Microsoft unveils Maia 300. Production deployment is planned for 2027.

The announcement separates Maia 200 already serving workloads from its successor’s deployment plan. It does not announce a cheaper chatbot subscription.

Microsoft’s October 12, 2026 announcement says Maia 200 is serving production workloads and Maia 300 will begin production deployments in 2027. The company plans to expand its Maia infrastructure at multi-gigawatt scale starting that year.

What is the new system for?

Microsoft describes Maia 300 as focused on inference: running models for workloads such as long-context reasoning and multi-step agent tasks. Its announcement discusses compute, memory, networking and power together, rather than a chip alone.

Read efficiency claims with their conditions

The company forecasts better token throughput per unit of power. Those are vendor claims for specified systems and workloads, not our independent measurements. We have not tested the hardware or compared it with another system.

What changes for a reader today?

Our interpretation: a future infrastructure deployment is not itself a promise about your subscription price, answer quality or response time. Check a service’s own release and pricing notices before assuming that a hardware announcement changes your account.

Published October 14, 2026. The official blog is dated October 12; Source Asia lists it on October 13. Planned Maia 300 production deployment is in 2027, not already completed.

Sources: Microsoft · Maia announcement.