Five Things Microsoft Bulls Are Overlooking

Free cash flow dropped from $74.1 billion to $71.6 billion year-on-year. A 3.4% decline sounds trivial. It is not. This is the first FCF decline Microsoft has posted in over five years, and it occurred despite revenue growing 15%. The culprit is capex — specifically, the $50+ billion annual commitment to data centre infrastructure for Azure and AI workloads.

Management frames this as investment in future growth. We agree. But the capital allocation question is whether the return on this incremental capex will match the returns Microsoft historically generated. Azure's incremental revenue per dollar of capex has been declining since 2023, which suggests the easy capacity additions are behind us. Every new data centre competes with hyperscaler builds from Amazon and Google, driving construction costs and power costs higher.

This dynamic was already visible. The trend has accelerated since then, and the capex guidance for fiscal 2026 suggests another step-up.

Microsoft 365 Copilot is the highest-profile AI monetisation bet in enterprise software. At $30 per user per month, the revenue potential is enormous — if adoption scales. Early data from enterprise deployments suggests usage rates of 30-40% among licensed users, with significant variation by company size and industry.

The risk is not that Copilot fails. The risk is that it works but does not justify the premium. If enterprises conclude that Copilot delivers 10-15% productivity gains rather than 30-40%, the pricing power erodes. We have seen this pattern in previous enterprise software rollouts — the pilot phase generates enthusiasm, the broad rollout reveals that actual usage concentrates among power users while casual users do not engage enough to justify the per-seat cost.

The Activision Blizzard acquisition closed for $69 billion — the largest deal in gaming history. Microsoft now operates the third-largest gaming business globally by revenue. The strategic logic — combining Game Pass subscriptions with the Activision catalogue — is sound in theory.

In practice, gaming EBITDA margins have been compressing industry-wide as content costs rise and the subscription model cannibalises higher-margin full-price sales. Microsoft does not break out gaming profitability in sufficient detail to assess the return on the $69 billion investment. When a company spends $69 billion and then reduces disclosure on the acquired segment, it is, frankly, not a great signal.

Azure revenue growth has been the primary stock catalyst for five years. But Azure is now a $100+ billion run-rate business. Growing a $100 billion business at 30% means adding $30 billion in annual revenue — the equivalent of creating a new Adobe every year. The maths gets progressively harder.

Amazon's AWS experienced the same deceleration curve from 2019 to 2023, and the stock underperformed during the transition. Microsoft bulls argue that AI workloads create a new growth vector. They are right — but the AI revenue sits on top of the existing base, and the capital intensity of AI infrastructure is materially higher than traditional cloud compute.

Operating margins of 47.1% are at or near the structural ceiling for a diversified software company. The gaming acquisition added lower-margin revenue. The AI capex cycle will pressure margins as depreciation charges ramp. And the rising proportion of consumption-based Azure revenue (versus subscription-based Office 365) introduces more variability into the margin profile.

We estimate operating margins will settle in the 43-45% range over the next two to three years — still excellent, but a 200-400 basis point compression from current levels. On $281.7 billion in revenue, each 100 basis points of margin compression costs $2.8 billion in operating income.