Microsoft Says Maia 300 Chip Will Run Its AI Services from 2027
Microsoft will make its own Maia chips the primary system for running AI in its products starting in 2027. The company set out the plan on Monday as it unveiled the Maia 300, the third generation of its custom AI chip. Microsoft plans deployments measured in gigawatts starting next year, which could mean more than a million Maia 300 chips in 2027.
Key Features of Maia 300:
- Inference-focused: Designed for running AI models once trained, catering to long conversations, reasoning, and multi-step tasks.
- Improved Performance: For each megawatt of power, Maia 300 generates 40% more tokens than top GPU systems.
- Efficient Design: Uses smaller chiplets for computing, memory, and data transfer, optimized across the entire system.
Comparisons and Impact:
- Power Efficiency: A full rack of Maia 300 chips delivers up to seven times the output of its predecessor, Maia 200.
- Competitive Landscape: Microsoft joins Google and Amazon in developing AI chips, with OpenAI having built its own chip with Broadcom.
- Factory Capacity: Microsoft has secured factory capacity with TSMC for manufacturing Maia 300 chips.
Maia 200 is already in use, serving Microsoft Copilot and running OpenAI models alongside Microsoft’s MAI models. The launch of Maia 300 is a significant step in Microsoft’s AI strategy, aiming to revolutionize its products with more powerful and efficient AI capabilities.
Microsoft plans to share more details about Maia 300 at its Ignite conference in San Francisco next month.