Nvidia delivered at GTC 2026 in San Jose last night — and if you thought Blackwell was big, buckle up. Jensen Huang took the stage and introduced the Vera Rubin platform, and the numbers sound like science fiction.
Vera Rubin: Seven Chips, One Platform
The Vera Rubin platform comprises seven chips, five rack-scale systems, and one supercomputer. At its core sit the new Vera CPU and BlueField-4 STX storage architecture. According to Nvidia, the platform delivers 10x more inference throughput per watt compared to Blackwell — and one-tenth the cost per token.
Huang expects purchase orders worth one trillion dollars for Blackwell and Vera Rubin combined through 2027. One trillion. Let that sink in.
DLSS 5, DGX Station, and a Desktop Supercomputer
Beyond the major chip announcements, there were some notable reveals: DLSS 5 brings photorealistic 4K rendering in real time to local hardware. The new DGX Station is a deskside supercomputer with 748 GB of coherent memory and 20 petaflops of AI performance. And the RTX PRO 4500 Blackwell Server Edition delivers 100x performance gains for vision AI.
Feynman: What Comes After Vera Rubin
Huang also teased the next-next generation: Feynman. Named after the physicist, it’ll come with a new CPU called Rosa (after Rosalind Franklin), the LP40 chip, and BlueField-5. The succession plan is already in the pipeline.
AI Data Centers in Orbit
The most surprising moment: Nvidia is working on Space-1 Vera Rubin — AI data centers in orbit. Yes, really. Accelerated computing is meant to expand from Earth-based data centers into space. Details are thin, but the vision alone is staggering.
Partnerships Everywhere
AWS is deploying over one million Nvidia GPUs. Microsoft has installed hundreds of thousands of liquid-cooled Grace Blackwell GPUs in Azure data centers and will be the first hyperscaler to power Vera Rubin NVL72 systems. Google Cloud is integrating Nvidia’s acceleration libraries. And on the autonomous driving front, BYD, Hyundai, and Nissan are joining as new robotaxi partners — plus a partnership with Uber.
My Take
GTC 2026 makes one thing crystal clear: the AI infrastructure boom isn’t slowing down. It’s accelerating. Nvidia isn’t just building GPUs anymore — they’re building the entire platform, from chip to rack to orbit. For developers and businesses, this means cost per token keeps falling, speed keeps rising, and use cases keep expanding in every direction.
The most interesting question isn’t whether Vera Rubin will succeed — that’s almost certain given Nvidia’s market dominance. It’s whether the rest of the industry can keep up.
Sources: