Jensen Huang erasing the competition in Las Vegas?
While half of Wall Street analysts are trying to understand if AI is a bubble, Nvidia's CEO and founder, Huang, took the stage at CES 2026 in Las Vegas and tried to show everyone who's in charge?
If until now people thought Nvidia was just a chip company, Huang is trying to explain why you should update your version?
Jensen claims he doesn't sell silicon, but the infrastructure of the modern world.
The world might have thought it had seen everything, but Jensen is trying to explain that the previous chapters were just the promo?
In an energetic keynote speech, Jensen not only presented chips, he announced an upgrade to the world economy!
He claims that $10 trillion of computing infrastructure from the last decade is undergoing an "upgrade" to AI.
The new engine?
Rubin platform
Jensen presented what Nvidia calls "Vera Rubin" - the first AI platform based on Extreme-codesigned.
What does that mean in plain English?
That Nvidia designed everything - from the individual chip to the server racks and software - as one organic unit to eliminate all bottlenecks.
Rubin platform,
It is a connection of 6 chips designed for data centers,
and includes:
Rubin GPU chip: 50 petaFLOPS performance according to NVFP4 standard.
Vera CPU: A processor that has undergone extreme optimization for data transfer and agentic processing.
Edge connectivity: NVLink 6 network for Scale-up, Spectrum-X photonics Ethernet for Scale-out, along with ConnectX-9 SuperNIC and BlueField-4 DPU.
Sounds complicated and technical - the story converges to one thing
The economy
Jensen claims that Rubin will reduce the cost of token creation to just 10% of the cost on previous platforms.
According to him, the integration of the new KV-Cache layer enables 5x token processing rate, 5x performance per dollar of TCO, and 5x improved energy efficiency.
Jensen is cutting the price and trying to fight anyone who tries to compete with him with inferior hardware.
The open ecosystem: the brain behind the machine
Nvidia is not just building hardware, it is becoming a "pioneering model builder".
Jensen unveiled the family of open models trained on the company's supercomputers, with the aim that every country and company can join the celebration.
The portfolio is divided into six strategic areas:
1 Clara: Healthcare. 2 Earth-2: Climate science. 3 Nemotron: Multimodal models for Inference. 4 Cosmos: Robotics and simulation. 5 GR00T: Embodied AI. 6 Alpamayo: Autonomous driving.
Huang noted that the number of downloads of these models jumps every six months as they become smarter.
This is the classic playbook - give them the software for free -> and sell them the most powerful computer in the world to run it.
The real drama is moving to the physical world? Jensen unveiled Alpamayo, the first open VLA model of its kind for Level 4 autonomous driving.
It includes AlpaSim, a high-fidelity simulation template.
The news on the ground?
The new Mercedes-Benz CLA will be the first vehicle to hit the road with Alpamayo this year in the US.
In robotics, Nvidia presented "NVIDIA Cosmos" which offers "world model" capabilities - realistic video generation, closed-loop interactive simulation, and trajectory prediction.
Nvidia's robotics stack is already adopted by companies such as:
Boston Dynamics.
Caterpillar.
LG Electronics.
Additional partners like Franka Robotics and NEURA Robotics.
To industrialize development, a robot policy evaluation system was launched along with an edge-cloud orchestration platform, which integrates with Microsoft's cloud.
Jetson Thor chip: the new brain for humanoids.
Jetson T4000 (Blackwell) chip: priced at $1,999, 4x performance of the previous generation (1,200 TFLOPS) and only 70W power consumption - for autonomous robots.
The IGX Thor: designed for industrial edge and functional safety, including application in Archer's autonomous flights.
Who's in?
The list of adopters includes all the "who's who":
Google, AWS, Meta, Microsoft, and Oracle (OCI), along with companies like OpenAI, Anthropic, and Elon Musk's xAI.
And also giants like Palantir, ServiceNow, CrowdStrike, and Symantec, which integrate Nvidia's AI deep into their products.
Jensen Huang doesn't just provide the chips - he provides the models, the simulation, and the edge infrastructure.
Is the transition from AI that writes poems to AI that drives trucks and flies planes the next growth engine?