Just after CES 2026, I am still overwhelmed by Jensen Huang's "big move" this time from NVIDIA. Instead of continuing to release consumer graphics cards like every year, the old man unveiled a 2.5-ton "bomb" on stage—the Vera Rubin computing platform. But the interesting part isn't the weight; it's the strategy behind it.

One notable point: NVIDIA usually changes only 1-2 chips per generation, but this time they redesigned 6 types of chips simultaneously. Why? Moore's Law is slowing down, AI models are increasing tenfold every year, so there's no choice but to innovate across the entire system—from CPUs and GPUs to networking.

But just talking about numbers gets boring. The real importance is that Vera Rubin solves real-world difficult problems. Previously, each supercomputer node required 43 cables, took 2 hours to assemble, and was prone to errors. Now, just 6 liquid cooling tubes and 5 minutes are enough. It sounds simple, but it’s highly effective.

As for costs, Vera Rubin needs only a quarter of the previous generation system to train a 100 trillion parameter model. Token calculation is only one-tenth. For data centers worth $50 billion, this means huge savings.

What really caught my attention is how Jensen Huang talks about open source and DeepSeek. He publicly praised DeepSeek V1 and compared it to Kimi k2 as top-tier models. This shows NVIDIA is no longer just selling hardware but is involved in the entire AI ecosystem—from open-source models like Nemotron to practical applications.

Regarding applications, the new physical AI is what makes this launch stand out. NVIDIA demonstrated Alpamayo, a truly reasoning autonomous system, not just following commands. A Mercedes CLA equipped with this technology will launch in the US in Q1 this year. NVIDIA’s "dual safety stack" design ensures that when AI is not confident enough, the system switches to traditional safety mode.

Robots are also a major part. Jensen Huang brought on stage humanoid robots, Boston Dynamics robots, and even Disney robots. He made a quite interesting statement: "The factory is the biggest robot." The idea is that in the future, chip design, system design, and even factory simulation will be accelerated by physical AI.

Overall, NVIDIA’s CES 2026 isn’t about selling "shovels" or graphics cards anymore. It’s a turning point—from focusing on the digital world to conquering the physical world. And with Jensen Huang’s presentation, it seems NVIDIA is ready for the next battle.