Microsoft is stealing Azure GPUs to feed its own AI — and leasing from AWS to survive

Microsoft's $190 billion AI capex is cannibalizing its own cloud.

CFO Amy Hood admitted Azure would have grown over 40% instead of 39% if the newest GPUs had gone to Azure customers instead of Microsoft's own AI products.

That is the world's largest cloud provider telling paying customers: you come second.

The numbers are stark. Stock down 24% year-over-year. Copilot still lags ChatGPT and Claude. GitHub facing repeated outages and pressure from Cursor and Claude Code.

So what does Microsoft do? They go shopping for capacity everywhere — including from Amazon and Google.

Amazon already bailed Microsoft out after a series of GitHub outages. The company explored leasing Oracle cloud infrastructure but walked away over security concerns.

This is not a capacity story. This is a priority story.

Microsoft has decided that its own AI products matter more than the enterprises paying for Azure. Every CISO and cloud architect running on Azure just learned where they sit in the queue.

Audit your cloud capacity agreements today. If your SLA assumes limitless Azure scale, that assumption just broke. Start negotiating multi-cloud failover now — before your next outage becomes Microsoft's next headline.

SOURCE: https://www.businessinsider.com/inside-nadella-hardest-year-microsoft-ai-azure-github-2026-7
VERIFIED: Business Insider, AI Weekly, Microsoft earnings call transcript
SIGNAL: The world's largest cloud provider is capacity-constrained and prioritizing internal AI over paying customers. Every enterprise on Azure needs a multi-cloud contingency plan.