Microsoft Corporation

Big picture performance (good + encouraging demand signals) - FY2026 closed $331B revenue (+18% YoY); Microsoft Cloud $214B (+27%); Azure $100B (+41%). - This quarter: revenue $90B (+18% YoY; +17% constant currency); operating income +18%; EPS $4.74 (+23%) (notably adjusted for OpenAI investment impact). - Cloud billings/collections strong: Cash flow from operations $55.4B (+30%); Free cash flow $19.6B, though still impacted by higher CapEx.

AI platform & monetization momentum (major opportunity) - Capacity scaling: 31 new data centers this quarter (total 88 for the year), plus ~1 additional gigawatt added; management claims dock-to-live for new GPUs ~50% faster and targets ~2x overall capacity in ~2 years. - Product stack traction: - Copilot: 30M+ paid M365 Copilot seats (net adds more than doubled QoQ); engagement and “high usage” time dropped from months to days over the past year. - Foundry (app/agent stack): 100,000 Foundry customers; revenue >2x YoY; 1T token annualized run rate increased 4x YoY. - Agent governance: Agent 365 has ~40M agents registered across tens of thousands of companies (started ~2 months ago). - Developer ecosystem (GitHub Copilot): 225M users total; 50M Copilot users; Copilot revenue +60% QoQ with notable improvement after pricing change. - Efficiency gains: examples include 4x higher throughput for Copilot workloads since start of year, and multiple “Flash” model claims (token/GPU cost reductions).

Strategic bet: “sovereign + substitutable models + enterprise control of IP” (surprising direction) - Microsoft emphasizes model flexibility: “model choice” across 11,000+ models, plus a substitutable model system to improve cost-to-outcome and resilience (business continuity). - AI sovereignty: partnership with Mistral to run models in Microsoft Sovereign Cloud, including public, customer-controlled, and fully disconnected environments. - Open vs custom model traction (from Q&A): Satya framed it as enterprises needing control of token/learning capacity, implying open models can gain traction but are evaluated based on outcomes and governance—not model ideology.