Rumor
Microsoft to build gigawatt-scale infrastructure for internal frontier model development
Friday, February 13, 2026 at 06:51 AM
Microsoft is planning to develop frontier foundation models internally to reduce its reliance on OpenAI. This strategy involves building gigawatt-scale compute infrastructure and leveraging specialized AI training teams for internal development.
Context
Microsoft is pivoting toward AI self-sufficiency with plans to launch proprietary, "frontier-grade" models in 2026. This strategic shift, confirmed by AI chief Mustafa Suleyman, aims to reduce the company's long-standing reliance on OpenAI by leveraging in-house foundation models and gigawatt-scale compute. While Microsoft maintains a 27% stake in OpenAI following its $13 billion investment, it is increasingly diversifying its AI portfolio, including a $5 billion stake in primary competitor Anthropic.
This independence move coincides with a cooling relationship between OpenAI and its primary hardware provider, Nvidia. CEO Jensen Huang recently clarified that a rumored $100 billion commitment to OpenAI was never a binding deal, with negotiations for massive data center clusters reportedly stalled. These developments signal a broader fragmentation in the AI supply chain as hyperscalers move to internalize development costs during an expected $650 billion industry spending blitz throughout 2026.
Sources (3)
Related Companies
Microsoft
MSFT
Nvidia
NVDA