Microsoft AI plans hindered by chip shortage
Archived — this story has rotated out of today’s deck. It is kept here in full.
The gist
Microsoft has 2.2m AI chips installed, less than half of some experts' estimates. The shortfall may limit Copilot and Azure AI services for enterprise customers.
Background
Microsoft has invested heavily in AI infrastructure, spending roughly $280bn since 2022 and claiming to have added 5GW of datacentre capacity. However, a Guardian investigation found that the number of advanced AI chips actually installed is far below what would be expected for that capacity, suggesting some datacentres may not be fully operational. This discrepancy raises questions about Microsoft's ability to meet demand for its AI products.
How it unfolded
- 2024Microsoft reportedly targeted having 1.8m AI chips installed by end of 2024.
- Jan 2026Microsoft unveiled its second-generation Maia 200 AI chip, built by TSMC using 3-nanometer technology.
- Aug 11, 2026Reports emerged that Microsoft plans to unveil its next-generation Maia 300 AI chip in September and is in talks with TSMC for manufacturing capacity.
- Aug 17, 2026Guardian investigation reveals Microsoft has 2.2m AI chips installed, less than half of what some experts expected, amid a $280bn expansion.
Who’s saying what
- Analysts
- The discrepancy suggests Microsoft's newest datacentres may not be fully operational or lack the chips they need.
Still unverified
The Guardian's investigation relies on internal documents and expert estimates; Microsoft has not confirmed the figures. The exact number of AI chips needed for 10GW capacity is an estimate.