Friday, 12 September, 2025
Microsoft Boosts In-House AI Model Development with Major Compute Investment
Microsoft is stepping up its ambitions to build frontier AI models in-house, unveiling plans for a much larger compute infrastructure. The company’s AI chief, Mustafa Suleyman, revealed that Microsoft’s MAI-1 preview was trained on about 15,000 Nvidia H100 GPUs, but future clusters will be six-to-ten times bigger. CEO Satya Nadella emphasized support for multiple models across products like GitHub Copilot.
Read full story at The Verge