Nvidia CEO Jensen Huang introduces a ‘big, big GPU’ that is ‘pushing the limits of physics’

For his two-hour long keynote address at Nvidia’s GTC developers conference yesterday at the packed SAP Center in San Jose, CEO Jensen Huang was clad in a black leather jacket that was just a bit more rock and roll than the plain versions he has sported over the past few years. With a few silver zippers here, a little shine and texture there, the jacket offered a small clue that, while Huang joked that he hoped the audience realized the gathering was not a concert, Nvidia is one of the past year’s biggest AI rock stars.

But Huang didn’t tout the company’s stock price, which has skyrocketed since OpenAI’s ChatGPT was released in November 2022 and generative AI exploded — with its accompanying hunger for Nvidia’s GPUs. There was no need to mention generative AI models require thousands of GPUs to run — and Nvidia holds over 80% of the GPU market, according to John Peddie Research. Or that OpenAI reportedly used 10,000 Nvidia GPUs to train ChatGPT. Or that Meta CEO Mark Zuckerberg recently said the company plans to buy 350,000 Nvidia H100 GPUs.

Huang said Nvidia’s new Blackwell GPU offers 30x performance increase

Instead, Huang sought to show off Nvidia’s dominance by unveiling the company’s newest AI chip — a “big, big GPU” called Blackwell, which, he said, offers a 30 times performance increase for LLM inference workloads compared to previous iterations and is “pushing the limits of physics.” Like the famous scene in “Jaws” when Roy Scheider’s Chief Brody says “You’re going to need a bigger boat,” Huang explained that the ‘new industrial revolution’ of generative AI requires super-fast data centers and bigger GPUs to handle the mammoth scale of generated tokens, that, in turn, create ‘incredibly valuable software’ in everything from healthcare to robotics.

With a tech-heavy presentation that went deep into the math behind the scale of today’s generative AI models, it was clear that Huang was speaking directly to the developer community that is core to Nvidia’s success. “I want to show you the soul of Nvidia, the soul of our company, at the intersection of computer graphics, physics and artificial intelligence, all intersecting inside a computer,” he said.

Huang did not focus on wider societal impact of AI

With the focus on building, accelerating, exploring strange new worlds, boldly going where no one has gone before — oh, wait, that’s Star Trek — it was also notable that Huang was not speaking to a wider public audience that has many questions about AI that don’t have to do with PFLOPS.

There were no comments about the societal impacts of AI — from impacts to the workforce (other than making it more efficient) and job skills to issues of ethics and trust, bias and misinformation. Other than many mentions of energy ‘efficiency,’ there was no mention of how the manufacturing and use of Nvidia products impact the environment.

Instead, after an appearance by Disney’s adorable Star Wars AI robots, which learned to walk in Nvidia’s robotics simulation platform Isaac Sim, Huang drilled down on his predicted “new industrial revolution” of accelerated computing — with a 100T generative AI industry powered by the Blackwell platform; Nvidia’s new NIM container microservices to easily distribute gen AI software; its NEMO framework and AI foundry; and Omniverse robotics platform.

The keynote ended with a video showing a small Star Trek-like aircraft piloted by an animated Jensen Huang reaching for the stars and “powering the new era of computing.” Nvidia’s developers will certainly be on board. The question is, is the rest of society ready for what might be a rough ride?