Microsoft Corporation (MSFT)

News Summary

• Azure grew 39% in constant currency in Q4 FY25, significantly beating expectations, with Microsoft explicitly capacity-constrained through at least H1 FY26; Q1 FY26 Azure growth is guided at approximately 37% constant currency, signaling sustained hyperscaler AI infrastructure demand that directly drives GPU procurement cycles relevant to Cerebras.
• Microsoft guided Q1 FY26 CapEx at over $30 billion, anchored by $368 billion in contracted backlog, with full-year FY26 CapEx growth moderating but H1-weighted — the scale and durability of this spend underscores the competitive intensity of the AI infrastructure buildout Cerebras is targeting.
• Satya Nadella cited a 90% improvement in tokens-per-GPU on GPT-4o through software optimization alone, a direct signal that hyperscalers are aggressively pursuing inference efficiency gains at the software layer, which compresses effective compute costs and raises the bar for hardware differentiation.
• Azure AI Foundry Agent Service reached 14,000 customers building agents, with Foundry APIs processing over 500 trillion tokens in FY25 — up more than 7x year-over-year — indicating that agentic workloads are scaling rapidly and broadening the base of AI inference demand beyond a handful of flagship applications.
• Microsoft stood up over 2 gigawatts of new data center capacity in the past 12 months across 400-plus facilities in 70 regions, making all Azure regions liquid-cooling capable — a scale and density posture that sets the infrastructure benchmark Cerebras must engage with as it pursues hyperscaler and enterprise AI hardware deals.

Financial Highlights

Earnings Call Highlights

Microsoft's FQ4 2025 call was a triumphant, confidence-saturated affair with management projecting expansive ambition rather than defensive caution — Azure grew 39% and beat significantly, commercial bookings crossed $100 billion for the first time, and Satya Nadella framed the AI infrastructure buildout as a multi-decade compounding platform play rather than a cyclical spend. Analyst tone was uniformly celebratory, with multiple analysts calling it the best quarter they had seen in years of coverage, and management used that goodwill to double down on aggressive CapEx guidance — over $30 billion in Q1 FY26 — without apology. The dominant theme was that AI demand is outpacing supply, that Microsoft remains capacity-constrained through at least H1 FY26, and that the software optimization layer (not just hardware) is the key differentiator. Amy Hood explicitly pushed back on attempts to identify a CapEx-to-revenue crossover point, signaling that Microsoft sees no reason to slow infrastructure investment given $368 billion in contracted backlog.

• Azure grew 39% in constant currency, 'significantly ahead of expectations,' driven by accelerated core infrastructure demand from largest customers; Microsoft explicitly stated demand remains higher than supply and expects to be capacity-constrained through H1 FY26, with Q1 FY26 Azure growth guided at approximately 37% constant currency.
• Capital expenditure guidance for Q1 FY26 is 'over $30 billion,' with full-year FY26 CapEx growth moderating versus FY25 but H1 weighted higher due to finance lease timing; Amy Hood framed this as directly correlated to $368 billion in contracted backlog, signaling durable GPU and CPU server demand.
• Satya Nadella cited a concrete AI inference efficiency milestone: through software optimizations alone on the GPT-4o model family, Microsoft is delivering 90% more tokens per GPU versus a year ago — a direct signal that software-layer yield improvements are compressing effective compute costs independent of hardware upgrades.
• Azure AI Foundry Agent Service has reached 14,000 customers building agents, and Foundry APIs processed over 500 trillion tokens in FY25, up more than 7x year-over-year — management cited this as evidence of broad platform diffusion beyond a handful of head applications, which is the metric they track most closely for durable AI infrastructure demand.
• Microsoft stood up more than 2 gigawatts of new data center capacity over the past 12 months, now operates 400-plus data centers across 70 regions, and has made all Azure regions liquid-cooling capable — framing this as a fungibility and flexibility advantage for AI workload density that competitors have not matched.