Microsoft's Most Powerful AI Chip, Maia 300, Unveiled

With the introduction of its powerful Maia 300 AI chip, Microsoft has taken a major step toward reducing its dependence on Nvidia chips.

With the introduction of the powerful artificial intelligence chip Maia 300, Microsoft has taken a big step towards reducing dependence on Nvidia chips.

Become a special member of Arjan Ambassador in order to benefit from and access special facilities and different departments in Arjan Ambassador.

Microsoft unveiled the third generation of its proprietary artificial intelligence chip named "Maia 300". This chip is designed with the aim of becoming the main infrastructure of artificial intelligence inference in Microsoft products and services, and in addition to reducing the cost and increasing the productivity of artificial intelligence processes in the cloud, it has a greater ability to implement new reasoning models and tasks based on artificial intelligence agents.

The supply of Maia 300 in operational environments is scheduled to begin in 2027. Focusing on energy efficiency, Microsoft expects the chip to process 40 percent more tokens per megawatt of power compared to leading GPU-based systems. Also, this chip in a full configuration with 96 accelerators can increase the throughput of rack-level token processing up to 7 times more than Maia 200.

Scott Guthrie, executive vice president of Microsoft's cloud computing division, says these improvements will soon allow the company to rely more on its own chips and reduce its dependence on Nvidia and AMD products. According to him, from 2027, Microsoft will start developing the Azure Maia chip family on a scale of several gigawatts, so that these chips will become the main infrastructure for inferring artificial intelligence in the company's products and services.

Maia 300 is the successor of Maia 200; A chip that was introduced only in January. Maia 200 is already used to process artificial intelligence workloads in various parts of Microsoft, including Microsoft Copilot, GitHub Copilot, Microsoft Foundry, and the company's proprietary AI models.

Guthrie says the Maia 200 is the best processing infrastructure in Microsoft's fleet in terms of real-world AI service performance metrics. However, Maia 300 is designed as a complete system for the new generation of artificial intelligence processing; From long-context reasoning and latency-sensitive conversational experiences to continuous execution of multi-step tasks by AI agents.

Microsoft has used the new chiplet architecture to achieve this increase in productivity. Maia 300 is built based on TSMC's N2P manufacturing process, and in its architecture, the processing, memory controller, and input and output parts are separated from each other. The Azure hardware and infrastructure team says this design frees up more chip space for processing, allowing the technology to be tailored to each department's task.

The chip also plays an important role in the development of Microsoft's own AI models, and could help the company catch up with rivals such as Google DeepMind, OpenAI, and Entropic. Mustafa Suleiman, CEO of Microsoft's artificial intelligence division, emphasizes that the company must develop all layers of technology, from models to the silicon level. According to him, at such a scale, Microsoft needs a dedicated platform that is designed along with the company's artificial intelligence models; For this reason, the implementation of Microsoft's exclusive models on the Maia chip is one of the main priorities of this team.

Technology, especially the part that moves on the edge, is considered the most attractive part of this world to me, but all areas, even the intersection of science and technology, can be very attractive. Besides these, I also like the world of movies, series and video games.