HawkInsight

  • Contact Us
  • App
  • English

Self-reliance! Microsoft launches first custom AI chip

On Wednesday (November 15) local time, Microsoft launched two custom-designed internal data center artificial intelligence chips at its Microsoft Ignite Global Technology Conference.

On Wednesday (November 15) local time, Microsoft launched a new self-developed artificial intelligence chip, which the company hopes to use to reduce its dependence on chip manufacturers。

微软芯片

At the Microsoft Ignite Global Technology Conference on Wednesday, Microsoft unveiled two custom-designed in-house data center AI chips: the Maia 100 and the Cobalt 100 CPU.。Maia 100 is a chip optimized for artificial intelligence tasks and generative artificial intelligence, while Cobalt 100 is designed to run general-purpose computing workloads on the Microsoft cloud。

Scott Guthrie, executive vice president of Microsoft's cloud and artificial intelligence group, said: "Microsoft is building infrastructure that supports AI innovation, and we are reimagining every aspect of the data center to meet customer needs.。Guthrie added: "In terms of the scale of our operations, it is important to optimize and integrate every layer of the infrastructure stack to maximize performance, diversify the supply chain and provide customers with infrastructure choices.。"

Microsoft says the Maia 100 and Cobalt 100 will begin rolling out in Azure data centers early next year。Originally providing support for Microsoft AI services such as Copilot (Microsoft's family of generative AI products) and Azure OpenAI Service (the company's fully managed offering for OpenAI)。

Microsoft says it is working with OpenAI to test the Maia 100 and will use that experience to build future chips.。The second generation of Maia and Cobalt chips are already under development。

Sam Altman, CEO of OpenAI, said: "We were very excited when Microsoft first shared the design of its Maia chip and we worked together to improve and test it with our models.。"Azure's end-to-end AI architecture has now been optimized to the chip through Maia, paving the way for training more powerful models and making those models cheaper for our customers."。"

Microsoft said that in addition to the chips, the company also manufactures the server boards and server racks where the chips are located。To reduce the heat generated by AI servers, Microsoft said it created a special cooling feature that allows coolant to pass through a series of pipes and radiators to absorb heat more effectively than fans.。And this should also help reduce electricity consumption, which will be more environmentally friendly。

For Microsoft's self-developed custom artificial intelligence chip, the market is not surprised。

In April, it was reported that Microsoft had been secretly developing AI chips, codenamed Athena, since 2019.。Earlier, in 2020, there were also media revelations that Microsoft had designed a series of ARM-based chips for its data centers and other devices.。

As to whether Microsoft will replace its NVIDIA chips in the cloud with the Maia 100, Microsoft responded that the company works closely with NVIDIA, while mentioning that Microsoft is expanding its industry partnerships, including with NVIDIA.。

Just a few days ago, Nvidia introduced the H200, its latest and most powerful GPU available to power AI training.。Nvidia is the market leader in artificial intelligence chips, and its chips support the research and development of many artificial intelligence companies.。Amazon, Google, Microsoft and many other large technology companies are Nvidia customers。

Microsoft said it is launching a preview of the new NC H100 v5 virtual machine family built for the Nvidia H100 division, and the company will add an upgraded Nvidia H200 to its fleet "next year to support greater model inference."。

·Original

Disclaimer: The views in this article are from the original author and do not represent the views or position of Hawk Insight. The content of the article is for reference, communication and learning only, and does not constitute investment advice. If it involves copyright issues, please contact us for deletion.