Microsoft plans Maia 300 AI chip rollout to cut infrastructure costs

Traders Union
All News
Financial News
Microsoft Maia 300 AI chip rollout

Highlights

Microsoft unveiled the Maia 300, its third-generation custom AI chip, using TSMC's 2-nanometer technology for key functions and planning multi-gigawatt deployments in 2025.
The Maia 300 racks will feature 96 processors drawing 167 kilowatts, exceeding Nvidia GB300 racks' 72 GPUs and 142 kilowatts per rack.
Microsoft claims the Maia 300 offers about 40% higher useful computing per megawatt versus leading GPU-based systems, as it expects over $175 billion in capital expenditures this year.

Chip design and deployment plan

As reported by Reuters, Microsoft on Monday introduced the Maia 300 as the third generation of its custom AI chips, built to handle intensive tasks such as coding and complex business operations carried out by AI systems. The company says the new chip will support internal AI models as well as OpenAI models delivered through its Azure cloud service, extending work already handled by the previous generation.

Microsoft is taking a different engineering route from Nvidia, whose AI accelerators rely on a large central computing chip paired with high-bandwidth memory. In the Maia 300, the main processor uses chiplets, smaller specialized chips bonded together, with key computing functions built on Taiwan Semiconductor Manufacturing Co's 2-nanometer technology, while other functions such as moving data across the system use lower-cost 3-nanometer technology.

Microsoft says it plans to begin multi-gigawatt deployments of the Maia 300 starting next year. The chips will be installed in racks containing 96 processors and drawing about 167 kilowatts of power, compared with 142 kilowatts for Nvidia's GB300 AI server racks with 72 graphics processing units.

Efficiency claims and AI infrastructure spending

At the system level, Microsoft says the Maia 300 delivers about 40% more useful computing per megawatt than leading GPU-based systems, though it does not identify the systems used for comparison. Rani Borkar, president of Azure hardware systems and infrastructure at Microsoft, tells Reuters the design is intended to deliver more intelligence for every watt, dollar and byte of memory by optimizing the full system.

The launch comes as cloud computing rivals and AI developers, including OpenAI, step up efforts to develop custom chips and weaken Nvidia's grip on the AI computing market. For Microsoft, the timing is also tied to the financial pressure of AI expansion, as the company has said it expects to spend more than $175 billion in capital expenditures this calendar year.

Our earlier coverage highlighted how the AI infrastructure buildout is increasingly constrained by financing, with Intel and Broadcom warning that rising capital costs and investor scrutiny may become a bigger bottleneck than chip supply or power availability. The article also noted that sustaining demand through the end of the decade will depend on proving tangible returns from AI spending and improving efficiency as deployments scale.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.