Scott Guthrie, executive vice president of cloud and enterprise at Microsoft, speaks at the Microsoft Build developer conference in Seattle on May 7, 2018. The Build conference, which marks its second consecutive year in Seattle, is expected to highlight the company’s cloud technologies and the artificial intelligence features within these services.
Grant Hindsley | Bloomberg | Getty Images
Microsoft announced the next generation of its artificial intelligence chip, a potential alternative to leading processors Nvidia and on offers from cloud rivals Amazon And Googling.
The Maia 200 comes two years after Microsoft said it had developed its first AI chip, the Maia 100, which was never made available to cloud customers to rent. Scott Guthrie, Microsoft’s executive vice president for cloud and AI, said in a blog post Monday that the new chip will have “greater availability to customers in the future.”
Guthrie called the Maia 200 “the most efficient inference system Microsoft has ever implemented”. Developers, academics, AI labs, and contributors to open-source AI models can request a preview of a software development kit.
Microsoft said its super intelligence team, led by Mustafa Suleyman, will use the new chip. The Microsoft 365 Copilot add-on for commercial productivity software bundles and the Microsoft Foundry service, for building on top of AI models, will also use it.
Cloud providers are facing rising demand from generative AI model developers such as Anthropic and OpenAI and from companies building AI agents and other products on top of the popular models. Data center operators and infrastructure providers are trying to increase their computing prowess while controlling energy consumption.
Microsoft is equipping its data centers in the US Central region with Maia 200 chips, and then they will arrive in the US West 3 region, with more locations on the horizon.
Using the chips Taiwan Semiconductor Manufacturing Co 3 nanometer process. Within each server, four are connected together. They rely on Ethernet cables instead of the InfiniBand standard. Nvidia is selling InfiniBand switches after acquiring Mellanox in 2020.
The chip offers 30% higher performance than alternatives for the same price, Guthrie wrote. Microsoft said each Maia 200 contains more high-bandwidth memory than a third-generation Trainium AI chip from Amazon Web Services or Google’s seventh-generation tensor processing unit.
Microsoft can achieve high performance by connecting up to 6,144 Maia 200 chips, reducing power consumption and total cost of ownership, Guthrie wrote.
In 2023, Microsoft demonstrated that its coding assistant GitHub Copilot could run on Maia 100 processors.
WATCH: Chinese AI models adapt without Nvidia
