Meet the muscle behind Pittsburgh’s AI
Steel City innovation is built on one chipmaker’s popular tool
While Pittsburgh and Carnegie Mellon University have long touted themselves as pioneers of artificial intelligence, much of the power behind that innovation has been driven by one company: Nvidia.
The trillion-dollar chipmaker that got its start in gaming has long propped up the city’s autonomous vehicle development. It is now partnering with or directly powering a slew of Pittsburgh AI projects, including Abridge’s notation platform for doctors, Ansys’ simulation software and Agility Robotics’ friendly warehouse humanoids.
“Any company in Pittsburgh that is using big data, AI or training AI models is going to be relying on Nvidia chips,” Marc Swinnen, Ansys’ director of product marketing, told me in a recent interview.
He said the technology differs from other industries like steel in its growth rate. As soon as one chip is released, a new one that’s “twice as big” is already in the works.
Competitors are only just now catching up, with Google, Intel and Meta all announcing new versions of their own AI chips this week. (At least one Pittsburgh startup is also trying to carve out its own chip profits by focusing on efficiency.)
Meanwhile, Nvidia, which has an estimated 80% of the AI chip market, is looking to take on the world of humanoid robotics, with Pittsburgh’s own two-legged Digit leading the way.
Digit joined a lineup of other human-like bots at Nvidia’s recent summit in California, where it unveiled Project Gr00t, a “comprehensive AI platform for leading humanoid robot companies” including Boston Dynamics and Digit’s maker, Agility Robotics, which designs and tests the robots in Lawrenceville.
“Our collaboration with Nvidia augments our existing computing and simulation tools, and essentially gives us a force multiplier on our ability to do the kind of model building and training that we need to accelerate Digit’s commercial capabilities,” Agility told me by email.
Nvidia has said the focus will help robots understand human speech and “emulate movements by observing human actions,” making it easier for them to move around our world.
Beyond its hardware, Nvidia also is training the Pittsburgh talent behind AI.
The company is supporting Carnegie Mellon University students through its fellowship program
and is backing startups like Downtown-based Preamble AI with development expertise.
Abridge, the startup using AI to help doctors process notes, said its investment from Nvidia — part of a recently closed $150 million funding round — is also a “strategic collaboration around health care AI.” The two are working on research efforts that could expand Abridge’s product to other health care applications.
Chris Paxton, an AI and robotics research scientist based in Pittsburgh who has worked for Nvidia and Meta, said the chips became ubiquitous in part because their software capabilities made them so easy to pair.
“Everyone is using Nvidia, absolutely everyone,” he said. “I would be shocked if you could find a single robotics company in Pittsburgh that does not rely on those cards.”
He noted that between games and AI, there was another big use case for Nvidia’s powerful processors: mining cryptocurrency.