Microsoft spent hundreds of millions on Azure infrastructure to make ChatGPT happen

Microsoft Corp. at the moment revealed the way it spent billions of {dollars} to allow OpenAI LLC to deliver ChatGPT to life.

The partnership between Microsoft and OpenAI has gotten numerous consideration not too long ago. In January, the Home windows maker introduced it was investing $10 billion into the unreal intelligence startup that created ChatGPT, an AI chatbot that has taken the internet by storm because of its spectacular, humanlike conversational capability.

Nonetheless, the partnership between Microsoft and OpenAI truly started a number of years in the past. Based on a report by Bloomberg, Microsoft had already spent “a number of hundred million {dollars}” previous to this yr on the computing infrastructure required to develop ChatGPT.

The cash was spent to construct a large supercomputer that might be used to coach ChatGPT, Bloomberg stated. And in a pair of weblog posts at the moment, Microsoft mentioned what went into constructing the AI infrastructure and the way it’s planning to make that system much more sturdy, so it may well energy extra superior fashions.

To create its Azure supercomputer that powers OpenAI’s initiatives, Microsoft purchased and linked collectively hundreds of graphics processing items made by Nvidia Corp. By linking all of those GPUs collectively, Microsoft was in a position to present the big computing energy OpenAI wanted to coach more and more succesful AI fashions.

Whereas the “tons of of tens of millions” spent by Microsoft may not look like a lot for an organization that may throw billions of {dollars} at promising startups, it additionally reveals how AI has develop into one in all its largest priorities.

That a lot is clear as a result of Microsoft revealed at the moment that the corporate is now working to enhance Azure’s AI capabilities and make it much more highly effective. Matt Vegas, principal product supervisor of Azure HPC+AI, stated Azure has not too long ago been upgraded with new digital machines that make use of Nvidia’s H100 and A100 Tensor Core GPUs which might be linked along with Quantum-2 InfiniBand, an accelerated networking structure revealed by the graphics chipmaker final yr. Microsoft reckons this may permit OpenAI and different AI companies that use Azure’s infrastructure to coach even bigger and extra advanced fashions.

Eric Boyd, Microsoft’s company vice chairman of Azure AI, stated in an announcement that the corporate foresaw the necessity for particular function clusters that concentrate on enabling massive coaching workloads, and that OpenAI was one of many earliest proof factors. “We labored carefully with them to study what are the important thing issues they have been in search of as they constructed out their coaching environments and what have been the important thing issues they want,” he stated.

Photograph: Emiliano Vittoriosi/Unsplash