This is the reason Microsoft is cutting its Claude token usage. Microsoft just announced it will scale its own AI chip, Maia, to multigigawatt deployments starting in 2027. Maia will become the main system running AI inside Microsoft's products and services. Maia 200 is already live in the United States, Europe and Australia, running Microsoft Copilot today. Microsoft says Maia 200 delivers up to 1.7x more tokens per watt than leading GPU systems in Teams and Copilot. The next chip, Maia 300, is built on $TSM's 2 nanometer process with eight HBM4 memory stacks. Microsoft says it will deliver 40% more tokens per megawatt than leading GPU systems. A full Maia 300 rack holds 96 chips and pushes up to 7x the token output of Maia 200. The Information reported that Microsoft cut its projected internal Claude spend by more than a third, down from over $1 billion a year. Its Cloud and AI group also cut per employee AI budgets from $100,000 a month to about $10,000. Microsoft is now building the chips and the models itself. Every token moved off Claude and onto Maia is money that stays with Microsoft. I think this is a long term margin story for $MSFT that most people are missing.