Microsoft Launches Maia 200: Custom AI Chip Takes On NVIDIA

Microsoft has officially launched Maia 200, its second-generation custom AI chip, marking a significant step toward reducing dependency on NVIDIA GPUs.

Maia 200 Specifications

• Performance: 10+ petaFLOPS at FP4
• Memory: 216GB HBM3e
• Process: 3nm manufacturing
• TCO Improvement: >30% reduction vs. current solutions

Multi-Vendor Strategy

CEO Satya Nadella emphasized Microsoft’s multi-vendor fleet approach: NVIDIA + AMD + Maia.

“We don’t want to be locked into any one thing… It’s not a one-generation game. You have to be ahead for all time to come.”

— Satya Nadella

Infrastructure Investment

Microsoft is investing $120B+ in FY26 CapEx, with Maia 200 central to its “tokens per watt per dollar” optimization strategy.

The chip will power Azure AI workloads across Microsoft’s 400+ datacenters in 70 regions worldwide.

Competitive Implications

While Maia 200 represents significant progress, Microsoft acknowledges its custom silicon remains 3-5 years behind Google’s TPUs. However, the vertical integr — as explored in how AI is restructuring the traditional value chain — ation allows Microsoft to optimize specifically for its enterprise AI workloads.

For a deeper strategic analysis, read Microsoft In The AI Stack on The Business Engineer.