed through, strong elasticity helping bookings growth.

  • 50m Uber One members, up 50%; better retention, 3x spending, 50% of gross bookings are now members.

  • Cross-platform users growing 1.5x faster than overall users.

  • In mobility, barbell strategy: low cost products like Wait and Save drive 75% more frequency than core products while premium features drive 3.5x the profit. All drives higher lifetime value.

  • Adding selection in sparse/suburban markets is growing gross bookings in these markets at 2x overall rate and at higher margins. See higher % of reserve and Wait and Save in these markets too - improves reliability.

  • Uber for Business now does $5bn of high margin gross bookings and is growing 45%.Target $10bn by 2028.

  • Can now order eats to pick up on the way in a taxi, and delivery to room in hotels.

  • Can book hotels in-app via Expedia partnership. Long wondered whether they could evolve from on-demand to planned services. Uber Reserve answered this - hence hotels. Giving most of the economics of this deal to Uber One members.

  • Launched in Finland and straight to #1 in the App Store - there is a global halo/network effect, not just local.

  • Delivery in 35 countries now and targeting more. Within Delivery, Grocery and retail growing fast with added selection and growing awareness. AI shoppers help.

  • Freight gross bookings +6%. Nearly as many new enterprise customers added in 1q26 as in all of 2025. “Significant opportunity to more tightly integrate Freight with the broader Uber platform to create a cohesive end-to-end logistics ecosystem”.

  • Earners +21%. Various new features to boost driver experience - AI assistant, Hourly, Women Preferences.

• AI:

  • Bumping up FY AI budget already.

  • “It’s creating…employees with superpowers,” an accelerator for every company.

  • Will hire fewer people than planned this year due to AI.

  • Can now predict where you are going 3/4 of the time.

  • Turn a recipe, image, or prompt into a curated cart.

  • Can now personalise user interfaces. Different users interact in different ways. To date they have had to build for the majority. Now they can build for everyone. Key for adding new services, driving cross-platform usage, personalised upsell, etc.

• AV “is another $1tn TAM”.

  • Live in 8 cities, 15 by yearend, rides up 10x y/y.

  • No negative impact when Waymo launches in a city. In markets where Waymo has been operating for a while (SF, LA) Uber’s category share is higher than 6 months ago.

  • Bottleneck is cars on the road.

  • 30 autonomous partners. Success in signing partners is because they have demand. Also Uber Autonomous Solutions lets partners focus on building the driver while Uber does everything else.

  • Santander deal paves way to AV fleet financing. Difficulty is lack of established residual value, but their ability to offer predictable demand (and higher than 1P networks) underpins attractive financing.

  • Working with Marsh and Apollo on insurance.

  • Working with Hertz on fleet management and securing depots in regulation-ready markets. Builds on work they have done for some time as drivers have switched to EV.

  • Data offering for training models “has scaled quickly” - some of the earner fleet has robotax-grade sensors creating a differentiated dataset in combination with Uber’s proprietary data.

  • AV investments need to drive audience acquisition, frequency, or margin. Investing capital for now, but vehicle investments are transferrable to finance partners in time.

• Personal assistants vs interacting with apps

  • “Talking to many of these third-party agents. We have a great market position… [we] often dictate the terms of trade in those discussions”.

  • “I think we’ll continue to see that the majority of our transactions come direct.”

    • Building an indispensable service in mobility and delivery over 70 countries.

    • Engaged users / Uber One bring people to the app.

  • Assistants will interact with Uber’s agents.

  • Fears of metasearch in travel were unfounded. Most of the value stayed with Expedia, Booking, Airbnb.

  • Google Maps had comparison shopping between Uber and Lyft, and it wasn’t the same experience as coming direct to the app.

Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Tag for finding my other articles on this stock: LNG

Key takeaways

EBITDA and DCF guide increased; execution excellent as usual; buyback executed with real discipline, averaging $202 in a quarter that saw the stock range from $194 to $297.

Subscribe now

Thesis and valuation update

Minor increase in 5y cash flows likely but no substantive change.

At last closing price of $240 my (simple) DCF shows a 15% IRR through 2030. Caveat: this assumes buybacks are at the current share price.

Share

Notes

• Ebitda $2.3bn, DCF $1.7bn. Net income negative due to noncash charges related to the accounting mismatch on their IPM contracts.

  • Bought back 2.7m shares for $535m ($202). Opportunistic. Overage on DCF or delays in FIDs = more buybacks.

  • Paid down $250m of debt.

  • Issued $1bn of 2036 notes at 5.2% and $750m of 2056 notes at 6%, their inaugural 30-year issuance. Used some proceeds to prepay $550m facility.

• Increasing FY guide to $7.25-7.75bn ebitda (up $500m) and $4.75-5.25bn DCF (up $400m) on higher production, higher marketing margins, and optimisation efforts. Also locked in 1mt of 2027 CMI volumes at good margins.

• Project progress

  • CCL Stage 3 project 97% complete. Train 5 substantial completion achieved in March and T6&7 on track for summer and fall - each a few weeks ahead of the schedule assumed in the initial guidance. First LNG at T6 imminent.

  • T8/9 and debottlenecking project now 37% complete.

  • Phase 1 expansions at SP and CC on track - “the most compelling risk-adjusted infrastructure investment opportunities on the Gulf Coast or maybe all of North America”.

  • After Phase 1 expansions, CC probably the focus, advantaged by having 500 acres of unused land with access to water, next to their power plant, and only 40 miles from their Sinton station for gas supply. SP has land, but it’s wetlands.

• Hormuz/Qatar

  • Demonstrates the value of geopolitically stable supply and the role of flexible US volumes in balancing the market - most Qatari supply goes to Asia so US cargoes rerouted there. These benefits are reflected in current commercial negotiations.

  • 13mtpa of supply off the market for up to 5 years.

  • Also likely to see delays to new capacity planned for Qatar and UAE.

  • 2026 is much tighter than thought and 27 “more structurally constrained”.

Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

What it does: enterprise software.

Elevator pitch: arguably one of the finest franchises ever built, with deep distribution advantages that let it deliver innovations cheaply to a huge customer base.

Mental model: moat (read about my mental models here).

Valuation and potential returns: 21x June 2027 EPS estimates and growing EPS 20% per year.

Exchange and ticker: Nasdaq; MSFT.

Stock price and market cap: $421, $3.1tn.

Do I own it? Yes

IR website: here.

Tag for finding my other articles on this stock: MSFT

About this blog: I have been investing for 25 years, professionally and personally. I look for stocks that have a high probability of compounding at 15% for at least 5 years with limited downside. I write these stocks up on my blog. You can find more about me, my philosophy, my mental models, and my portfolio structure on my site.

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Part 1 of this review introduced the layers of AI and dived into two, silicon and infrastructure. It concluded that

1. Microsoft is behind in silicon because it started late, but Maia 200 is promising and Microsoft will close the gap.

2. Infrastructure will commoditise slowly, and a new utility compute industry will eventually allow hyperscalers to go (partway) back to being capital light.

Part 2 dives into the middle two layers: LLMs and Agents.

Subscribe now

LLMs

Large Language Models are the engines behind AI. Is this a winner-take-most market, or will LLMs be commoditised? If it’s winner-take-most Microsoft has a real problem. It doesn’t own a leading LLM, so it won’t be the winner; in order to sell AI to its customers, therefore, it will have to buy from the winner, who will have the whip hand in negotiations and might bypass Microsoft entirely to own the customer relationship directly.

Winner-take-most could come about in two ways:

1. One LLM lab reaches artificial general intelligence first, it could pull ahead of the others in performance terms. AGI isn’t well-defined, but in this context it means the LLM would improve itself faster and at lower cost than is possible with humans managing the process.

2. Several LLM labs run out of money, leaving only one.

I think it is vanishingly improbable that either of these things will happen, for a variety of reasons.

The first is that it is in literally nobody’s interests for one LLM to dominate the future. Here’s a short list of people who would absolutely hate this outcome:

• Every consumer.

• Every enterprise.

• Every voter.

• Every government.

AI is an incredibly potent technology. AGI will be terrifying. The idea that one company would be allowed to monopolise that sort of power is effectively impossible. Governments, customers, competitors, and regulators will work hard to ensure this doesn’t happen. And if it looks like it’s going to happen the company concerned will be broken up, so winner-take-most isn’t even really in the interests of the winner.

Share

The next reason is simply that I don’t think technology works this way. Technology seems to be inherently replicable. Like the 4-minute mile or the 2-hour marathon, once one person can do something, others follow. I can’t think of any major technologies that were cutting edge 5, 10, or 20 years ago that have not been replicated. I am not sure AI will be any different.

The evidence so far suggests that LLM advances are indeed replicable. There are perhaps 4-8 frontier labs. None has so far carved out a significant lead - on the contrary, models from these labs leapfrog each other in capability. Further, xAI is only 3 years old and Meta has stumbled, fallen behind, and is apparently catching up again. None of these datapoints argues for winner-take-most. On the contrary, the evidence suggests that hiring the right people and throwing money at them gets you a seat at the front of the race. And the number of “right people” will inevitably grow over time.

Equally important is the proliferation and performance of second-tier models. While the frontier models are breaking new ground at great cost, second-tier models are about a year behind and cost far less. This one-year gap varies by task and over time, but it doesn’t appear to be trending longer. This is key. A certain level of intelligence is required for any given task. Eventually, many models will be good enough for each task. Intelligence that is impossible this year might be frontier next year and commoditised the year after. In addition there is evidence that the software surrounding a model can be as important as the model itself: for example, Microsoft Critique uses multi-agent review to achieve better results than the underlying models alone, suggesting that combinations of commoditised models may become extremely capable.

I believe all of the above argues strongly for a rolling wave of commoditisation behind the ever-advancing frontier LLMs. Microsoft seems to agree: I think their strategy is to become the distributor of commoditised AI. This is a strategy shift from the early days of AI in 2022/3. Back then they built everything on OpenAI models, arguing this drove utilisation and efficiency. Why has the strategy changed? A cynic might argue it’s because they don’t have a frontier model of their own and their relationship with OpenAI has become rather less cozy. But there might be a simpler answer: initially Microsoft didn’t have a choice, but 18 months later they did. ChatGPT kicked off the AI era in November, 2022. It took about a year for all the other frontier models to become available and months more for powerful open-weight models to proliferate. But proliferate the models did, and in May 2024 Microsoft publicly announced a model-agnostic architecture for the first time. Today there are over 11,000 models available on Microsoft’s platform. That number alone is strong evidence that LLMs will commoditise.

Subscribe now

To distribute commoditised AI, Microsoft puts together several things. First, the models, available via API. Second, the harness in which the models run. Third, the reams of enterprise data stored in Microsoft’s systems. And fourth, the customers. Who benefits? The commoditised LLM benefits from broad distribution. The customer benefits from access to cheap models and from not locking themselves into a frontier lab that might not be competitive in the future. And Microsoft benefits, because amongst other things:

• Inference explodes. Models are cheap to run. Labs don’t have to invest in distribution. Low costs drive AI adoption.

• Costs can be optimised by routing tasks to the optimal model.

• Performance can be optimised by using domain-specific architecture and/or more than one model.

• By building proprietary low-cost models and agents for specific use cases, Microsoft can build vertical niches with full-stack economics.

There are, of course, counterarguments to the commoditisation thesis. Perhaps the best is the most obvious: the independent frontier labs are growing revenue like weeds. Anthropic’s run-rate revenue reached $30bn in April. This is over 10% of the revenue run rate Microsoft has taken decades to build, and is up 3x in 4 months. OpenAI is reportedly on a similar trend. On the face of it, exploding revenues suggests these will be powerful companies in the future. But the more revenue the labs make the longer they stay alive, the more models there are, and the more likely commoditisation is. Also, I am not arguing that the frontier becomes commoditised. Indeed, the labs may be able to keep growing revenue by solving ever-more impressive problems at the frontier. But behind the frontier I predict rolling model commoditisation as domain-specific systems are built on second-tier models to solve problems that were frontier 12 months ago. That is what I think Microsoft is positioning for.

Share

Another counterargument is that models trained on data nobody else has access to are likely to outperform with regard to that dataset/use case. I don’t think this argues against commoditisation of models in general, but it is an argument in favour of companies like Meta and Google driving increasing performance out of their legacy businesses.

A third counterargument is that vertical integration might be key to optimising model economics. Complex queries are exponentially more expensive to run than simple ones, so a small fraction of super-users can dominate the cost of running a model. This means flat pricing does not work and even consumption-based pricing doesn’t perfectly solve the issue since inference pricing depends on datacentre utilisation: a complex task run at a slow time is cheaper than a complex task run at a busy time. Demand prediction is key and adding a distributor between the model and the customer obfuscates the data. I don’t think this kills a multi-model architecture. I find it very hard to believe that Microsoft can’t figure out a pricing model that produces stable average margins across its huge customer base. Indeed, they have already begun to introduce consumption-based pricing, which will help with this.

Finally, Mark Zuckerberg argues that for competitive and safety reasons, frontier AI won’t always be fully available via APIs. I can believe that. But I think second-tier models are likely to remain available via API, at least to grade-A distributors like Microsoft who can ensure safety and provide a route to market. In fact, as second-tier models get more powerful, I could imagine strict regulation of how they are distributed, which would benefit scale players like Microsoft.

The endgame may be commoditisation even in an AGI scenario. AGI should reduce the cost of developing frontier models. In addition, compute is only going to get cheaper. That argues for more and cheaper frontier models, not fewer and dearer. Cheaper frontier models would lead to a massive expansion in TAM for the industry, and probably excellent economics for the distributors.

Subscribe now

OpenAI

What does commoditisation mean for OpenAI? Microsoft’s contractual relationship with OpenAI keeps being weakened, so I don’t ascribe much value to it. However Microsoft does have a strategic interest in keeping multiple model providers alive, and it has an equity stake in OpenAI.

OpenAI’s frontier model performance is, broadly, matched by its main competitors. ChatGPT has a huge consumer user base (>800m) but engagement is low, there is no network effect, and switching costs for AI chat assistants are nil (I’ve personally gone from ChatGPT to Grok to Copilot to Claude and halfway back to Copilot). Unlike Meta and Google, OpenAI has no legacy business on which to unleash its LLMs to drive productivity and cash flow. OpenAI does have growing enterprise revenues but appears to be behind Anthropic, and anyway both have to compete with Google which can afford to subsidise its Gemini LLM aggressively, as it demonstrated when it reportedly undercut Anthropic when Apple was looking for a model to power Siri. OpenAI is burning cash fast as it invests to advance the frontier, and even on its own projections won’t achieve cash flow breakeven before 2030. Rolling commoditisation may prevent independent frontier labs like OpenAI and Anthropic from selling legacy models at prices that would support continued investment in their frontier models. If this is the case, they will have to generate a return on each frontier model in the year or two before its capabilities are commoditised.

None of this reads well. But big cash generative firms like Microsoft, Amazon, and Nividia have two good reasons to keep OpenAI alive: they benefit from advancing the AI frontier because this expands their TAM, and they benefit from preventing Google developing a vertically-integrated monopoly. Arguably, it is better that they share the costs than that each bears them alone. My base case is that they will keep OpenAI alive, at least for now.

Share

In the long run I see three possible outcomes:

1. There is a positive return on developing frontier LLMs, in which case OpenAI will survive and Microsoft’s equity will be worth something.

2. There is a negative return on developing frontier models, in which case OpenAI will die and Microsoft will buy it, shut down the frontier effort, and sell OpenAI’s legacy models to customers at strong margins.

3. Only Google and maybe Meta can earn a return on frontier models, by deploying them to improve their legacy businesses. I think this is unlikely - if there is a return on developing frontier models it will apply across the economy and thousands of companies will pay for it - but if it happens Google becomes an AI monopoly and gets broken up.

Conclusion on LLMs

While frontier models will remain an oligopoly with capital as a barrier to entry, trailing models will commoditise. As the frontier moves forward, more and more use cases will be addressable by commoditised models. This rolling wave of commoditisation will be wonderful for customers but also potentially very good for distributors. Microsoft will distribute these commoditised models by allowing customers to build bespoke apps and agents on them and by bundling cheap intelligence into existing services. Customers will choose Microsoft over independent labs to avoid getting locked into a single model vendor and to benefit from using the right model for each specific task. As commoditised models improve, Microsoft’s TAM will increase. Eventually every process in every enterprise will be addressable.

Subscribe now

Agents

This section is about everything Microsoft does to make agents possible, including its cloud platform, productivity suite, vast data lake, proprietary LLMs, and finally the agents themselves.

Why am I lumping all of these things together? Because while several of them are powerful standalone businesses in their own right, I think it will become increasingly clear that they all exist to support agents. In the AI era agents will drive productivity, agents will expand the TAM, and agents are where the money is.

Cheap models make powerful agents economically viable. As the AI frontier rolls forward agents will be able to do more and more complex tasks. In addition as silicon advances and datacentres become more efficient, cost per token will continue to fall. This combination - more powerful agents that cost less - will cause an explosion in use cases. To capture this opportunity, Microsoft has developed products to address every stage in the process of creating, deploying, using, and managing an agent:

• Azure is Microsoft’s cloud platform and the foundational software infrastructure layer. It provides the global cloud infrastructure, security, identity, and scalable compute on which every other product in the agentic stack runs.

• OneLake stores enterprise business and operational data, natively or via mirroring. Fabric ingests, transforms, analyses, and models this data to surface business-ready insights. Fabric IQ adds business semantics so te data has consistent meaning. Foundry IQ turns scattered enterprise content into usable data for agents. Work IQ uses Microsoft 365 emails, meetings, and documents to help agents understand a user’s role, relationships, work patterns, and how their organisation actually operates.

• Foundry lets developers build, deploy, and govern code-heavy agents. Copilot Studio lets users build low-code agents that live in Microsoft 365 apps. Agent 365 lets organizations register, manage, secure, and monitor agents from any source.

• Microsoft 365 Copilot embeds agents directly into Word, Excel, Outlook, Teams, and other productivity tools.

• Microsoft Agent Factory is a program (and procurement bundle) that brings Work IQ, Fabric IQ, and Foundry IQ together under a single metered plan, helping organizations move quickly from experimentation to deployment.

One important caveat: the customer, not Microsoft, owns the data in OneLake and Work IQ. Microsoft does not use it for training models or agents that it makes available to other customers. The data can only be used to train models and agents for that specific customer, with that customer’s permission. Nonetheless Microsoft’s agent stack is immensely powerful, combining infrastructure, data, data preparation and analytics, human and organisational insights, tools for building simple and complex agents, surfaces on which to deploy them, and tools to control and manage them.

Share

Microsoft is building two business models on this platform:

1. The platform, sold to customers to create their own agents.

2. Proprietary agents, sold to customers directly.

Customers building their own agents will be a huge business. As a matter of competitive necessity, every enterprise will apply agentic AI to every process to drive efficiency. Each enterprise has specific use cases and ways of working, so they will develop specific agents that suit their needs.

Building proprietary agents is potentially transformational. Microsoft has always been a horizontal software company. It builds software that can be used by all enterprises across all sectors. Outlook, Excel, security - all enterprises need these. The first thing proprietary agents can do is transform the capabilities of the existing horizontal productivity suite. By adding AI chat to Windows and agents into major applications like Excel, Microsoft increases the value-add and pricing power of its legacy products.

The second thing proprietary agents can do is help Microsoft go vertical. By building agents that automate sector-specific workflows, Microsoft can build large vertical software businesses. Microsoft Copilot inside Excel now does AI-powered DCF modeling and financial statement parsing. Copilot inside Word does contract review and case law research. Developing vertical software no longer requires domain engineers and years of development; it now requires an agent and a customer relationship. Microsoft has both.

To date Microsoft’s proprietary agents don’t have a great reputation. For example, Copilot chat has a reputation for hallucinating. I think this has caused analysts to underestimate the potential of this business. To date, Copilot has run on OpenAI’s GPT series of models. These are highly competitive but Copilot underperforms because Microsoft has limited its context window (presumably to control costs).

Three things will make Microsoft’s proprietary agents vastly better. The first is Microsoft’s move to a seat + consumption pricing model, which will allow Microsoft to charge more for users who need better performance. The second is cheaper inference. And the third is the rolling commoditisation of LLMs, which will be able to address exponentially more use cases over time.

As Satya Nadella said on the 3q26 earnings call: Agent Mode in Excel “sort of didn’t work until it started working” because the model got better. Many use cases will follow this pattern.

Subscribe now

Competition in agents

Microsoft is not the only company offering an agent platform or proprietary agents. At minimum, it will have to compete with the other hyperscalers plus Anthropic and OpenAI, and a host of single-product startups building agents for specific uses. I think Microsoft is well-equipped to compete, with advantages that include:

1. Distribution. Microsoft can bundle agents into existing subscriptions and products. More on this in Part III.

2. Trust. For decades, enterprise IT departments have been built around Microsoft’s data privacy, permissioning, and security architectures. This is not an unassailable moat, but it will take time for each competitor to build trust.

3. Data. With a customer’s permission, Microsoft can build agents that know who you are, what you have worked on, who you interact with, what permissions you have, and how your organisation works.

4. Scale. Microsoft can leverage the cost of developing and maintaining its agent platform and its proprietary agents across thousands of existing customers, globally.

5. Switching costs. Agents reduce these initially, by handling migration tasks that used to be a headache. But agents that handle workflows will become deeply embedded in how each organisation operates. Once large numbers of agents are working together to coordinate and execute complex tasks, switching costs will be significant. Customers know this, which is an incentive to work with Microsoft rather than getting locked into Anthropic and OpenAI (which may not stay solvent, let alone with the LLM race).

Despite this, customers will likely build agents on Microsoft’s platform and experiment with other vendors. Some of these new relationships might become sticky, or Microsoft might win these customers back as its proprietary models and agents improve. Either way, the market is big enough for several players.

Overall, I think that as the infrastructure and LLM layers commoditise, value will shift to “compound systems” like Microsoft’s agent platform.

Share

Economics

If that is right, then the economics are very attractive.

For starters, the market is huge. As agents do more of the work that humans can do, enterprises will devote more of their operating budget to tech spending. Agentic efficiency will also help them develop new products and services, growing revenue and boosting spending power. To get an idea of the scale, consider Microsoft 365 Copilot. Microsoft 365 has 450m customers. Over 20m of them pay for M365 Copilot. That number rose 35% in the last quarter. Paid M365 Copilot costs $30 per month. That makes it a $7bn product already. As it gets more powerful, I expect adoption and possibly pricing to rise. M365 Copilot alone could theoretically be a $50-100bn product. That’s just the general, horizontal Copilot. There are others for specific verticals, and that’s before you consider all the other agents being built on Microsoft’s platform.

If the competitive arguments above are right then margins should be strong. The platform itself has some of the economic characteristics of legacy software. Each new customer involves additional compute costs, which can be covered by consumption-based pricing. However there is no additional software development cost for each new customer. Like Windows and Office, the software component of Azure, Foundry etc. can be sold millions of times at near 100% gross margins. The cost of coding is falling fast, but that’s been true for decades; code has never been Microsoft’s competitive advantage. As long as there are some remaining advantages around distribution, trust, data, scale, and switching costs, margins should be attractive.

On the proprietary agent side, the same applies: the same advantages, and the same low marginal cost of the software component. Some of the more complex agents may be able to achieve outcome-based pricing, capturing a portion of the value they deliver rather than a fixed price. In addition, the cost to run agents will fall as models commoditise and compute gets cheaper. All of this suggests strong margins.

One concern is that Microsoft does not have a frontier LLM itself and therefore has to pay external providers for frontier intelligence. The benefit of this is that it doesn’t have to fund the essentially speculative development of frontier models. The downside is that once the OpenAI deal expires in 2032, Microsoft might have to pay market price for frontier intelligence and its gross margins might fall. But by then, I think Microsoft will have built an immense business distributing commoditised models. These will be hugely capable and cheap, allowing high gross margins for a distributor with a moat. In addition, Microsoft is building proprietary models optimised for specific large use cases. These are cheap to develop and efficient to run, and will give Microsoft full-stack economics in specific verticals.

Another concern is that for decades, Microsoft has sold its products on a per-seat basis. If one agent can do the work of several people, customers will need fewer “seats”. But if agents make humans more productive enterprises might employ more of them, not fewer (Jevons Paradox). Second, Microsoft is already migrating to a seat + consumption model, and agents will consume a lot.

On the 3q26 results call, Microsoft’s CFO Amy Hood said that AI margins are better now than they were at the same stage of the cloud transition. She implied that this isn’t widely understood, and that consumption-based pricing will improve margins. If she’s right, these statements have enormous importance.

Subscribe now

Conclusion on agents

Agents vastly expand Microsoft’s TAM. Microsoft offers a full stack for third parties to build their own agents without locking themselves into one model provider. It is also developing its proprietary agents which improve the value offered by its productivity suite and allow it to go much deeper into large verticals where it will own the full stack. As commoditised LLMs get more and more powerful, Microsoft will be able to offer better and better proprietary agents at strong margins.

Agents are Microsoft’s future. In Part III, we will examine Microsoft’s superpower: how it will distribute them. We will also look at the competitive advantages of combining different layers of the AI stack, and briefly discuss valuation.

Links to previous Reviews

1. IRSA

2. Brookfield

3. Cheniere

4. Uber

5. Howard Hughes Holdings

6. Millrose Properties

7. Microsoft Part 1

8. My notes

Thanks for reading - if you enjoyed reading this please subscribe, like, and restack, and do get in touch if you have questions.

Pete

This email is a summary of my activity this month.

Trades

• Sold a tracker in BMW3. Can’t get comfortable with Chinese competition and the commoditisation of AV vehicles.

• Bought a tracker in Goodwin $GDWN.LN after a massive spike and drawdown. Family run compounder in high-end steel foundry.

  Subscribe now

Articles I wrote

• Can Microsoft Compete? Part 1 - revisiting a stock I have held since 2010.

• Millrose Properties. Don’t like the risk/reward one bit.

• Uber Freight. Counterpoint to the article linked below.

• Brookfield 4q call summaries. Done by hand, not by AI!

• Howard Hughes Holdings - shaping up to be a good compounder, if you can live with the fees.

What I found interesting this month

• on Millrose Properties - the bull case to my bear.

• on Howard Hughes, Progressive and Fairfax, and PE-backed life insurers.

• on the other side of the private credit coin.

• on offshore drillers.

• on Goodwin.

• on how dividends destroy shareholder value.

• on nuclear.

• on cheap Chinese cars - relevant to my Uber thesis.

• on how to think about volatility and the alt managers.

• on moats in luxury.

• on Uber’s logistics platform.

  Subscribe now

Thanks for reading - and please get in touch if you have questions.

Pete

What it does: enterprise software.

Elevator pitch: arguably one of the finest franchises ever built, with deep distribution advantages that let it deliver innovations cheaply to a huge customer base.

Mental model: moat (read about my mental models here).

Valuation and potential returns: 21x June 2027 EPS estimates and growing EPS 20% per year.

Exchange and ticker: Nasdaq, MSFT

Stock price and market cap: $401, $3.2tn.

Do I own it? Yes.

IR website: here.

About this blog: I have been investing for 25 years, professionally and personally. I look for stocks that have a high probability of compounding at 15% for at least 5 years with limited downside. I write these stocks up on my blog. You can find more about me, my philosophy, my mental models, and my portfolio structure on my site.

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Introduction

I have owned Microsoft since 2010. What gave me the confidence not to sell - even as the company transformed and the multiple expanded - was a deep conviction in Microsoft’s competitive moat. Not its technology. I don’t think technology ever constitutes a sustainable advantage in itself. The moat was distribution. Microsoft sells to enterprise. Enterprises move slowly. Microsoft is deeply embedded: enterprise IT departments are built on its products, and it has a huge reseller network reaching millions of SMEs. Microsoft leverages this difficult-to-replicate incumbency by copying innovations (it is the quintessential second-mover) and bundling them cheaply alongside existing products. This has been remarkably hard for standalone competitors to fight.

Distribution + time to react + cheap bundling = a moat full of crocodiles.

Does AI change this?

Subscribe now

Satya

Satya Nadella is the CEO of Microsoft and the chief architect of its renaissance over the last 15 years. Two things he has said over the years rattle round my brain:

1. The software industry - and by implication, Microsoft - has no franchise value.

2. With agentic AI, Microsoft’s TAM is every process in every organisation.

Both of these comments are mind-blowing. One is a stark warning. The other is exceptionally bullish. They appear contradictory, yet both are right. This makes analysing Microsoft a fascinating challenge.

Share

Why I wrote this

Microsoft is a largeish position for me, and I have a long term time horizon - 5 years at minimum, and ideally 10 years or more. However Microsoft’s business is changing rapidly as AI alters its competitive landscape. I needed a better framework for understanding how the industry is evolving. A vast amount has been written about this, although a lot of it is very short term (Azure only grew 39% in 2q26, not 40%!). This piece is an attempt to synthesise what I have read into a framework for interpreting newsflow over the next few years: what will be the signs that Microsoft’s competitive position is strengthening, or weakening?

I am a generalist, not a tech expert. This can make interpreting short term newsflow harder, but might make interpreting long term trends easier, because what will drive long term trends is not the latest tech advance but classic competitive dynamics. The key principle is simple: if too many companies can do a thing, pricing will fall until the return on capital equals the cost of capital. The main beneficiaries of innovation will be the customer, not the innovator.

This piece focuses on each layer of the AI tech stack and whether long term competitive advantage will be carved out in any of them (or by combining any of them). It does not address the idea that AI drives the value of everything to zero and software is dead. This is deliberate. I see limited evidence in favour of this theory so far. On the contrary, AI seems to be strengthening the classic tech competitive advantages of network effects and access to unique data. In addition I think it is almost impossible to analyse this thesis sensibly given the comprehensive advances in technology that would be necessary and the unpredictable ways incumbents might find to defend themselves. I do not dismiss the possibility, but rather than letting it scare me out of investing in the greatest productivity advance in human history I prefer to think through the wider impacts and find ways to hedge. But that’s a separate topic entirely.

This article will be published in 3 parts:

1. Introduction, layers, silicon, and infrastructure.

2. LLMs and agents.

3. Distribution, combining layers, and conclusion.

Subscribe now

Layers

There are multiple layers in the AI future:

We need to frame whether real competitive moats can be created within each layer or by combining layers. And to do that, we need to start at the bottom.

Share

Silicon

Having the cheapest compute is a clear advantage in the AI race. Evidence for this is already showing up in earnings releases. The undisputed champion of highly advanced general-purpose AI silicon is Nvidia and it does not look like this will change any time soon. All AI infrastructure builders will probably need buy Nvidia for a long time. This levels the playing field to some degree, because they’re all using the same chips for large parts of their workload.

That said, by designing silicon in-house hyperscalers like Microsoft can keep pressure on Nvidia pricing and maximise infrastructure efficiency by tightly integrating chip design with the hardware and software that comprises the rest of the system. One of the core themes of this article is that several layers of the AI stack will at least partly commoditise, and in commoditised industries efficiency is the key competitive battleground.

Amazon, Google, and Microsoft all have their own in-house silicon design programmes which are emerging as core competitive advantages and significant businesses in their own right. Microsoft’s programme is currently the laggard. It unveiled its first chip in late 2023, several years after Amazon and Google. It is therefore promising that Microsoft’s Maia 200 chip, released in January 2026, already appears to be competitive with Google and Amazon’s inference chips. However, the data comes from Microsoft’s internal testing, the exact parameters for these tests have not been released, and there is limited third party validation. In addition, Microsoft’s production volumes are much smaller than Google’s or Amazon’s.

In short: this is a weak point for Microsoft but they are addressing it. Given the importance of inference efficiency, what we know about Maia, and the Microsoft programme’s comparative immaturity, I think Microsoft is likely to close the competitive gap rather than slip further behind. I think all 3 companies will have highly competitive silicon, and this will be a competitive advantage against compute providers who don’t have in-house silicon. I am less certain that any of these 3 companies will develop a lasting silicon advantage over their peers.

Subscribe now

Infrastructure

The hundreds of billions being spent on AI infrastructure are the stuff of legend. Hyperscalers have little choice but to build today: the world is short compute, and if they don’t build it they can’t sell AI to their customers. That is death. Nonetheless, there are two big questions regarding this spend:

1. Will current spending produce a worthwhile return on investment?

2. Does owning infrastructure confer a lasting competitive advantage?

I think the answers are yes and maybe.

Share

Infrastructure question 1: ROI

Here’s how today’s capex could earn a poor ROI:

• Supply gets overbuilt.

• Demand collapses.

• Obsolescence.

The supply overbuild argument is predicated on the idea that rising capex multiplied by improving efficiency = explosive growth in compute capacity. Currently, all the evidence is that more capacity is needed: hyperscalers say they are capacity constrained, Microsoft has to choose between allocating capacity to internal uses and selling it to customers, and older generations of chips aren’t being retired, suggesting they can still be rented out for attractive prices even though better chips are available. So far, so good. But will this last?

Bears point to the huge overbuild of fibreoptic capacity in the dot-com bubble as an obvious analogy. The problem wasn’t just that capacity was overbuilt initially; two additional factors combined to make things worse. One was that the capacity bottleneck was the send-receive technology at either end of the cable. Rapid improvements in cheap send/receive technology therefore multiplied capacity for years after the cables were installed. The second factor was that fibreoptic cables last decades if they’re installed well. As a result, overcapacity never self-corrected even as internet traffic exploded.

The datacentre buildout looks very different to my eyes. This is partly because there are clear braking factors on capacity growth, notably power availability, chip supply, and the fear of overbuying one generation of chips before a more efficient one comes along. Also, when we break the capex into its component parts some important differences with the fibreoptic buildout emerge. Roughly:

• 15% goes on land and buildings. This is the component that looks most like fibreoptic cables: the compute capacity that goes in the buildings will keep getting better and the buildings will last decades. But even if a few too many get built, the loss on the excess capacity won’t be total: land and buildings can be repurposed and retain value.

• 25% goes on electrical and cooling equipment. In the event that capacity is overbuilt this will suffer significant impairment, although much of it can be repurposed/used for parts.

• 60% goes on the actual compute, including GPUs. This is where most of the capex is going and it is where the fibreoptic analogy really breaks down. Although software improvements can meaningfully improve the performance of installed chips, there is no bottleneck equivalent to send/receive in fibreoptic. Instead there is a hard ceiling set by the physical hardware, so capacity cannot multiply for years after installation. And the useful life of chips, while subject to much debate, is far shorter than that of fibreoptic cable. As a result, compute capacity degrades over time rather than compounding. Even if capacity gets overbuilt, this will bring supply and demand back into balance relatively quickly.

What about demand collapse? The bearish argument is that training frontier models has an unproven ROI, and if investors get tired of pumping cash into OpenAI and Anthropic (in particular) they will have to cut back on training their next generation models. However, I think the risk of demand collapse is diminishing:

1. Inference has grown from 1/3rd of compute usage in 2023 to 2/3rds in 2026. Inference is driven by adoption of AI, not training new models. In my opinion, adoption has barely begun. Even if training spend drops, inference will keep growing, and compute built for training can be retasked to run inference workloads (albeit less efficiently).

2. It is becoming increasingly clear that AI has vast real-world uses. OpenAI and Anthropic are growing revenues rapidly - Anthropic has grown annual recurring revenue 30x in 15 months and 3x in the last 4. This might be the fastest revenue ramp in technology history. These companies are nowhere near self-funding, but as revenues grow their dependency on investors falls and their attractiveness to investors rises.

Finally, obsolescence is the idea that if tomorrow’s data centre is vastly more cost-effective than yesterday’s, then tomorrow’s datacentre will set pricing and yesterday’s will struggle to earn a decent ROI. What great leaps forward will render today’s capacity obsolete? Two possibilities come to mind. The big one is the efficiency of leading-edge chips. These will inevitably obsolete older chips. That’s a problem if the step up from one generation to the next is huge, and if the generations come in rapid succession. The evidence so far is the opposite: chips remain competitive long enough to more than earn their cost of capital. My guess is this will continue, but of course it is linked to demand: the combination of a huge step up in chip efficiency and a significant slowdown in demand would render a generation of chips obsolete. That said, this would be a one-time event and would collapse the cost of compute, stimulating demand and raising margins in other parts of the business. It’s a risk, but not an existential one.

The other potential great leap forward is some paradigm shift, such as Elon Musk’s goal of putting datacentres in space to benefit from 24/7 solar power and infinite radiative cooling into deep space. While this is theoretically coherent, it is a wildly complex idea to deliver cheaply. To put it politely, I suspect we are in for a wait.

In conclusion: the risk of a short to medium term ROI collapse is likely overstated by the bears. Demand is growing ahead of supply and may do for years. Indeed, the real risk is not spending, not being able to sell AI, and losing customers that can’t be regained. Nonetheless, at some point supply and demand will meet. ROI will compress, but capacity degradation will correct this fairly quickly. Also, any reduction in ROI on infrastructure may be offset by margin improvements in the layers that use compute. More on this in Part II.

Subscribe now

Infrastructure question 2: does owning infrastructure confer a lasting competitive advantage?

Question 1 dealt with the short to medium term ROI on current capex. Question 2 is about long term ROI on infrastructure: does owning infrastructure confer competitive advantage such that ROI stays high, or does infrastructure commoditise?

There doesn’t seem to be much competitive advantage in owning the “bare metal” - the land, buildings, cooling systems, racks, and chips that make up a datacentre. But there is competitive advantage in integrating it all together. Running 100,000 GPUs as a coherent unit with optimised interconnect, utilisation, cooling, and so on is extraordinarily hard, and doing it better drives significant efficiency gains. Microsoft, Amazon, and Google have spent a decade learning how to do this. A new entrant cannot replicate this knowhow overnight even if they buy the same chips. But whether this edge is sustainable is another matter. We are in the early innings of a massive buildout. Multiple players are throwing capital at this and need to solve it if they are to survive. I think it is likely that the optimisation gap between the leaders and followers will narrow over time.

The other way to think about this is by workload. Infrastructure for frontier models and the most demanding large-scale inference needs to be optimised to perfection and there are genuine gaps between the leaders and the rest. On the other hand, infrastructure for standard inference done by midsize models is already commoditising. Competition is mainly on price and reliability. In addition, as small models improve and chips get more energy-efficient, inference will move to the edge driven by advantages in latency, security, and cost. (In other words, for simple tasks models will run on your computer or phone rather than in a datacentre.)

These two trends - the narrowing of the optimisation gap, and the relentless forward march of the frontier leaving ever more commoditised inference in its wake - suggest to me that over time, more and more infrastructure will become commoditised. If so, a new utility industry will emerge selling commoditised compute with a mix of long term contracts and instantly-available capacity at spot. There may be competitive advantage in owning frontier infrastructure, but as the industry matures hyperscalers like Microsoft will be able to choose what they own and outsource the rest.

This has implications for ROI and free cash flow. The market sees the hyperscalers ramping capex and worries that ROIC and FCF must fall. I think this framing might be wrong. What’s really happening is that the hyperscalers are incubating a second business. While AI is compute constrained, hyperscalers must own compute in order to guarantee supply and sell AI. But in the long term, the decision to own compute will be driven by ROI. If there is long term competitive advantage in owning compute, the ROI will stay high and the hyperscalers will retain ownership. If not, utility compute can be separated from the original capital light businesses, either by selling/spinning compute or simply by reducing capex and buying third party compute. My money is on at least partial commoditisation and separation.

What really matters, therefore, is that the original capital light business is still there, still profitable, and has a vastly expanded TAM. We will explore this in Part II.

Subscribe now

Links to previous Reviews

1. IRSA

2. Brookfield

3. Cheniere

4. Uber

5. Howard Hughes Holdings

6. Millrose Properties

7. My notes

Thanks for reading - if you enjoyed reading this please subscribe, like, and restack, and do get in touch if you have questions.

Pete

What it does: finances land banks for homebuilders

Elevator pitch: homebuilders want to go capital light. Millrose is happy to buy their land and sell it back to them when they need it, for a fee.

Mental model: n/a - I was hoping it would be value (read about my mental models here).

Valuation and potential returns: trades slightly below book value with a 10% yield. I think the upside is capped at 10% and the downside is significant.

Exchange and ticker: MRP, NYSE

Stock price and market cap: $31, $5.1bn

Do I own it? No.

IR website: here.

About this blog: I have been investing for 25 years, professionally and personally. I look for stocks that have a high probability of compounding at 15% for at least 5 years with limited downside. I write these stocks up on my blog. You can find more about me, my philosophy, my mental models, and my portfolio structure on my site.

Subscribe now

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

I meant to write about Microsoft this week, but got distracted by this. I don’t plan to write about stocks I don’t own very often, but I enjoyed thinking through this business model and the risks.

What it does

Millrose finances land banks for homebuilders. Homebuilders find a property they want; Millrose buys it; the homebuilder pays a deposit and an option premium; Millrose funds the development of horizontal infrastructure (roads, utilities, landscaping etc.); the homebuilder can then build and sell homes; and then, finally, the homebuilder exercises its option to buy the property.

Share

How the finances work

Millrose’s P&L and balance sheet are almost completely separate (at least when things are going well - more on that later).

The P&L is quite simple. It looks like this: regular option premium payments, less management fees, less interest on debt, equals dividends. There is not much tax or retained earnings because Millrose is a REIT.

The balance sheet is also fairly simple: home sites on the asset side and equity + debt + deposits (in that order) on the liability side.

The cash flow statement is more interesting. Balance sheet activity creates massive inflows and outflows of cash. Millrose receives deposits, buys land, pays for horizontal infrastructure up to a pre-defined limit, and then sells the land at total cost. It looks like this:

In rough terms, the cost of the land and the cost of the basic infrastructure are about equal. I derive this from the fact that at year end Millrose had about $8bn of land on the balance sheet and will receive about $16bn if all options relating to that land are eventually exercised. The difference is the cost of developing basic infrastructure: another $8bn.

These asset cycling cash flows can be billions of dollars over a year. Because Millrose sells the land + infrastructure at cost, these transactions don’t usually touch the P&L. However, as we will see, there are real risks with this activity.

Subscribe now

Why Millrose exists

Millrose was spun out of Lennar. It buys land from or for Lennar and other homebuilders. This allows the homebuilders to control a land bank without carrying it on their balance sheet. The idea is that the stock market rewards capital light businesses with higher valuations. This is true, but the market is not stupid (most of the time). It understands that in capital-heavy businesses, there is a cost, financial or strategic, to going capital light. And homebuilding is a fundamentally capital-heavy business: you can’t build homes without land, and the land needs to be good.

To go capital-light, therefore, homebuilders need to get land off their balance sheet while keeping access to it, keeping the economic upside of ownership, and transferring most of the risk. And that is exactly what Millrose lets them do.

Landbanking is not new. What’s new is listing a landbanking vehicle. Most landbanks are funds. They raise capital, buy land, sell it, and return the capital to investors. Millrose, being listed, doesn’t return the capital. Instead it recycles it into the next land acquisition. The other difference is that Millrose, being a REIT, taps into a large market of yield-hungry investors, particularly (I suspect) retirees. This may reduce the sophistication of Millrose’s investor base, and its cost of capital.

Share

The bull case

Homebuilders are willing to pay 8-11% of the land value per year in option fees, which is sufficient for Millrose to pay a good dividend to its shareholders.

Millrose’s addressable market is theoretically huge. The value of the land and horizontal infrastructure underlying annual home starts in the US is $170bn. That’s all got to be on someone’s balance sheet, and why shouldn’t it be Millrose’s? The potential for growth seems huge. And indeed, according to management on results calls, homebuilders have been beating a path to Millrose’s door since the spin in early 2025.

A hard-asset company, trading at a discount to book value, yielding 10%, with a long runway for growth? What’s not to like?

Let’s find out.

Subscribe now

The problems, in no particular order

Millrose pays Kennedy Lewis a management fee worth 1.25% of tangible capital. Not net tangible capital, and not tangible capital per share - just tangible capital. This seems to create two gloriously perverse incentives:

1. to issue cheap shares and buy expensive assets. There is some protection for the first 18 months post-spin, because Lennar kept a right to be compensated with extra shares if Millrose issued new shares below Lennar’s in-price. But once that date is past, I imagine the stock certificate printer at Millrose will go brrrrrrr. This might make it hard for the stock to rerate.

2. to carry assets at inflated valuations. Note that Millrose has not obtained an independent appraisal as to the value of its land. In effect, we have to trust the value that Lennar put on the land when it span Millrose out. Future acquisitions will be carried at cost (until they’re not).

Homebuilders sell land to Millrose, or bring Millrose ideas for land to buy. I think we can safely assume that they keep their best land on their balance sheet. There is a clear risk that they will dump their weaker assets on Millrose, and Kennedy Lewis will buy them to boost fees. Worse, Kennedy Lewis run traditional landbanking funds which compete with Millrose for acquisitions, so it is not even clear that the manager’s best deals will go into Millrose.

Millrose will struggle to grow book value or dividends per share. REITS have pay out most of their profits, so they can’t retain capital to grow. However most REITS do at least benefit from inflation in land prices and rents. Millrose does not: when they sell land, they only recoup their cost. Any land price inflation during Millrose’s ownership accrues to the homebuilder. Inflation protection is one of the main reasons to own real estate of any kind, and Millrose doesn’t offer it. This is one of the reasons Millrose’s yield is higher than most REITS: most REIT yields are implicitly inflation-linked (because underlying rents are expected to rise with inflation over time) but Millrose’s is not.

Millrose only has two ways to grow on a per share basis, and both are limited:

1. It can take on debt, but this adds risk and Millrose cap themselves at 33% debt-to-capital. They’re already at 25%.

2. It can shift its business mix away from Lennar towards other homebuilders, who pay higher options yields. However, Lennar has the right to reserve a significant part of Millrose’s capital. The exact calculation is complex, but the impact is that Millrose can’t grow its business with other homebuilders very much without issuing new shares. Since issuance may well be below book value, the resulting dilution is likely to offset the mix shift benefit.

Homebuilders don’t have to exercise options. They don’t even have to build homes on the land. If a project is uneconomic homebuilders can simply walk away. If they do, they lose their deposit and pay a termination fee, but in a major market downturn or if there are extreme project-specific issues walking away might be the economic or even the only option. It is inevitable that some options will not be exercised. In these cases, Millrose keeps the land but almost certainly loses money. They may be forced to sell the land for less than they paid, or worse: when a homebuilder walks, Millrose is liable for all the costs related to the land, which might include claims from unhappy homebuyers, the cost to remove partially-completed construction, or environmental reclamation costs.

In theory Millrose is protected from homebuilders walking away from options because land parcels are pooled: if a homebuilder walks away from one option it loses all the deposits in the pool, and Millrose can terminate the other options in the pool. Diversification within pools is meant to ensure that the assets aren’t correlated. Pooling ought to incentivise homebuilders to exercise, not walk, but it is imperfect:

1. Millrose cannot refuse to sell assets to Lennar, even if Millrose dispute Lennar’s right to buy based on pooling cross-termination provisions. All Millrose can do is litigate to recover losses. There is no guarantee this will succeed, and doing it could wreck Millrose’s relationship with their major customer. In addition, Lennar’s payments per pool appear to be capped, which presumably limits the value of the pooling provisions to Millrose.

2. When Millrose was spun out of Lennar, it was Lennar that chose which assets went into which pool. It is prudent to assume that they designed the pools to their benefit. Pooling offers little protection if all the bad assets are in one pool.

3. The following language appears in the 10k: “Lennar retains substantial discretion in selecting pool properties and setting pool terms. We may have limited ability to negotiate pooling conditions with Lennar…and we may not be able to negotiate pooling terms at all with Other Counterparties”. While this is to some extent boilerplate wording, it’s also true: pooling is by definition done through negotiation with homebuilders, who are not stupid. They will not accept pool terms that significantly increase their risks.

There are two implications of the homebuilder right to walk. The first is that far from growing, Millrose’s per share metrics are likely to decline over time. Millrose does not make a profit when a homebuilder exercises their option to buy a homesite - it simply gets back what it spent. But when a homebuilder does not exercise their option, Millrose is highly likely to take a loss. Millrose cannot offset inevitable losses on bad projects with profits on good ones. It can replenish the lost capital by issuing shares, but there is no incentive to do this above book value. Between losses on projects and share issuance below book value, it seems highly likely that book value per share will erode over time. And if book value per share erodes, so eventually must dividends per share, since the dividends are derived from asset ownership.

The second implication of the homebuilder right to walk is that Millrose has a potential cash flow timing mismatch which could be quite serious if there is a downturn or dislocation in the housing market. If homebuilders stop exercising their options, the cash inflow from selling land + horizontal infrastructure stops. However Millrose does not control the timing of the cash outflow to build horizontal infrastructure: if the homebuilder wants it built, Millrose has to fund it. Millrose could therefore face a situation where billions of dollars flow out but not in.

Millrose’s dividend is exposed to both housing market conditions and interest rates, which are likely to be somewhat correlated. Under some market conditions Lennar can suspend monthly option payments or reduce them by 50%. Other homebuilders may simply stop paying their option premiums. In addition, Millrose competes with alternative sources of finance. If interest rates fall, option premium rates on new deals must fall to remain competitive with debt finance. Millrose’s P&L is operationally and financially levered, so any reduction in option premiums will drive an even greater reduction in the dividend. Millrose’s dividend might behave more like a levered floating rate bond coupon than an equity dividend.

Finally, Millrose’s share structure means the company will never be subject to shareholder activism or sold to the highest bidder. Millrose has two classes of stock. The A shares trade publicly and have 1 vote per share. The B shares are 99% owned by the Miller family, which founded, runs, and effectively controls Lennar. Currently the B’s have 10 votes per share and 43% of the total votes. No matter how many A shares are issued, the B’s will never drop below 35% of total votes. Any amendment to Millrose’s Charter requires a 2/3rds supermajority, meaning in effect that the B’s have a veto. The Charter forbids anyone from owning more than 9% of Millrose unless exempted by the board - and anyway the B shares have a collective veto over various things, including a sale of the company.

Share

Financial engineering

The comparison between the finance Millrose offers and the finance a lender offers is stark.

Option yields (premiums/land value) are 8-11%. Compared to debt, this seems like an expensive way to fund a land bank. However, the spread is significantly lower once you include Millrose’s significant investment in horizontal infrastructure, and it is arguably not adequate compensation for the additional risk Millrose takes compared to a lender.

Subscribe now

How it could go wrong:

1. Kennedy Lewis grow Millrose as fast as possible, levering up and issuing shares to buy every land bank a homebuilder offers them without sufficient concern for downside risk.

2. The economy turns down and interest rates follow.

3. Some homebuilders stop exercising option contracts but Millrose still has to fund horizontal developments at the projects that are moving ahead. Cash makes a giant sucking sound.

4. To fund itself, Millrose sells land for whatever it can get. The company doesn’t take impairments, but it becomes clear its assets aren’t worth their marks. With leverage, mark-to-market book value is seriously impaired.

5. Lennar reduces its options payments by half. Rates come down so new deals, if there are any, are done at low yields. Millrose’s management fee and debt interest are fixed, however, so net income basically evaporates. The dividend is cut to 0. Panicked retirees rush for an increasingly crowded exit.

6. The stock drops 90%. Lennar offer a 20% premium to buy their land bank back for a song. There’s no counterbid because the Class B shares would veto it. Desperate shareholders vote for the deal.

This isn’t a prediction. But it wouldn’t be the first time, either.

Share

Conclusion

Millrose yields 10%, but can’t grow value per share because it can’t retain profits. In fact, value per share is likely to decline in the long run because the manager has a strong incentive to issue shares below book value and because the company cannot offset the inevitable losses on some deals with gains on others. So the upside appears capped at +/- 10% per year, and that might be generous.

The potential downside, however, appears significant. There are protections in place such as pooling but the structure seems prone to a cashflow crisis, balance sheet impairments, and/or significant dividend reductions. I am not certain that any of these things will happen, but if they do, they may all happen at once.

On top of that, neither the manager nor the owner of supervoting shares seem well-aligned with Class A shareholders.

Contrast the risk/reward with Berkshire Hathaway, my go-to comparison stock. In my estimation Berkshire has a good chance of compounding intrinsic value at +/- 10% over time, so the upside is similar, but the chance of Berkshire becoming seriously impaired is vanishingly small. In fact, at just the sort of time when Millrose might become impaired, Berkshire is likely to be increasing its intrinsic value by buying assets on the cheap.

Also, a big part of investing is positioning yourself to make good decisions in tough times. With Berkshire, I know I won’t panic sell at the wrong time. With Millrose, I don’t.

There is only one way to find out whether this company will do well in a major downcycle: wait for one. I am happy not to own it while I do that.

Subscribe now

Links to previous Reviews

1. IRSA

2. Brookfield

3. Cheniere

4. Uber

5. Howard Hughes Holdings

6. My notes

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Uber Freight has two parts:

1. Digital freight brokerage was launched in 2017 and operates as a marketplace; shippers post loads and carriers accept them through the Uber Freight app. Uber earns the spread between what it charges shippers and what it pays carriers, so it is exposed to spot pricing cycles.

2. Managed transport is a SAAS+services business which manages the entire carrier network for large shippers. This is deeply embedded into sticky customers and is insulated from spot pricing cycles.

Subscribe now

Uber Freight’s right to win

Freight is a large and potentially attractive business, but it’s highly competitive with established scale players.

Uber’s 10k states: “We believe that Freight is revolutionizing the logistics industry. Freight powers a managed transportation and logistics network and connects Shippers and Carriers in a digital marketplace to move shipments while leveraging our proprietary technology, brand awareness, and experience revolutionizing industries. Freight provides an on-demand platform to automate and accelerate logistics transactions end-to-end while providing visibility and control of logistics networks. Freight connects Carriers with Shippers’ shipments available on our platform, and gives Carriers upfront, transparent pricing and the ability to book a shipment with the touch of a button. Freight serves Shippers ranging from small- and medium-sized businesses to global enterprises. By leveraging logistics solutions expertise and value-add solutions, Freight enables Shippers to create and tender shipments, secure capacity on demand with real-time pricing, and track those shipments from pickup to delivery. Freight operations are principally based in North America and Europe. We believe that all of these factors represent significant efficiency improvements over traditional transportation management and freight brokerage providers.”

“Revolutionising.”

“Significant efficiency improvements.”

This sounds good.

Is it reflected in the evidence?

Not really:

Now in fairness, freight is a cyclical industry. 2023 saw a post-covid boom in shipping rates that boosted revenue (but apparently not ebitda). 2023/4/5 saw a predictable hangover which may be masking progress in revenue growth and margins at Uber Freight. CH Robinson is a competitor in the brokerage industry and shows a similar cyclical trend:

Three things strike me from these figures:

1. Even allowing for the cycle, Uber Freight shows no obvious evidence of being a disruptive technology that’s “revolutionising” an industry.

2. Uber Freight appears to be subscale, at least compared to CH Robinson.

3. CH Robinson is far more profitable than Uber Freight, despite focussing on brokerage, supposedly the lower-quality part of the business.

Share

Going deeper

It is tempting to point to obvious inefficiencies in logistics and conclude that technology - and specifically, Uber’s technology - can solve them. However:

• Uber Freight is not uniquely positioned to address inefficiencies in logistics - as I discuss below, there are multiple large players with decades of data and plenty of capital who are building solutions, too.

• Some of the inefficiency is structural, creating the illusion of opportunity where in fact there is none. We will never get to the point where every truck is full all the time and no load ever has to wait. It is physically impossible.

Brokers match people who need something moved to people who have a truck. In theory there is a network effect here: the bigger the network, the easier it is for the broker to minimise inefficiencies such as half-empty runs and long journeys to collect loads. However there are no barriers to entry in freight brokerage, and there is no evidence to suggest that network effects are strong enough to drive winner-take-most dynamics. There are an estimated 16,000 brokers in the US. The top 10 firms control about a third of the market. Beyond that, brokerage is wildly fragmented. Even tiny players can build local networks strong enough to keep them alive. In addition there are “load boards”, neutral marketplaces where loads can be posted and accepted, which allow small brokers to widen their reach and also allow shippers and truckers to find each other without brokers at all.

The big brokers (CH Robinson, RXO, J.B.Hunt, etc.) are large, well-capitalised, and very capable of keeping up in the technology race. The same goes for the dominant load board (DAT One, owned by Roper Technologies). The digital native insurgents that tried to dislodge the leaders in the 2010s have died - Convoy was probably best in class and collapsed in 2023. It is not obvious to me why Uber would have transformationally better tech than any of the other big players. Uber might take share from the tail, but CH Robinson has been doing that for decades and still “only” has 17% share.

In managed transport, while I can accept that customers are sticky, the same argument applies: I see no reason why Uber will be transformationally better than the competition, so while it might keep its existing customers, I do not understand why it should take share or have real pricing power. In addition we have no idea whether Uber is profitable within this segment. If it is, then brokerage is losing money hand over fist, which suggests brokerage is a much worse business than its peer CH Robinson.

I accept that the industry is in a downturn and Uber Freight have spent several years integrating an acquisition and rebuilding their technology. The business may well perform better in the next cycle. But have they built a platform for dominance? I don’t see it.

Subscribe now

Where it gets interesting

There is one area where Uber Freight can genuinely do something different. By linking with Uber’s gig-economy driver network, Uber can add last-mile to its Freight offering. The long-run vision is a single platform that orchestrates first-mile, long-haul, and last-mile delivery across transport modes and geographies. This is virtually impossible for other freight brokers to replicate.

The problem is, I am not sure how many customers really need freight and last mile in the same package. Most users of freight are shipping from a factory to a warehouse or a retailer. And if they do need last-mile delivery direct to the customer, why would they choose Uber over or Amazon, which already offers consumer discovery, factory-to-doorstep logistics, trust, and returns all in one service?

That said, I do think there is a logic here. Direct to consumer is growing. By combining discovery and logistics, Uber can be a part of that trend. But a big, profitable part? Uber Freight has a lot to prove before we can conclude that.

Share

Conclusion

Uber Freight operates in a large market where others have built profitable, capital light businesses. It may well become one of those. But there is very little evidence that it is a disruptive force, able to deliver materially greater efficiencies than competitors and take market share.

As an Uber shareholder I would absolutely love to be wrong on this. Do you think I am? Persuade me!

Subscribe now

Links to previous work

1. IRSA

2. Brookfield

3. Cheniere

4. Uber

5. Howard Hughes Holdings

6. My notes

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Share

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

BN

• FY25

  • DE $6bn, DEBR $5.4bn = $2.27/share, +11%. DEBR CAGR since 2021: 18%.

  • Raised $112bn, financed $175bn, sold $91bn, and deployed $126bn.

  • Insurance assets now $145bn.

  • Operating business DE $1.6bn.

  • $11.6bn unrealised carried interest. Will realise significant amounts in next 3 years.

  • Bought back over $1bn at $36 per share.

  • Issued CAD1bn of 7- and 30-year notes.

  • 17% increase in the dividend to 28c/year.

• “Splitting market capitalisations” across multiple securities doesn’t work in the age of indexing. Now streamlining. “Initial step” was merging Business Partners and Corp. Next they will merge the insurance entity into BN, which will allow insurance to fully benefit from BN’s capital base and grow. Will assess other subsidiaries in the future and they are already issuing Corp shares to buy LP units in both BIP and BEP.

• Long term stock returns: 19% 30y CAGR; $1m turned into $285m. Only works if you pick a good business, run it well, and don’t interrupt compounding, which requires excess capital and a flexible balance sheet.

• Business relationships are “increasingly with the best”: Google, NVIDIA, JP Morgan, Microsoft, the US government.

• Liquidity has returned in both debt and equity markets and interest rates are starting to come down, which “will do wonders” for the US economy.

• Real Estate: with limited new supply and growing demand, asset values are “set to rise substantially”.

  • Sold $24bn in 2025, invested $33bn, financed $42bn.

  • FFO catching up to NOI on higher rents, tighter financing spreads, and some deleverage. Also, 20-25% of the debt is floating so 25bps lower rates adds $35m FFO.

  • Signed 17msf of office leases in 2025 at 18% spreads in super core and core plus. Virtually no new office supply coming in gateway cities globally, several of which are seeing prime office rents 50% above pre-covid levels.

  • “Capital markets get stronger by the day. Those really involved in the business [are starting] to appreciate what’s going on in the office market...I think as that…broadens, you’re going to see transaction activity really pick up…[and] we will be poised to monetize a number of assets”.

  • Have derisked the NA resi business by selling MPCs; now more capital light.

• Wealth Solutions:

  • DE $1.7bn, up 24%. 2.25% spread, 15% ROE.

  • DE will exceed $2bn in 2026 on $20bn of capital.

  • Have worked hard to diversify the liability side across product types and geographies so they can allocate capital to wherever the cost of funding is lowest. Pricing can move fast.

  • In the UK, £500bn of pensions will come to the risk transfer market over the next 10 years.

  • Growing footprint in Asia where demographics make savings products key.

    • $3tn of life and savings insurance on Japanese insurer balance sheets alone. After 3 years of work they have strong relationships with half a dozen insurers and first deal done.

    • Asia Pac has broad trends towards investing in financial assets - building relationships.

  • P&C business delivers $8bn of float at “virtually no” cost, and should make underwriting profits in future. Will grow organically and buy other assets cheap in down markets, with a path to $20-25bn of float by the end of the decade. P&C is less competitive than annuities.

BAM

• FY:

  • Raised $112bn, invested $66bn, sold $50bn.

  • FBC +12%, FRE +22%, DE +14%. FRE up 28% in the q.

  • Issued $500m of 5 year notes at 4.65% and $400m of 10-year notes at 5.3%.

  • Dividend +15%.

• 2026 looks very strong

  • Oaktree, Just Group, and acquisitions in Q4 will alone add $200m to FRE.

  • Expect a “further step change” in fundraising and likely dealmaking.

  • Closed flagship RE and Transition funds, giving dry powder to deploy at a very attractive point in the RE cycle.

  • Launched or launching new infrastructure and PE flagships.

  • Operating leverage is driving margins up in each business but consolidated margins will fall on

    • Acquisition of Oaktree. Oaktree is structurally lower margin, and is at a (counter)cyclical low.

    • Breakout of partner manager economics. BAM will now report their revenues and costs, not just FRE, and they are currently subscale and low margin, but rising.

• By broadening the platform, “we have built a business that can raise capital more consistently and deliver an earnings profile that is more predictable, more resilient, and better positioned to grow across economic cycles”.

  • 90% of 2025 fundraising was non-flagship.

  • In 4q they raised $35bn across 50 strategies; in 2026 they will raise across 60 strategies compared to 4 10 years ago.

  • They have 2,500 institutional relationships, up 10x over 10 years, plus 70k clients in private wealth and 800k in insurance.

  • Wealth is “an absolutely amazing opportunity” in terms of potential scale across 3 channels: retail/HNW, insurance policy/annuity, and 401(k)/retiree benefit.

• “A step change in growth is emerging across our infrastructure and private equity platforms”.

  • Strong investor demand.

  • AI needs infrastructure.

  • PE needs change management to deal with AI. Operators will win not financial engineers. Half of Brookfield’s returns historically have been driven by operating improvements, “a real differentiator”.

  • Unlike many, they have returned $10bn of PE investor money over the last 2y.

• AI is a strong net positive for the business.

  • Very little exposure to things that might be disrupted (software), much exposure to things that might benefit (industrials), and ability to deploy capital against long term contracts with world class counterparties (datacentres).

  • Huge client appetite for AI - first fund has $5bn committed of $10bn target, likely >$20bn with co-invest.

  • “There is no question. The bottleneck to AI growth today is not capital. It is not demand. It is electricity supply.”

• Private credit demand remains very robust, driven by huge capital requirements to build out assets.

  • Demand outweighs supply in real asset and asset-backed credit.

  • Spreads are very tight in commoditised areas, but distress here is driving increased activity in opportunistic credit.

• Teskey becomes CEO, formalising a change 4 years in the works. Flatt remains Chairman, and BN CEO.

• Investments in partner managers are likely to slow but secondaries would be near the top of the list if something comes along.

BIP

• FFO $2.6bn, $3.32 per unit, up 10% lfl. Distribution up 6%, 17th year over 5%. Payout 66%.

• Sold $3.1bn, invested $2.2bn including $1.5bn inorganic, completed $16bn of financings.

• On the NYSE, BIP has delivered 14% total returns since inception; higher for BIPC and on the TSX.

• 3 D’s “are driving an infrastructure investment super cycle that is broadening in both scope and scale”.

• Backlog is $3.9bn for Intel and $5.3bn for everything else. Intel comes online late this year; the rest commissions at a rate of $1.5-2bn per year. Should support a return to 10%+ FFO growth.

• Step change in data business in 2025.

  • More to come in 2026, with significant take-or-pay contracts signed on existing capacity, high-return densification projects, and lease-up of their land bank for development.

  • Develop datacentres at 9-10% yield on cost, and sell for 5.5-6%. Development leverage is 70% so this translates to high teens RoEs, or in the 20’s when everything goes right.

• Agreed to sell 1 of their 4 Brazilian transmission concessions for $150m net to BIP, a 45% IRR, and an 8x MM. Will close 1q26.

• A more pro-market energy policy for Alberta could boost their Midstream business.

• Issuing equity in BPIC to buy back BIP 1-1.

Subscribe now

BBU

• FY LFL ebitda up 5% to $2.1bn. EFO $1.2bn.

  • Sold $2bn, repaid $1bn of corporate borrowing, bought 4 new investments for $700m, and repurchased $235m of stock at $26 per share, a 50% discount to their view of fair value.

  • $20bn of refinancings at a 50bp cost saving, with spreads very tight.

• Have always focused on improving businesses. Two things make that even more important now: deglobalisation requires capital and change management, and AI beneficiaries will be those who can implement it. “This is the environment we were built for.”

• Clarios: 30% of NAV. Ebitda is up 40% since they bought it, or $700m, can repeat this over the next 5 years with investment in advanced batteries, and enhanced recycling and critical mineral recovery. Will generate $5bn of FCF over that time before tax credits, more than the $4.5bn 2025 up-financing.

• Nielsen margins are up 350bps since acquisition on $800m of cost savings.

• Evaluating IPO’ing BRK - window seems to be opening and the business is derisked and growing strongly.

• New investments

  • Fosber: equipment for corrugated packaging industry, 2/3rds aftermarket with high cost of downtime, carve-out at 10x ebitda, clear opportunities to improve performance.

  • First National: Canadian resi mortgages, asset light, $160bn of mortgages under administration.

  • Chemelex, heat management equipment, aftermarket, carve out at 11x ebitda, opportunities to improve margins.

Share

BEP

• 4q FFOPU $0.51, +14%. Raising dividend 5%, 15th year in a row.

• FY FFOPU $2.01, +10%.

  • Deployed or committed $8.9bn ($1.9bn net), sold $4.5bn ($1.3bn net) for 2.4x invested capital and returns above target, financed $37bn.

  • Issued $450m of 10-year debt in March 2025.

  • Raised $650m of new equity in November.

  • Issued CAD500m of 30-year debt in Jan 2026 at 5.2% and their lowest spread ever.

• “It is now clear that power is a strategic priority around the world and is the bottleneck to growth for both governments and corporates”, driven by electrification, industrial growth, and AI, not just replacing fossil fuels as it was a few years ago.

• Capital is an increasingly important competitive advantage and the environment for acquiring development assets is very attractive. Listed assets, carve outs from utilities, and “lower quality” developers lacking capital and operating scale but with big pipelines are particularly attractive.

• Sales proceeds will become more programmatic/recurring as they scale. Have already agreed an $860m ($210m net) sale in 2026 and a framework for future asset sales up to $1.5bn to the same buyer. Expect to sign more of these frameworks soon. They derisk capital recycling and speed up the time from development completion to sale. “I will go out on a limb and say I think this is going to be a huge differentiator for our franchise”.

• By asset type

  • Solar and onshore wind are low-cost and fast to market. Scaling development capacity from 8Gw in 2025 to 10Gw in 2027. US permitting for solar is accelerating; for onshore wind it is slowing somewhat, but still happening.

  • Have not done much offshore wind but opportunities are growing in Europe. Might include end-of contract assets than can be bought assuming merchant economics and then re-contracted.

  • “The value of hydro is being recognized more than ever before.”

  • Westinghouse: US govt commitment “provides long-term demand certainty, helping unlock supply chain investment”.

  • Battery costs are down 95% since 2010 and 60% over 2 years. In addition, over the last 2-3 years revenue models have switched from merchant to take-or-pay. Batteries are the fastest growing part of BEP today. Neoen was the largest acquisition in BEP history and included a large battery pipeline.

• Will issue of $400m BEPC shares to repurchase BEP LP units for a net reduction.

Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Share

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.

Trades

Quite a lot of activity this month as my energy positions outperformed the rest and I recycled some capital. These were mostly small trades.

• Reduced Transocean, bought more Howard Hughes, bought more Pershing Square.

• Bought Saga Plc.

• Sold Noble.

• Bought more Ashmore.

• Sold Cheniere.

• Reduced Peyto.

• Bought more Blackstone, KKR, and Apollo.

• Bought more Ashmore.

  Subscribe now

Articles

• Howard Hughes Holdings review.

• Brookfield investor day summaries.

• Uber review.

• Vail Resorts first look.

• Cheniere Energy review.

What I found interesting this month

• LongYield on Alibaba.

• Grow or Die on Uber.

• Cayucos Capital on Jardine Matheson.

• Future Cognitive Capital on Microsoft.

  Subscribe now

Thanks for reading - if you enjoyed reading this please like and restack, and do get in touch if you have questions.

Pete

Disclaimer: This post is for informational and educational purposes only. Building Arks is not licensed or regulated to provide any financial advisory service and nothing published by Building Arks should be taken as a recommendation to buy or sell securities, relied upon as financial advice, or treated as individual investment advice designed to meet your personal financial needs. You are advised to discuss your personal investment needs and options with qualified financial advisers. Building Arks uses information sources believed to be reliable, but does not guarantee the accuracy of the information in this post. The opinions expressed in this post are those of the publisher and are subject to change without notice. The publisher may or may not hold positions in the securities discussed in this post and may purchase or sell such positions without notice.