Microsoft tops Q3 estimates, says AI business up 123% year over year

Microsoft reported Q3 EPS of $4.27 on revenue of $82.89B, ahead of consensus estimates of $4.04 and $81.46B, while its AI business reached a $37B annual run rate, up 123% year over year. Copilot paid seats exceeded 20 million, and the company lifted remaining performance obligations to $627B, although the stock fell more than 1% after initially rising. Microsoft also said it reworked its OpenAI agreement, removing revenue-sharing payments while losing exclusive access to OpenAI intellectual property and models.

Analysis

The market is reading this as a clean beat, but the more important signal is that Microsoft is trying to decouple Azure economics from OpenAI dependence while still keeping the strategic upside. Removing revenue-share leakage improves long-duration monetization, but losing exclusivity lowers the probability that Azure captures the full incremental AI training and inference spend; that is a subtle negative for cloud share quality even if near-term guidance remains intact. In other words, Microsoft is transitioning from a privileged AI landlord to a platform toll collector, which is a better business model than headline partnership optics suggest but a slightly worse one than the market previously capitalized.

The real second-order issue is capex intensity versus monetization velocity. At this scale of spending, the stock needs evidence that AI dollars are converting into higher mix revenue, not just higher asset base and depreciation drag over the next 4-8 quarters. If enterprise AI seat expansion or RPO growth slows even modestly, the market will start questioning whether returns on incremental GPU/CPU investment are compressing faster than the AI revenue line is accelerating.

For competitors, the loosening of OpenAI exclusivity reduces Microsoft’s moat and marginally improves the negotiating position of alternate clouds and model providers. That creates a broader redistribution of AI workloads across hyperscalers, which should help Amazon and Google capture incremental share over the next 6-12 months if they can price aggressively and secure model partners. For PC supply chain names, the memory shortage is a near-term margin tailwind for component vendors but a demand headwind for low-end OEMs, suggesting a bifurcation rather than a broad PC recovery.

The contrarian view is that the post-earnings dip may be too small if investors have been treating AI monetization as a one-way call option. The cleaner way to express the trade is to own the AI capex beneficiaries and hedge the platform premium risk, because Microsoft now looks more like a high-quality compounder with moderating exclusivity benefits than an unassailable AI monopolist.

Key Decisions for Investors

• Trim MSFT on strength over the next 1-2 weeks; the setup is now better for a hold than an outright add because the market may need proof that AI revenue growth offsets lower exclusivity and heavier capex intensity.
• Initiate a pair trade: long AMZN / short MSFT for a 3-6 month horizon, betting that cloud workload diversification and model optionality improve Amazon’s relative AI share capture as Microsoft’s OpenAI moat narrows.
• Add to NVDA on any AI-related cloud pullback, since multi-cloud competition should keep GPU demand elevated even if Microsoft no longer captures the full OpenAI funnel; use a 2-4 month horizon and size against cyclical volatility.
• Consider short-dated MSFT puts or put spreads into the next 4-8 weeks if the stock re-rates back above pre-earnings levels, as the market may be underpricing the longer-term dilution from reduced exclusivity.
• For the PC memory squeeze, prefer component exposure over OEMs: long MU / short lower-end PC assemblers for 1-2 quarters, expecting pricing power to accrue to memory suppliers while subscale PC vendors absorb margin pressure.