Microsoft Reveals Maia 300 AI Chip As It Works To Reduce Dependence On Nvidia
(Tuesday October 13, 2026 @11:00AM (BeauHD)
from the dependency-reduction dept.)
- Reference: 0186291476
- News link: https://hardware.slashdot.org/story/26/10/13/0443232/microsoft-reveals-maia-300-ai-chip-as-it-works-to-reduce-dependence-on-nvidia
- Source link:
An anonymous reader quotes a report from Yahoo Finance:
> Microsoft has [1]debuted its latest AI chip, the Azure Maia 300, as it looks to reduce its reliance on third-party processors from the likes of Nvidia and AMD. Unveiled at the OCP Global Summit 2026 in San Jose on Monday, the Maia 300 [2]will join Microsoft's Maia 200 and Maia 100 chips as part of what the company said will be multigigawatt deployments beginning next year. The Windows maker said that will make its Maia chips the "primary AI inference system for Microsoft products and services."
>
> The company said Maia 300 provides 40% more tokens per megawatt than "leading GPU-based systems," a clear swipe at Nvidia, and that a 96-chip rack offers seven times the rack-scale token throughput of Microsoft's Maia 200. [...] The Maia 200 chip currently powers its Microsoft production workloads, including Microsoft 365 Copilot, as well as OpenAI models, Microsoft's MAI models, and open-source models. That provides 1.7 times more tokens per watt for AI capabilities in Microsoft's Teams, Copilot, and Cowork software versus other GPU-based systems, the company said.
[1] https://azure.microsoft.com/en-us/blog/deep-dive-into-azure-maia-300-architecture/
[2] https://finance.yahoo.com/technology/article/microsoft-reveals-maia-300-ai-chip-as-it-works-to-reduce-dependence-on-nvidia-amd-130000212.html
> Microsoft has [1]debuted its latest AI chip, the Azure Maia 300, as it looks to reduce its reliance on third-party processors from the likes of Nvidia and AMD. Unveiled at the OCP Global Summit 2026 in San Jose on Monday, the Maia 300 [2]will join Microsoft's Maia 200 and Maia 100 chips as part of what the company said will be multigigawatt deployments beginning next year. The Windows maker said that will make its Maia chips the "primary AI inference system for Microsoft products and services."
>
> The company said Maia 300 provides 40% more tokens per megawatt than "leading GPU-based systems," a clear swipe at Nvidia, and that a 96-chip rack offers seven times the rack-scale token throughput of Microsoft's Maia 200. [...] The Maia 200 chip currently powers its Microsoft production workloads, including Microsoft 365 Copilot, as well as OpenAI models, Microsoft's MAI models, and open-source models. That provides 1.7 times more tokens per watt for AI capabilities in Microsoft's Teams, Copilot, and Cowork software versus other GPU-based systems, the company said.
[1] https://azure.microsoft.com/en-us/blog/deep-dive-into-azure-maia-300-architecture/
[2] https://finance.yahoo.com/technology/article/microsoft-reveals-maia-300-ai-chip-as-it-works-to-reduce-dependence-on-nvidia-amd-130000212.html
Not really (Score:2)
by CEC-P ( 10248912 )
To anyone wondering whose ass MS pulled an entire chip fabrication plant out of, it's exactly what you suspected. They're making anything; TSMC is. And they're out of fab space constantly.
Some issues (Score:2)
by TheMiddleRoad ( 1153113 )
1. Between hardware and software development and then deployment of non-standard hardware, this has got to be costing way more than just buying what NVIDIA already makes.
2. MXFP4 is less efficient and less accurate than NVFP4, so they're not exactly making the best decisions.
3. Copilot AI generally sucks.
Good luck, but I don't think it will work (Score:2)
by OrangeTide ( 124937 )
If you want to ditch Nvidia, you should probably use AI to make a new CUDA competitor instead of wasting time designing silicon.
Re: (Score:2)
Can you run Doom on one of them? If not, it's just a slop generator (which, I think, there are enough o f already).
Re: (Score:2)
It's marketing as they try to stay relevant in the new AI era. Standard playbook but this time they have no control of the players because their OS is not relevant.
When they'd just announced Maia 200 earlier this year and now the 300 which the claim is that it's faster than the Maia 200 it's fluff and puffing hopping to be noticed and seen as 'trying".
After all, they're claiming it'll be used in their own data centers and nothing more. And I'm sure Microsoft doesn't like being just another line item on NVi