
Microsoft's Maia 300 to Lead AI Inference in 2027
Microsoft announced Maia 300, its third in-house Maia chip, which will handle inference — running trained AI models for users — across Microsoft products and services. Rolling out in 2027, it succeeds Maia 200 and uses TSMC's N2P manufacturing, with a chiplet design that splits compute, memory control and input-output into separate blocks.
Published