Microsoft unveils Maia 300 AI chip to reduce reliance on Nvidia

Microsoft has introduced the Maia 300, its third-generation in-house AI chip, and it has a very specific job in mind. The company says the chip is designed to serve as the “primary AI inference system” behind Microsoft’s own products and services. What Microsoft is promising The headline metric is energy efficiency. Scott Guthrie, executive vice president of Microsoft’s cloud group, put a number on it. “We expect Maia 300 to deliver 40 percent more tokens per megawatt compared to leading GPU-based systems.” For the uninitiated, tokens are the small chunks of text an AI model reads and writes. More tokens per megawatt means more AI output for the same electricity bill, which matters when power has become one of the tightest constraints on data center growth. Note the framing, though. Guthrie said Microsoft expects that result. It is a target, not a benchmark anyone has independently verified yet. The chip will start rolling out to production in 2027. So customers and investors are evaluating a promise with a lead time measured in quarters, not weeks. From tens of thousands to hundreds of thousands The real story may be scale. Research findings indicate Microsoft is negotiating with ...