Microsoft to rely more on homegrown AI chips, reducing dependence on Nvidia
- ID
- 34473
- Status
- summarized
- Published
- 12 Oct 2026, 10:04 PM
- Fetched
- 12 Oct 2026, 10:28 PM
- Provider
- CNBC Technology
- Category
- technology
- Original URL
- https://www.cnbc.com/2026/10/12/microsoft-will-mainly-use-its-maia-chips-for-ai-inference.html
- Source URL
- https://www.cnbc.com/id/19854910/device/rss/rss.html
Summary
- Score
- 5.0
- Created
- 12 Oct 2026, 10:29 PM
- Tags
- Audience
- developersai_ml_learnersai_agent_userssaas_startup_founders
What happened
Microsoft said it will eventually have multiple gigawatts of its Maia AI chips and intends Maia to be its primary architecture for running AI models. The third-generation Maia accelerator was unveiled Monday, and the next-generation Maia 300 chip is scheduled to enter production in 2027. Microsoft will mainly use its own silicon for incoming AI requests, reducing its reliance on Nvidia GPUs in Azure and Copilot services.
Why it matters
For builders using Azure AI, GitHub Copilot, or Microsoft 365 Copilot, the only concrete date here is Maia 300 production in 2027; there is no stated pricing, benchmark, API-compatibility, or regional-availability change. Treat this as a roadmap signal to watch, not an immediate reason to re-architect or switch cloud AI vendors, and ask Microsoft for inference benchmarks and Malaysia/Azure-region availability before making cost decisions.
Discussion angle
What evidence would make Maia a credible default for your inference workloads—published per-token pricing, latency/throughput benchmarks, or availability in your Azure region? The article provides none of those yet.