Nvidia watchers have a lot to celebrate at CES this week, with news that the company’s latest GPU, the Vera Rubin, is now available Fully put into production. After all, those powerful AI chips—the picks and shovels of the AI boom—helped create NVIDIA The world’s most valuable company.
But CEO Jensen Huang made it clear again in his keynote speech that Nvidia does not consider itself just a chip company. It’s also a software company with a presence at nearly every layer of the AI stack, and a major bet on physical AI: artificial intelligence systems that operate in the real world, including robots and self-driving cars.
In a press release promoting Nvidia’s CES announcement, Huang cited a quote declaring that “robotics’ ChatGPT moment has arrived.” Breakthroughs in physical artificial intelligence — models that understand the real world, reason and plan actions — “are unlocking entirely new applications,” he said.
However, in his keynote speech, Huang was more cautious, saying that the ChatGPT moment for physical artificial intelligence is “coming soon.” This may sound nit-picky, but the distinction is important—especially considering the Said at last year’s CESwhen he introduced Nvidia’s Cosmos World Platform and described robotics’ “ChatGPT moment” as “coming soon.”
So has that moment really arrived, or is it still just out of reach?
Huang himself seems to acknowledge this gap. “The challenges are clear,” he said during his keynote address yesterday. “The physical world is diverse and unpredictable.”
In terms of physical artificial intelligence, Nvidia is not a flash in the pan. Over the past decade, the company has laid the foundation by developing an ecosystem of artificial intelligence software, hardware and simulation systems for robots and autonomous vehicles. But it was never about building your own robots or self-driving cars. As Rev Lebaredian, vice president of analog technology at Nvidia said wealth Last year, the strategy was still to provide picks and shovels.
There is no doubt that Nvidia has made progress in this area over the past year. In terms of autonomous driving, today the Alpamayo series of open artificial intelligence models, simulation tools and data sets were launched, aiming to help autonomous vehicles operate safely in a range of rare and complex driving scenarios, which are considered some of the most difficult challenges for autonomous driving systems to safely address.
Nvidia also released new Cosmos and GR00T open models and data for robot learning and inference, with strong promotions including Boston Dynamics, caterpillarFranke robot, humanoid robot, LG Electronics and NEURA Robotics debut new robots and autonomous machines based on Nvidia technology.
Even as models, simulation tools, and computing platforms become more capable, Nvidia is not building self-driving cars or robots themselves. Automakers still must transform these tools into systems that can operate safely on public roads in response to regulatory scrutiny, real-world driving conditions and public acceptance. At the same time, robotics companies must turn artificial intelligence into machines that can reliably manipulate the physical world at scale and at a cost that makes commercial sense.
This work—integrating hardware, software, sensors, safety systems, and real-world constraints—remains very difficult, slow, and capital-intensive. It’s unclear whether faster advances in artificial intelligence will be enough to overcome these obstacles. After all, ChatGPT moments aren’t just about the models behind the scenes. These have been around for several years. It’s about user experience and a company that can capture lightning in a bottle.
Nvidia has caught lightning in a bottle before — and it turns out that GPUs are an unlikely but perfect engine for modern artificial intelligence. Whether that luck can be repeated in the messier and less standardized field of physical artificial intelligence remains an open question.