TD Cowen reiterates Buy on Microsoft stock, cites Azure growth

Microsoft reiterated a capacity-constrained outlook through at least end-2026, but management said added Azure capacity, faster Fairwater ramping, and a new Anthropic workload could support growth acceleration in 2H 2026. Copilot adoption improved last quarter, and Microsoft expects more net subscriber additions in the June quarter than the roughly 5 million added in March. TD Cowen kept a Buy rating and $540 price target, while Microsoft also added Carmine Di Sibio to its board and expanded its AI-related ecosystem updates.

Analysis

The incremental positive here is not the headline AI demand; it is the improving conversion of AI capex into monetizable software consumption. Microsoft is still effectively rationing compute, but the mix shift toward higher-value Azure workloads and agentic usage suggests operating leverage can re-accelerate even before supply is fully unconstrained. That makes this less about near-term revenue upside and more about sustained margin durability as capacity allocation tilts away from low-ROI internal usage and toward externally billable demand.

The second-order winner is NVIDIA: every meaningful increase in Microsoft’s usable capacity implies faster absorption of accelerators, networking, and ancillary data-center components. If Anthropic meaningfully scales inside Microsoft’s stack, the market is likely underestimating how diversified the large-model training/inference buyer base has become; this reduces OpenAI concentration risk and supports a longer runway for AI infrastructure spending. The loser is anyone hoping for a broad deceleration in AI capex—this read-through argues the spend is becoming more disciplined, not smaller.

The main risk is timing mismatch: the market may have already priced the 2026 acceleration while the stock is still hostage to near-term capacity constraints and product-cycle noise. If enterprise buyers continue shifting from seats to agents, reported growth can look lumpy even as underlying wallet share expands, which creates room for volatility around guidance. The contrarian angle is that the bull case is now less about software multiple expansion and more about a hidden infrastructure toll-booth model; if that thesis is right, Microsoft can keep compounding without needing heroic upside surprises, but the stock may not rerate until investors see proof that AI monetization is outpacing incremental capex.

From a trading perspective, this is a relative-value setup more than an outright momentum trade. The best risk/reward is long MSFT against a software basket or against slower AI adopters, with a 3-6 month horizon into evidence of improved Azure capacity utilization and Copilot conversion. For pure expression, buy NVIDIA on any pullback tied to concerns about Microsoft capex fatigue; the data imply the capex cycle is extending, not peaking. Near-term, upside in MSFT likely requires confirmation that enterprise agent adoption is translating into cleaner quarterly billings, not just narrative improvement.

Key Decisions for Investors

• Long MSFT / short a software basket (e.g. WDAY, CRM, NOW) for 3-6 months: express the view that Microsoft’s AI monetization is more durable and more directly tied to infrastructure leverage than peer SaaS names; target 8-12% relative outperformance, stop if Microsoft guidance implies prolonged demand normalization.
• Buy NVDA on 5-8% pullbacks over the next 1-2 months: Microsoft’s ability to absorb more compute and diversify AI workloads supports a longer capex runway; risk/reward favors scaling in rather than chasing strength.
• Add MSFT call spreads 4-6 months out: use upside structures to capture a re-rate if the market starts rewarding improved Azure utilization and Copilot conversion, while capping downside from capacity-related volatility.
• Pair long MSFT / short ORCL for 3 months: Microsoft is better positioned to convert AI infrastructure into recurring revenue, while Oracle is more exposed to headline-driven AI enthusiasm; this is a cleaner quality-vs-trading multiple expression.
• Avoid outright shorting MSFT into weakness: the bearish case is now dependent on a multi-quarter capex digestion narrative that is less likely if enterprise agent adoption keeps improving; downside should be expressed via hedges, not naked shorts.