The brief

Analyst view

• Microsoft launched seven frontier AI models June 2026, trained in-house from scratch with zero competitor distillation.
• GitHub Copilot reached 50M users by July 2026; 1 in 3 PRs now involve an AI agent.
• Maia 200 (3nm, 216GB HBM3e) ships to Azure US Central, claims 3x FP4 vs Trainium3.

Technical approach

As reported

Frontier Models

In-house AI model family

Seven models trained from scratch: MAI-Thinking-1 (reasoning), MAI-Code-1-Flash (coding), MAI-Image-2.5, MAI-Voice-2, transcription. Zero distillation.

Custom Silicon

Maia inference accelerator

TSMC 3nm chip with 216GB HBM3e, FP4/FP8 cores, 750W TDP. 10+ petaFLOPS FP4, targets cost-per-token economics.

Superfactory

Distributed training infrastructure

Wisconsin and Atlanta sites linked by 120,000+ miles AI WAN. Datacenters coordinate training as one machine.

Coding Agents

Copilot autonomous work

Agents draft PRs autonomously. 1 in 3 GitHub PRs involved agents by July 2026. GitHub moved to metered credits.

Show all 4