# Microsoft's Maia 300 rollout

AI & tech · Heat 1.1 / 10 · 5 shows · 22m airtime

Microsoft announced on October 12, 2026 that it will deploy its homegrown Maia 300 AI inference chip at gigawatt scale starting next year. The chip is built on TSMC's N2P node and delivers up to 7x the rack-scale token throughput of the Maia 200 at 96 accelerators, with Microsoft claiming 40% more tokens per megawatt than leading GPU systems. Nadella announced it himself, calling it Microsoft's biggest bet on custom silicon, and production starts in 2027. Microsoft will use it for internal workloads like Copilot and GitHub and will not rent the compute to third parties. Reporting since says Microsoft tried to convince Anthropic to become a Maia customer and that no Azure customer has signed on to rent the…

Page: https://boilpoints.com/point/microsofts_maia_300_rollout_2267536040

## Segments

- [Bloomberg Tech](https://boilpoints.com/u/bloombergtech), 2026-10-12, 1:23 to 4:18: **Walks through the third-generation Maia chip**. Bloomberg's Matt Day walks through Microsoft's third-generation Maia chip: gigawatt-scale deployment starting next year, 40% more tokens per megawatt than leading GPU systems, internal-only use for Copilot and GitHub, and no third-party rental. Sets it against Google's TPU head start and Amazon's silicon, and notes Microsoft's roughly 12 gigawatts of capacity today. New angle: Microsoft's Maia 300 promises 40% more tokens per megawatt than leading GPU systems and will not be rented to third parties.
- [Tech Brew Ride Home](https://boilpoints.com/u/techbrewridehome), 2026-10-12, 5:36 to 7:56: **Details the Maia 300 silicon**. Goes deeper on the Maia 300 silicon than the announcement coverage: TSMC N2P process node, a chiplet architecture splitting compute, memory controller and I/O, and up to 7x the rack-scale token throughput of Maia 200 in a 96-accelerator configuration. New angle: The Maia 300 is built on TSMC's N2P node and delivers up to 7x Maia 200's rack-scale token throughput at 96 accelerators.
- [The Information's TITV](https://boilpoints.com/u/theinformationstitv), 2026-10-12, 1:25 to 13:10: **Adds Microsoft's Anthropic pitch for Maia**. The reporter behind the August Maia scoop walks through the confirmation and adds what the announcement left out: Microsoft tried to land Anthropic as a Maia customer over the summer, and no Azure customer has signed on to rent the chips. New angle: Microsoft tried to convince Anthropic to become a Maia customer; no Azure customer has signed on to rent the chips.
- [Closing Bell](https://boilpoints.com/u/closingbell), 2026-10-12, 21:08 to 24:40: **Reports the Maya 300 claims and timeline**. Kate Rooney reports Microsoft's third-generation Maya chips: the Maya 300 is claimed at up to seven times more AI tokens per rack, production starts 2027, and Evercore sees homegrown silicon at about 10% of Microsoft's AI chip spending by early next year. New angle: Microsoft's Maya 300 is claimed at up to seven times more AI tokens per rack, with production starting in 2027.
- [The AI Daily Brief](https://boilpoints.com/u/theaidailybrief), 2026-10-13, 8:10 to 9:13: **Adds Nadella's on-stage framing**. Adds the on-stage framing of the Maia 300 announcement: Nadella himself unveiled the chip Monday, pitched it as Microsoft's biggest bet on custom silicon, and Chamath Palihapitiya's reaction that chip companies should expect to move up the stack. New angle: Nadella announced the Maia 300 himself, calling it Microsoft's biggest bet on custom silicon.