Microsoft goes all in on AI. Final yr, the corporate unveiled a customized AI chip to spice up the efficiency of its infrastructure to coach massive language fashions (LLMs) cost-effectively. In what could be known as one other step in the direction of bringing prices down and avoiding expensive reliance on third-party chipmakers, Microsoft is alleged to be engaged on AI server tech to run its programmes.
In response to a report by The Data, Microsoft is growing server community playing cards to exchange those supplied by Nvidia, which can assist the corporate avoid wasting money. Server playing cards transfer knowledge shortly between servers, the report mentioned, highlighting that these might additionally improve the efficiency of its Nvidia chip servers in addition to of its Maia AI chips.
“The software program large is growing a networking card to make sure that knowledge strikes shortly between its servers, as an alternative choice to one provided by Nvidia,” the report claimed, citing an individual with direct data.
“Along with doubtlessly saving Microsoft cash, the corporate hopes the brand new networking gear will enhance the efficiency of its Nvidia chip servers,” it added.
The transfer comes as Nvidia is using the elevated demand wave for its AI chips. Not solely Microsoft, however a number of firms have been reported to be engaged on in-house chips and different infrastructure to spice up the efficiency of their techniques and reduce their dependence on Nvidia.
Microsoft’s AI infrastructure plans
At Microsoft Ignite 2023, the corporate unveiled the Microsoft Azure Maia 100 AI Accelerator, which is optimised for endeavor AI duties.
At the moment, Scott Guthrie, govt vice chairman of Microsoft’s Cloud + AI Group mentioned that the corporate is constructing the infrastructure to help AI innovation, and Microsoft is “reimagining each facet of our knowledge centres to fulfill the wants of our prospects.” Microsoft introduced that the chips will begin to roll out this yr.
In response to a report by The Data, Microsoft is growing server community playing cards to exchange those supplied by Nvidia, which can assist the corporate avoid wasting money. Server playing cards transfer knowledge shortly between servers, the report mentioned, highlighting that these might additionally improve the efficiency of its Nvidia chip servers in addition to of its Maia AI chips.
“The software program large is growing a networking card to make sure that knowledge strikes shortly between its servers, as an alternative choice to one provided by Nvidia,” the report claimed, citing an individual with direct data.
“Along with doubtlessly saving Microsoft cash, the corporate hopes the brand new networking gear will enhance the efficiency of its Nvidia chip servers,” it added.
The transfer comes as Nvidia is using the elevated demand wave for its AI chips. Not solely Microsoft, however a number of firms have been reported to be engaged on in-house chips and different infrastructure to spice up the efficiency of their techniques and reduce their dependence on Nvidia.
Microsoft’s AI infrastructure plans
At Microsoft Ignite 2023, the corporate unveiled the Microsoft Azure Maia 100 AI Accelerator, which is optimised for endeavor AI duties.
At the moment, Scott Guthrie, govt vice chairman of Microsoft’s Cloud + AI Group mentioned that the corporate is constructing the infrastructure to help AI innovation, and Microsoft is “reimagining each facet of our knowledge centres to fulfill the wants of our prospects.” Microsoft introduced that the chips will begin to roll out this yr.
