This new model was presented two years after its first version, the Maia 100, which was never made available to customers.
Microsoft presented the Maia 200, the second generation of its artificial intelligence chip () with which it seeks to reduce dependence on Nvidia and compete with Google and Amazon in the cloud.
This new model was presented two years after its first version, the Maia 100, which was never made available to customers.
“It’s the most efficient inference system Microsoft has ever implemented,” he said. Microsoft’s executive vice president for cloud and AI, in a publication cited by the EFE agency.
According to the company, the chip was designed as an inference processor and promises up to 30% more performance per dollar than Microsoft’s current hardware.
Additionally, it is already being deployed in Azure data centers across the United States to power services like Microsoft 365 Copilot, Foundry, and OpenAI’s latest GPT models.
Large cloud providers are struggling to develop their own chip that does not depend exclusively on Nvidia, the market leader.
Launches from Amazon, Google and now Microsoft increase competition in an expanding market.
