Microsoft has unveiled its third generation custom processor and says it could install more than a million units next year to run its own AI services.
The Microsoft Maia 300 AI chip, the third generation of the company’s custom processors, was unveiled on Monday, and Microsoft plans to deploy potentially more than a million of them next year to back its internal AI services, according to Reuters. Investing.com reported that the rollout is meant to reduce reliance on outside hardware vendors and lower operating costs.
Why Is Microsoft Building Its Own AI Chips?
According to Reuters, the Maia 300 is a custom AI chip made for AI data crunching such as coding and carrying out complex business tasks on a user’s behalf. Its predecessor already runs Microsoft’s own AI models and OpenAI models delivered on the Azure cloud service, Reuters said. Rani Borkar, president of Azure hardware systems and infrastructure at Microsoft, told Reuters the chip delivers “more intelligence for every watt, every dollar, and every byte of memory.”
Reuters said the launch comes as cloud rivals and AI firms such as OpenAI build their own chips to challenge Nvidia’s grip on AI computing. Investing.com reported that Google, Amazon and Meta Platforms have also sped up their chip programs. The Next Web reported that Maia will become the main system for running AI in Copilot and other Microsoft products from 2027.
Inside the Microsoft Maia 300 AI Chip
Reuters reported that Microsoft used a different design from Nvidia, whose accelerators pair one large compute chip with surrounding memory chips. The Maia 300 is built from chiplets, which are specialized chips bonded together. Key computing functions use 2 nanometer technology from Taiwan Semiconductor Manufacturing Co (TSMC), while chiplets that move data around the system use cheaper 3 nanometer technology, according to Reuters.
The processors go into racks of 96 chips that draw about 167 kilowatts, compared with 142 kilowatts for Nvidia’s GB300 racks holding 72 GPUs, Reuters said. Microsoft claims about 40% more useful computing per megawatt than leading GPU systems, though it did not name the systems it used for comparison, Reuters reported. Yahoo Finance said a full rack delivers seven times the token throughput of a Maia 200 rack.
How Large Is the AI Inference Rollout?
Reuters said Microsoft plans deployments measured in gigawatts starting next year, and has said it expects capital spending above $175 billion this calendar year. Investing.com said the plan for Microsoft AI chips covers its global data center footprint.
TokenPost reported that Scott Guthrie, Microsoft’s executive vice president for Cloud and AI, said Azure Maia will scale to gigawatt deployments from 2027 as the primary system for AI inference across Microsoft products and services. The Next Web said The Information reported in August that Microsoft was in talks with TSMC to make more than 300,000 Maia 300 chips for 2027. The Next Web described a total above one million as a possibility, not a confirmed order.
From Maia 100 to Maia 300
Redmond Magazine reported that Maia 100 arrived in 2023 and that Maia 200, an AI inference chip built on TSMC’s 3 nanometer process, was introduced in January 2026. Guthrie wrote that Maia 200 delivers 30% better performance per dollar than the latest hardware in Microsoft’s fleet, the magazine said. CRN reported that Microsoft first deployed Maia 200 systems in its US Central region near Des Moines, Iowa.
Yahoo Finance said Maia 200 now powers Microsoft 365 Copilot, OpenAI models and Microsoft’s own MAI models. TokenPost reported that Maia 300 raises compute performance by 36% over Maia 200.
Will Microsoft Stop Buying Nvidia Chips?
Not entirely, according to Yahoo Finance. It reported that customers still want several kinds of chips for different workloads, so Microsoft and other cloud giants will keep buying Nvidia processors, which remain the most in demand. Yahoo Finance added that Nvidia regularly earns upwards of 50% of quarterly revenue from hyperscalers, and that Amazon and Google now rent their custom chips to outside customers. Its market data showed Microsoft shares up 1.08% and Nvidia down 1.44% in late morning trading on Monday.
To view more coverage on technology visit Tamrah News Technology section.
FAQs: Microsoft Maia 300 AI Chip
What is the Microsoft Maia 300 AI chip?
It is Microsoft’s third generation custom AI chip, designed for AI data crunching such as coding and complex business tasks, according to Reuters. Microsoft says Maia will be its primary AI inference system.
When will Microsoft deploy its custom AI chip at scale?
Microsoft said Azure Maia deployments measured in gigawatts begin next year, which is 2027, according to Reuters and TokenPost.
How many Microsoft AI chips could be installed in 2027?
Reuters reported potentially more than a million Maia 300 chips next year. The Next Web said The Information reported talks with TSMC for more than 300,000 chips.
How does the Maia 300 compare with Nvidia GPUs?
Microsoft says it gets about 40% more useful computing per megawatt than leading GPU systems, which it did not name, Reuters reported. Maia 300 racks draw about 167 kilowatts, against 142 kilowatts for Nvidia GB300 racks.
Will Microsoft still buy Nvidia chips as Azure Maia grows?
Microsoft and other cloud providers will keep stocking Nvidia processors for different workloads.
Which Microsoft products will use the Maia 300 AI chip?
Maia will be the main system for AI in Copilot and other Microsoft products from 2027. Yahoo Finance said Maia 200 already powers Microsoft 365 Copilot.
How much will the Microsoft Maia 300 AI chip save in costs?
No dollar savings figure appeared in the coverage Tamrah News reviewed. Microsoft pointed to efficiency, citing 40% more useful computing per megawatt.
Who makes the Maia 300 chip?
TSMC supplies the 2 nanometer and 3 nanometer technology for its chiplets, according to Reuters.