Microsoft just dropped its Q4 FY2026 earnings, and the numbers are massive. Quarterly revenue hit $90 billion, up 18% year over year, with net income of $35.8 billion — a 31% jump. Both figures blew past Wall Street expectations, and shares popped over 8% in after-hours trading.
The growth is largely powered by cloud and AI. Azure revenue grew 43%, beating estimates, and for the first time ever, Azure’s full-year revenue cracked $100 billion. Microsoft 365 Copilot paid seats surged to 30 million, up from 20 million last quarter.
But Microsoft also revealed how it's keeping AI costs in check. The company decoupled its models from the runtime, context, memory, and tooling systems, making it easy to swap underlying models on the fly. Routine tasks are handled by smaller, specialized models; only the toughest challenges get kicked up to large models like GPT-5.4.
Microsoft says this approach can cut GPU costs by 50% to 90% without sacrificing quality. PowerPoint image generation costs dropped as much as 84%, and Dynamics 365 speech costs fell up to 89%. That helps Microsoft absorb its growing AI compute spending — and reduces its dependence on any single model supplier.