Original review: Part 1, Part 2, Part 3.
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.
Key takeaways
Revenue +19%, EBIT +22%, operating EPS +19%.
Operating leverage more than offsets increased depreciation.
New segment reporting: Agents and Infrastructure, which is the core, grew Revenue and EBIT 25%.
Minor red flag: no updated Copilot paid seats disclosure.
Call highlights
Demand exceeds supply. Variability in quarterly growth rates is therefore driven by volatility in token pricing, GPU delivery and build out timing, the extent to which efficiencies release extra capacity for sale in the quarter, and ARPU changes from model mix. (My view: none of these really matter over time.)
Enterprise adoption of AI is broadening. Microsoft is “helping every organization stand up their own infinite SaaS factory”.
Competition between models is increasing. Harnesses that route tasks to the optimal model force models to compete for every task, not once a quarter.
More and more customers are using open source, and Microsoft’s own models now target being 3-6 months behind the frontier at much lower cost.
Consumption-based pricing brings multiple additional enterprise cost lines into Microsoft’s TAM, but also make sales motions more complex.
Copilot is getting a lot of developer attention and becoming far more capable. This is important as it leverages Microsoft’s distribution advantage.
Arguably the big positive surprise of the July quarter was the explosive growth in Copilot paid seats. It’s disappointing they did not update this disclosure, and even more disappointing that analysts did not ask why. Did growth slow? Revenue numbers suggest not. But it is something to watch, because Copilot is how Microsoft puts commoditised intelligence on the desk of every knowledge worker.
Long term economics
I think of Microsoft as earning profits in two buckets:
Bucket 1: selling compute, which is commoditised and capital intensive but demand currently exceeds supply so returns are strong.
Bucket 2: selling the tech stack that lets customers apply AI efficiently and safely, which is not commoditised and offers strong operating leverage as demand grows.
This is key for understanding what happens when compute supply exceeds demand and the ROI on capex falls. I think this might be a very long way off: I think we are still right at the start of AI adoption. But long history tells us that when the barrier to capacity expansion is mainly capital, eventually supply will eventually exceed demand.
The market worries that when this happens, the ROI on capex will collapse. I think the market has it backwards. I think that’s the best thing that could happen.