According to a Bloomberg article, Microsoft invested hundreds of millions of dollars in constructing a giant supercomputer to support OpenAI’s ChatGPT chatbot. Microsoft describes how it built Azure’s potent artificial intelligence infrastructure, which OpenAI uses, and how its systems are becoming even more reliable in a pair of blog articles posted on Monday.
Microsoft claims to have connected hundreds of Nvidia graphics processing units (GPUs) on its Azure cloud computing platform to create the supercomputer that supports OpenAI’s initiatives. As a result, tools like ChatGPT and Bing’s AI capabilities were “unlocked” by OpenAI, enabling it to train ever-more potent models.
Microsoft’s vice president of AI and cloud, Scott Guthrie, claimed the business invested hundreds of millions of dollars on the project. And while that might seem insignificant compared to Microsoft’s recent extension of its multiyear, multibillion-dollar investment in OpenAI, it does show that the company is prepared to invest even more money in the AI market.
With the release of its new virtual machines that utilize Nvidia’s H100 and A100 Tensor Core GPUs and Quantum-2 InfiniBand networking, a project both companies previewed last year, Microsoft is already striving to increase the strength of Azure’s AI capabilities. Microsoft claims this should allow OpenAI and other businesses that depend on Azure to train bigger, more sophisticated AI models.
Corporate vice president of Azure AI at Microsoft, Eric Boyd, says, “We identified that we would need to construct special purpose clusters focusing on supporting big training workloads. OpenAI was one of the early proof points for that. We carefully collaborated with them to identify the key things they needed and the important things they were looking for as they built out their training environments.”
Nidhi Chappell, a different Microsoft official, claimed that the initiative made ChatGPT possible and will lead to the development of more AI models.
