3 months ago
2
Microsoft connected Monday unveiled the 2nd procreation of its in-house artificial quality chip, along with bundle tools that instrumentality purpose astatine 1 of Nvidia’s biggest competitory advantages with developers. The caller “Maia 200” spot comes online this week successful a information centre successful Iowa, with plans for a 2nd determination successful Arizona, Microsoft said. It is the 2nd procreation of an AI spot called Maia that Microsoft introduced successful 2023.
The Maia 200 comes arsenic large unreality computing firms specified arsenic Microsoft, Alphabet’s Google and Amazon.com’s Amazon Web Services - immoderate of Nvidia’s biggest customers - are producing their ain chips that progressively vie with Nvidia. Google, successful particular, has garnered involvement from large Nvidia customers specified arsenic Meta Platforms, which is moving intimately with Google to adjacent 1 of the biggest bundle gaps betwixt Google and Nvidia’s AI spot offerings.
For its part, Microsoft said that on with the caller Maia chip, it volition beryllium offering a package of bundle tools to programme it. That includes Triton, an open-source bundle instrumentality with large contributions from ChatGPT creator OpenAI that takes connected the aforesaid tasks arsenic Cuda, the Nvidia bundle that galore Wall Street analysts accidental is Nvidia’s biggest competitory advantage.
Like Nvidia’s forthcoming flagship “Vera Rubin” chips introduced earlier this month, Microsoft’s Maia 200 is made by Taiwan Semiconductor Manufacturing Co utilizing 3-nanometer chipmaking exertion and volition usage high-bandwidth representation chips, albeit an older and slower procreation than Nvidia’s forthcoming chips. But Microsoft has besides taken a leafage from the playbook of immoderate of Nvidia’s rising competitors by packing the Maia 200 spot with a important magnitude of what is known arsenic SRAM, a benignant of representation that tin supply velocity advantages for chatbots and different AI systems erstwhile they tract requests from a ample fig of users.
Cerebras Systems, which precocious inked a $10 cardinal woody with OpenAI to proviso computing power, leans heavy connected that technology, arsenic does Groq, the startup that Nvidia licensed exertion from in a reported $20 cardinal deal.
