Microsoft to deploy more than a million in-house Maia 300 AI chips in 2027
Microsoft unveiled the third generation of its custom AI chip. It says Maia 300 delivers about 40% more per megawatt than leading GPU systems, a claim not independently verified.
Symbolic image · Photo: Hay Kranen / Wikimedia Commons, CC BY 4.0 (cropped and resized)
Microsoft on Monday unveiled the Maia 300, the third generation of its custom artificial intelligence accelerator. The company could deploy more than one million of the chips next year to back its internal AI services, Reuters reported, while Microsoft itself spoke of multi-gigawatt deployments starting in 2027. Its predecessor, the Maia 200, already runs Microsoft's in-house models and OpenAI models served on the Azure cloud platform.
The chip is built from several chiplets: TSMC makes the main compute die on its 2-nanometer process, while other parts use cheaper 3-nanometer technology. According to Notebookcheck, the Maia 300 carries 288 gigabytes of HBM4 memory with 18 terabytes per second of bandwidth. A rack holds 96 chips and draws about 167 kilowatts, compared with 72 GPUs and 142 kilowatts for Nvidia's GB300 racks. Microsoft says it delivers about 40% more useful computing per megawatt than "leading GPU-based systems" but did not name the systems compared.
"It's basically delivering more intelligence for every watt, every dollar, and every byte of memory," Rani Borkar, president of Azure hardware systems and infrastructure at Microsoft, told Reuters. Cloud providers and AI firms such as OpenAI are developing custom chips to reduce their reliance on Nvidia. Microsoft expects capital expenditures of more than $175 billion this calendar year.