The Rubin GPU isn’t going it alone, it comes as part of an ambitious six-chip platform including the Vera CPU, CX9 SuperNIC, NVLink 144 switch, a silicon photonics processor, and the Spectrum-X switch.

During the Q2 2026 earnings call, Nvidia supreme leader Jensen Huang said Rubin would be the third-gen NVLink rack-scale AI supercomputer and insisted everything was on schedule. The goal, apparently, is to “scale into” a $3 trillion to $4 trillion AI infrastructure market.

Huang said: “The chips of the Rubin platform are in fab. Rubin will be our third-generation NVLink rack-scale AI supercomputer with a mature and full-scale supply chain.”

He promised Rubin would bring “a whole bunch of new ideas” and told the cocaine nose jobs of Wall Street to stay tuned for more hype at the next GTC event.

Production has already started for the GB300 AI GPU, part of the Blackwell Ultra platform, which Huang said had “extraordinary” demand. The GB300 is generating tens of billions in revenue while sliding nicely into existing racks built for the GB200.

Nvidia CFO and EVP Colette Kress said the transition had been “seamless” and that by early August, factory builds were cranking out 1,000 racks a week, with that number expected to grow. Full production is already underway and shows no sign of slowing.

Meanwhile, the gaming division hit a record $4.3 billion, up 14 per cent from the last quarter and a whopping 49 per cent year-on-year. That leap is thanks to the ramp-up of Blackwell RTX GPUs and a fresh splatter of GeForce RTX 50 cards flooding the shelves.

“Our gaming revenue was a record $4.3 billion. This was driven by the ramp of Blackwell GeForce GPUs and strong sales continued as we increase supply availability,” Kress said.