Microsoft spent a whole lot of tens of millions of {dollars} on a ChatGPT supercomputer

To construct the supercomputer that powers OpenAI’s initiatives, Microsoft says it linked collectively hundreds of Nvidia graphics processing models (GPUs) on its Azure cloud computing platform. In flip, this allowed OpenAI to coach more and more highly effective fashions and “unlocked the AI capabilities” of instruments like ChatGPT and Bing.

Scott Guthrie, Microsoft’s vice chairman of AI and cloud, mentioned the corporate spent a number of a whole lot of tens of millions of {dollars} on the challenge, based on an announcement given to Bloomberg. And whereas which will seem to be a drop within the bucket for Microsoft, which just lately prolonged its multiyear, multibillion-dollar funding in OpenAI, it actually demonstrates that it’s keen to throw much more cash on the AI house.

Microsoft’s already working to make Azure’s AI capabilities much more highly effective

Microsoft’s already working to make Azure’s AI capabilities much more highly effective with the launch of its new digital machines that use Nvidia’s H100 and A100 Tensor Core GPUs, in addition to Quantum-2 InfiniBand networking, a challenge each firms teased final yr. Based on Microsoft, this could permit OpenAI and different firms that depend on Azure to coach bigger and extra complicated AI fashions.

“We noticed that we would wish to construct particular function clusters specializing in enabling giant coaching workloads and OpenAI was one of many early proof factors for that,” Eric Boyd, Microsoft’s company vice chairman of Azure AI, says in an announcement. “We labored carefully with them to be taught what are the important thing issues they had been in search of as they constructed out their coaching environments and what had been the important thing issues they want.”