Microsoft Announces Maia 300 AI Chip, With Deployments Planned for 2027

Maia 300 sets the next stage for Microsoft's AI services

Microsoft announced Maia 300, its third-generation custom artificial intelligence chip, on October 12, 2026. The company says large-scale deployments will begin in 2027, making this an announcement of planned infrastructure rather than immediate availability.

Microsoft intends Maia to become its primary system for AI inference—the work of running a model to produce responses. That goal puts the chips at the center of how the company plans to deliver AI through its products and services.

Microsoft says Maia 300 is designed for AI agents that carry out tasks in multiple steps, along with models handling long documents or conversations. Those are the types of work the new chip is meant to support.

Microsoft's output claims focus on power and scale

Microsoft says Maia 300 will deliver 40% more AI output at the same power consumption as leading systems based on graphics processing units, or GPUs. The claimed benefit is more AI work from the same amount of electricity.

The company also says a full rack of Maia 300 chips will deliver up to seven times the AI output of Maia 200. That comparison covers a rack of chips; it does not establish a sevenfold gain for each chip or each user request.

These are Microsoft's performance claims. The supplied reporting does not identify the GPU systems or testing conditions behind the efficiency comparison, limiting how precisely readers can assess it.

Maia 200 is already doing the work

The shift is already underway with Maia 200. That generation powers Microsoft 365 Copilot and runs both OpenAI models and Microsoft's own models.

Microsoft says Maia 200 is being deployed across the United States, Europe and Australia. GitHub Copilot and Microsoft Foundry are scheduled to move onto Maia 200 by the end of 2026, ahead of the planned Maia 300 deployments.

For customers, access currently comes through supported Microsoft services: Maia is not available for outside customers to rent, according to GeekWire. The announcement therefore does not offer businesses a Maia chip service they can rent directly.

A staged rollout with suppliers still involved

Microsoft designs Maia chips, while Taiwan Semiconductor Manufacturing Co. manufactures them. Nvidia also remains a key supplier for Microsoft's data centers, alongside the company's plans for its own chips.

The next milestones are the scheduled Maia 200 move for GitHub Copilot and Microsoft Foundry by the end of 2026, followed by large-scale Maia 300 deployments in 2027. Those plans describe a gradual shift; they do not establish when Maia will become Microsoft's primary inference system.