Microsoft to Unveil Maia 300 AI Chip Amid Push for 1M Units by 2027

Microsoft is preparing to unveil its next-generation AI-specific processor, the Maia 300, in an effort to strengthen its in-house chip capabilities and reduce reliance on Nvidia’s costly graphics processing units.

Timing and production targets

According to reports citing individuals familiar with the company’s plans, Microsoft intends to reveal the Maia 300 chip this autumn, possibly as soon as next September. The company is currently in discussions with Taiwan Semiconductor Manufacturing Company (TSMC) to secure sufficient manufacturing capacity to produce more than 300,000 units for delivery in 2027. Microsoft’s ultimate goal is to reach over one million Maia 300 chips, though that figure remains contingent on component supply availability and ongoing negotiations with TSMC over production capacity.

Strategic shift toward internal silicon

This move follows Microsoft’s November 2023 debut of the first Maia-series chip and reflects its broader effort to build an internal chip ecosystem, granting greater control over the infrastructure powering its AI services. However, Microsoft has progressed at a slower pace compared to rivals Google and Amazon, both of which have expanded internal use of custom-developed processors for AI workloads.

Google began recording revenue from direct sales of its Tensor Processing Units (TPUs) during the fiscal quarter ended last June. Amazon, meanwhile, has seen increased adoption of its internally developed Trainium chips.

Broad industry trend toward self-reliance

These developments signal a wider industry shift: rather than depending entirely on Nvidia for AI computing power, major technology firms are pursuing proprietary chip development to potentially lower costs and gain tighter control over performance and supply chains.

Commercial expansion beyond internal use

Microsoft is also seeking to deploy the Maia 300 beyond its own data centers. It is engaging major cloud computing customers—including Anthropic, developer of the Claude AI model—to adopt the new chip, positioning Maia 300 for roles extending well beyond internal deployment.

Evolution from Maia 200

The Maia 300 follows the Maia 200, launched in January 2026—less than a year prior. The newer chip incorporates a substantial amount of SRAM memory, a high-speed memory type enabling rapid data access, which may deliver significant advantages for AI systems handling large volumes of concurrent user requests.