Microsoft to boost Maia 300 inference chip output ahead of fall launch

August 13, 2026 · mediapost.com

Maia 300, Microsoft's next-generation AI chip focused on inference, is set for a fall release, and the company plans to increase production ahead of launch, The Information reported, citing an unnamed source. The chip follows Maia 200, announced in January as part of Microsoft's "heterogenous AI infrastructure" for running AI models, including the latest models from OpenAI. Microsoft has positioned Maia as part of a strategy for continuous computational processing that supports each AI-driven function in Copilot and search. Meta is pursuing a similar path with a new chip called Iris. One report says Iris is optimized to run AI tasks across Facebook, Instagram and WhatsApp, handling content ranking, ads and generative AI to free up third-party GPUs. Meta hasn't confirmed that Iris belongs to its Meta Training and Inference Accelerator (MTIA) chip family. The same report says Iris follows that strategy.

Lower ad-serving costs and lower server costs could follow as inference-specific chips replace expensive, general-purpose hardware for real-time ad optimization and creative tools, per the report. Meta's ambitions reach further into how compute itself gets priced. Mark Zuckerberg's manifesto for AI superintelligence, published Monday, promises free basic AI tools for billions of people and a "dynamic auction mechanism" for heavy compute users, working like a continuous stock exchange that matches available computing power with demand and prices it in real time. Zuckerberg says the model will guarantee the lowest price for the intelligence and compute people use while keeping capacity allocated to what is collectively most valuable.