New York, Jan 6, 2026, 16:21 EST — After-hours
- NVDA down about 0.5% after hours; traded $186.84–$192.13 in the regular session
- CFO says U.S. officials are “working feverishly” on export licenses for H200 shipments to China
- Nvidia says its next-gen Rubin platform is in full production; partner systems due in 2H 2026
Nvidia (NVDA) shares dipped in after-hours trading on Tuesday after CEO Jensen Huang said purchase orders, not a public declaration, would signal when Chinese customers can buy the company’s H200 data-center AI chip, a predecessor to its current flagship Blackwell products. The stock was down 0.5% at $187.32, after moving between $186.84 and $192.13 in the regular session.
The update puts the spotlight back on China, where sales of top-end accelerators depend on U.S. export licenses — government approvals to ship. CFO Colette Kress said officials were “working feverishly” on applications and Nvidia still does not know when approvals will land, after President Donald Trump last year reversed a ban on shipping advanced AI chips to China. Nvidia has called for $500 billion in sales from its current Blackwell generation and the forthcoming Rubin chips by the end of the year, she added. MarketScreener
Nvidia used CES to sketch its next step after Blackwell, launching the Vera Rubin platform that pairs a new Vera CPU and Rubin GPU with networking and security chips, the company said. Nvidia said Rubin could reduce inference token cost — what it costs to generate each chunk of text or data in an AI response — by up to 10 times and train mixture-of-experts models, which route requests to specialist sub-models, using about a quarter of the GPUs versus Blackwell. Rubin-based products will be available from partners in the second half of 2026, Nvidia said.
Huang also argued that the next generation would ease data center cooling, saying “no water chillers are necessary for data centres.” The remark knocked shares of Johnson Controls and Trane, two companies that sell cooling equipment into data centers. Barclays analyst Julian Mitchell wrote that “one should not take their comments lightly,” given Nvidia’s influence across the AI supply chain. Reuters
Nvidia and Siemens announced an expanded partnership to build what they called an industrial AI operating system, aimed at speeding up design, simulation and factory operations. The companies said they plan to start in 2026 with a blueprint site at Siemens’ electronics factory in Erlangen, Germany, using Nvidia’s Omniverse software and AI infrastructure.
Investors are also watching how fast rivals such as Advanced Micro Devices can close the gap in data center accelerators, and whether large cloud customers lean more on in-house chips. If Rubin shifts spending toward networks and storage as well as GPUs, that could ripple across suppliers.
For Nvidia stock, the near-term catalyst is concrete: export-license approvals and purchase orders that show whether H200 shipments can restart at volume. Traders will also look for signals that Blackwell supply and pricing remain firm as big customers plan next-wave data center builds.
But license delays or a change in U.S. policy would push out China revenue and complicate inventory planning. The efficiency claims around Rubin and data center cooling are still targets, and customers will need to validate them at scale.
Nvidia’s next scheduled update is its fourth-quarter fiscal 2026 results on Feb. 25. Investors will look for guidance on China shipments, Blackwell demand and the timing of early Rubin deployments.
