Microsoft extended its partnership with Nvidia by bringing the latter's generative artificial intelligence (AI) foundry service to Microsoft Azure. The move is part of a broader joint initiative at Microsoft and Nvidia focused on offloading general purpose computing and accelerating software to improve data center energy efficiency, increase performance, minimize carbon emissions and lower costs for customers.

The generative AI (genAI) foundry service combines Nvidia's AI foundation models, NeMo frameworks and tools with DGX Cloud AI supercomputing services to offer an end-to-end platform for building custom genAI models on Microsoft Azure.

Microsoft CEO Satya Nadella brought Nvidia CEO Jensen Huang on stage during the opening keynote at Ignite 2023 to highlight the growing accelerated computing partnership between the two tech giants. "Our collaboration extends across the entirety of the stack," Nadella said.

And together, Microsoft and Nvidia claim to have built the world's two fastest AI supercomputers — "one in your house, one in my house," Huang said. And while the process of planning and standing up machines like that traditionally takes a few years, these two tech powerhouses only needed a matter of months. And "seemingly without even trying, it's the third-fastest supercomputer on the planet," Huang touted.

Nonetheless, AI and accelerated computing is a data center challenge that spans the entire stack. "From chips to APIs, everything has been transformed as a result of genAI," Huang said. He described genAI as "the single most significant platform transition in computing history."

As genAI has opened up opportunities for any enterprise to leverage AI, the technology "for the very first time is now useful, versatile and quite frankly easy to use," Huang said. Modern enterprises are approaching AI in three main ways: through public cloud services like ChatGPT, embedded into applications like Microsoft's Copilot software, and by using proprietary data to train and leverage custom AI models.

With the release of Nvidia AI Foundry on Azure, Nvidia aims to help customers build proprietary large language models (LLMs) in the same way TSMC helps Nvidia build GPUs. "We'll be a foundry for AI," Huang said.

AI has brought with it the demand for an entirely new type of data center. "Unlike the data centers of the past, this data center is dedicated to one job, and one job only: running AI models and generating intelligence," Huang said. "It's an AI factory."

If the new hardware segment is AI factories, then the new software segment is copilots — "brand new things that the world has never had the opportunity to enjoy." He's referring to the tool originally launched as GitHub Copilot, an AI-powered code-writing tool for developers that spurred some early controversy around who owns the code used to train Copilot's AI model.

Since then, Microsoft has taken the copilot concept and applied it to a much broader spectrum of tasks assisted by genAI. And copilots will kick off the second wave of AI adoption, according to Huang. (The first wave consisted of the rise of genAI startups like OpenAI.)

The upcoming third wave, however, will be the largest, he said. "This is where Nvidia Omniverse and genAI [are] going to come together to help heavy industries digitalize and benefit from from genAI."