Microsoft Unveils Maia 300 AI Chip, Aiming to Reduce Reliance on Nvidia

Microsoft unveiled Maia 300, its latest custom AI accelerator, and said deployments will begin in 2027. Microsoft said Maia 300 will become the primary inference system for Microsoft's own products and services. Microsoft says Maia 300 will deliver 40% more AI tokens per megawatt than leading GPU-based systems. Maia 300 is designed to minimize unnecessary data movement between chips and memory, a growing bottleneck as AI models become more complex.

Microsoft's plan to power its AI products with custom silicon shows how every major Nvidia customer is now a competitor. Those customers seek lower costs, improved performance and greater control. The shift is particularly significant for inference, the computing needed to generate AI responses, as more sophisticated models and agents increase computing demand. Nearly all major AI players are seeking alternatives to Nvidia.

Google has been developing Tensor Processing Units for years. Google's latest generation includes separate versions optimized for training AI models and running inference. OpenAI is developing its Jalapeño chip with Broadcom, initially focusing on inference workloads. OpenAI said in August that OpenAI's first chip was outperforming Nvidia hardware on certain workloads. OpenAI said OpenAI was already developing subsequent generations.

Amazon has its own Trainium family of AI chips, which customers including Anthropic are using. OpenAI has committed to using Amazon's custom silicon as part of a massive infrastructure agreement. Anthropic confirmed in August that Anthropic is exploring its own chip designs while continuing to rely on Nvidia, Amazon and other suppliers. Meta also has an internal chip development program while continuing to buy substantial computing capacity from Nvidia and other suppliers.

Nvidia is not necessarily losing business because Nvidia's customers are developing competing chips. Nvidia's revenue more than doubled to $96 billion in the second quarter. Overall demand for AI computing leaves room for Nvidia chips and new offerings from Microsoft and others. Most Nvidia competitors, including Microsoft, focus on inference, where margins are highest. Nvidia remains dominant in training. Nvidia has begun using its balance sheet to finance the AI infrastructure expansion. Competition that even slightly reduces Nvidia's strength could weaken Nvidia's central role in continued growth.

Rani Borkar, Microsoft's president of Azure Hardware Systems and Innovation, said competitive AI hardware requires designing entire systems, optimizing memory, networking, software and power consumption. Borkar said Microsoft prioritizes Maia for inference. Borkar described inference as the largest and fastest-growing AI computing opportunity, where Microsoft can distinguish itself through cost and efficiency. Borkar said Microsoft wants a broad ecosystem while investing deeply in Microsoft's own silicon, rather than choosing between those approaches.

Microsoft's plans to make Maia its primary inference platform apply specifically to Microsoft's own services, not necessarily to third-party Azure customers, including OpenAI. Borkar cautioned that the transition will not be immediate. Borkar said Maia 300 deployments will begin in 2027, but reaching multi-gigawatt scale will take time, extending beyond next year.

Companies developing custom chips remain dependent on the same limited manufacturing capacity, advanced packaging and high-bandwidth memory that Nvidia relies on. Custom chips can reduce dependence on Nvidia but do not eliminate broader supply-chain constraints. Nvidia remains central to the AI boom. However, Nvidia's biggest customers increasingly want greater control over the technology powering those customers' services and a greater share of the economic benefits.