Microsoft Maia 300 Targets Azure AI Rollout in 2027

Microsoft plans to begin large-scale Azure data-centre deployment of its Maia 300 custom AI inference chip in 2027. Built on TSMC’s N2P process with a chiplet design, Maia 300 is intended for workloads such as conversational AI and long-context reasoning. Microsoft says it can process 40% more tokens per megawatt than leading GPU systems and deliver up to seven times the throughput of its Maia 200 predecessor; these performance claims have not been independently verified. The chip is expected to support Copilot, GitHub Copilot and models developed by Microsoft and OpenAI. Maia could account for about 10% of Microsoft’s AI chip spending by early 2027, while Nvidia and AMD are still expected to supply hardware, particularly for AI training. Earlier reports said Microsoft was negotiating with TSMC for more than 300,000 chips in 2027, with potential longer-term volumes above one million, and had discussed Maia with external customers including Anthropic. Those volume estimates and customer discussions remain unconfirmed, and no external deals have been announced. For traders, Maia 300 highlights continued investment in custom AI chips and could eventually reduce Microsoft’s reliance on Nvidia and lower infrastructure costs. The commercial effect will depend on manufacturing capacity, deployment pace and verified efficiency gains. The announcement has no direct catalyst for a specific cryptocurrency, though it adds to broader market attention on AI infrastructure and technology spending.