Microsoft Pushes Maia 300 AI Chip to Market This Fall

Microsoft's move to scale its custom silicon marks a turning point after years of reliance on Nvidia. The company introduced its first Maia chip in November 2023, but meaningful progress has been slow compared to rivals. The second-generation Maia 200, which began rolling out earlier this year on TSMC's 3-nanometer process, represents incremental maturity. The new Maia 300 arrives at a moment when the industry is grappling with a severe memory shortage that has inflated GPU costs—a pressure that makes owning proprietary silicon more strategically valuable.

Yet Microsoft faces a credibility challenge. Booking capacity for over 300,000 chips, with a stated ambition for over a million, signals real commitment rather than a science project, as Azure Maia general manager Andrew Wall suggested the reported figures "don't reflect the scale of our program." However, the company declined to share specific production numbers, leaving outsiders uncertain about the reality behind the negotiations. Capacity negotiations with TSMC are still ongoing and could constrain scaling speed. More immediately, major customers Microsoft is courting—like Anthropic—already hold commitments to Amazon and Google's AI chips, which limits the addressable market for Maia as a product line beyond Microsoft's own cloud infrastructure.