It’s hard to say whether Nvidia has ever truly been subtle with its announcements. At last year’s CES, CEO and founder Jensen Huang stunned the industry with the debut of the GeForce RTX 50 series alongside Nvidia Cosmos, its ambitious world-model initiative. This year’s show was more restrained on the consumer GPU front, but the message to CES 2026 attendees was still unmistakable: Nvidia wants it all.
“All” isn’t hyperbole. Nvidia is now the first company ever to surpass a $5 trillion valuation — an almost inconceivable figure — and Huang and company show no signs of slowing down. The company’s ambitions now span factories, autonomous vehicles, robotics, and nearly any domain that can be trained, tested, or perfected in simulation before ever touching the real world. If something can be modeled, Nvidia wants to power it.
Nvidia’s real obsession is physical AI
The biggest buzzword of the night was “physical AI,” Nvidia’s term for AI systems that don’t just generate content but actually act. These models are trained in virtual environments using synthetic data, then deployed into physical machines once they’ve learned how the world works.
Credit: Bridget Bennett/Bloomberg via Getty Images
Huang showcased Cosmos, a world foundation model capable of simulating environments and predicting movement, alongside Alpamayo, a reasoning model specifically designed for autonomous driving. This is the tech Nvidia says will power robots, industrial automation, and self-driving vehicles, as demonstrated by the Mercedes-Benz CLA, which was shown running AI-defined driving on stage. The company also revealed plans to test its own robotaxi service with a partner as soon as 2027, using Level 4 autonomous vehicles capable of driving without human intervention in limited regions.
Nvidia hasn’t announced where the service will launch or with whom it’s partnering, but the move signals a shift from being a behind-the-scenes supplier to actively participating in the self-driving race. Huang has already described robotics — including autonomous vehicles — as Nvidia’s second-most important growth category after AI itself.
No New GPUs
If you were waiting for new consumer GPUs, you probably noticed very quickly that there weren’t any. Nvidia didn’t announce a single new GeForce card, and that felt entirely intentional. Instead, Huang spent most of the keynote talking about Rubin, Nvidia’s next-generation AI platform that’s already in full production.
Rubin is described as more than just a chip, but an entire system. GPUs, CPUs, networking, and storage, all designed together to handle the immense (and environment-altering) compute demands of modern AI models at data center scale. Nvidia framed this as essential to keeping up with skyrocketing AI demand, where training costs, energy use, and bottlenecks are becoming existential problems.
The absence of gaming hardware shouldn’t be considered a snub, but it is clear that Nvidia is no longer driven by gamers. It’s kind of been clear that’s been the case for a while, but today’s conference really drove the nail in the coffin. Instead, the company’s ambitions are driven by hyperscalers, governments, and anyone trying to automate everything that moves.
‘Open’ AI, powered by Nvidia hardware
The third major takeaway was Nvidia’s ongoing push to make itself unavoidable through openness — or at least Nvidia’s version of it. Huang repeatedly emphasized that the company isn’t just selling hardware, but open AI models that developers can actually use, fine-tune, and deploy (not to be confused with ChatGPT developer OpenAI). Nvidia now has open models spanning healthcare, climate science, robotics, embodied intelligence, reasoning AI, and autonomous driving, all trained on Nvidia supercomputers and released as foundational building blocks. They’ve practically become the corn of tech.
Even personal AI agents got some stage time, with demos of local agents running on Nvidia’s DGX Spark hardware. Nvidia aims to be the platform beneath every AI system, from massive data centers to individual desktops. It’s an elegant strategy — sell openness, but still own the pipes.
Taken together, the keynote felt like a declaration. Nvidia isn’t chasing CES hype cycles any more. It’s positioning itself as the backbone of an AI-powered world, where the most important announcements don’t happen on stage, and the most impactful products aren’t meant to fit under your desk.
