NEW YORK, January 6, 2026, 09:42 EST — Regular session
- Nvidia shares dip early Tuesday after CEO Jensen Huang says Rubin platform is in “full production” at CES.
- Company says Rubin aims to cut AI inference costs and curb the number of GPUs needed for some model training versus Blackwell.
- Investors watch for more detail at a J.P. Morgan CES fireside chat and Nvidia’s Feb. 25 results.
Nvidia shares edged lower in early trading on Tuesday after Chief Executive Jensen Huang said at CES in Las Vegas that the chipmaker’s next AI computing platform, called Rubin, is now in full production. The stock was down 0.4% at $188.12 as of 9:42 a.m. EST.
The update matters because Nvidia’s valuation hinges on keeping its product cadence and avoiding delays as customers build out data centers for artificial intelligence. “This CES announcement around Rubin is to tell investors, ‘We’re on track,’” said Austin Lyons, an analyst at Creative Strategists who writes the Chipstrat semiconductor newsletter. WIRED
Nvidia said Rubin combines six chips — including a Vera CPU and Rubin GPU — into a rack-scale system and targets steep cuts in AI computing costs versus the current Blackwell generation. It said Rubin can reduce “inference” costs — running a trained model to generate an answer — by up to 10 times, and train some mixture-of-experts models with four times fewer GPUs; a token is a unit of text that models process. “Rubin arrives at exactly the right moment, as AI computing demand for both training and inference is going through the roof,” Huang said. NVIDIA Newsroom
The company also pointed to early demand signals from cloud and AI developers as it tries to extend its lead in data-center chips. It said Rubin-based products will be available from partners in the second half of 2026 and listed Amazon Web Services, Google Cloud, Microsoft and Oracle Cloud Infrastructure among early adopters, alongside AI cloud provider CoreWeave. “Intelligence scales with compute,” OpenAI CEO Sam Altman said in Nvidia’s announcement. NVIDIA Investor Relations
In a technical briefing, Nvidia framed Rubin as a broader “AI factory” platform that ties compute to networking and infrastructure, not just a faster GPU. It described a Spectrum-X Ethernet Photonics network fabric — optics integrated into switching systems — designed to cut power use and improve uptime, and outlined an “inference context memory storage” tier meant to keep large models responsive when they handle longer prompts. NVIDIA Developer
Chip stocks were firmer overall, even as Nvidia lagged. The iShares Semiconductor ETF was up 1.4%, while Advanced Micro Devices fell 1.1% and Broadcom dropped 1.2%; the SPDR S&P 500 ETF was up about 0.6%.
Traders will look for follow-through at a J.P. Morgan CES fireside chat later on Tuesday, where investors expect fresh color on the Blackwell ramp, supply constraints and what “full production” means in practice for Rubin. Any hints on timing, pricing and gross-margin impact are likely to drive near-term positioning. NVIDIA Investor Relations
But the new platform is still a 2026 story, and the risk is that schedules slip, customer spending cools, or large buyers accelerate efforts to design more of their own chips and reduce reliance on Nvidia’s hardware.
