Key Highlights
- Trump administration developing comprehensive licensing system requiring federal approval for AI chip exports to nearly all countries worldwide.
- Proposed regulations would trigger reviews for orders exceeding 1,000 Nvidia GB300 GPUs; deployments above 200,000 units need host nation approval.
- Production of H200 chips for Chinese market discontinued at TSMC, with manufacturing resources redirected to Vera Rubin chip line.
- CFO Colette Kress disclosed Nvidia hasn’t earned any revenue from China operations despite receiving US authorization for certain H200 shipments.
- Jensen Huang indicated Nvidia’s $30 billion OpenAI stake could be final investment before anticipated IPO.
Shares of Nvidia $NVDA tumbled approximately 1.7% on Thursday following a pair of reports that raised fresh concerns about the chipmaker’s global market access and China strategy.
A Bloomberg report revealed the Trump administration’s work on sweeping export regulations requiring government authorization for AI chip sales to practically every nation worldwide. The news pushed NVDA and $AMD, which fell roughly 2%, into negative territory during afternoon sessions.
These potential regulations would transform existing restrictions—currently applicable to approximately 40 nations—into a comprehensive global licensing mechanism. According to the draft proposal, shipments containing up to 1,000 Nvidia GB300 GPUs would undergo regulatory review, though certain exemption pathways may exist.
Larger transactions face heightened scrutiny. When deployments surpass 200,000 GB300 units controlled by a single entity within one nation, the host country’s government must participate in the authorization process.
Washington would only approve these massive exports to partner nations that provide security guarantees and commit to investments in US-based AI infrastructure—although specific investment thresholds remain undefined in the current draft.
While not constituting an outright prohibition, these regulations would grant the Commerce Department extensive authority over AI chip distribution that powers platforms like ChatGPT and Gemini.
Chinese Market Revenue Remains at Standstill
In a separate Financial Times report, sources indicated Nvidia has silently discontinued H200 chip manufacturing for the Chinese market at Taiwan Semiconductor Manufacturing Co., redirecting production capacity toward its successor Vera Rubin architecture.
These chip families employ distinct technologies and manufacturing processes—H200 utilizes CoWoS-S packaging with previous-generation high-bandwidth memory, whereas Vera Rubin leverages CoWoS-L alongside the advanced HBM4 specification—meaning the production reallocation doesn’t directly impact either line’s availability.
Nvidia’s Chinese operations have remained in uncertainty for an extended period. Following Trump’s December authorization of H200 exports to China—contingent on a 25% government revenue share—Nvidia had previously marketed the reduced-capability H20 chip in China until the Trump administration prohibited those sales last April.
Despite obtaining regulatory clearance, actual transactions haven’t materialized. During last week’s quarterly earnings discussion, CFO Colette Kress confirmed Nvidia has “yet to generate any revenue” from Chinese operations and expressed uncertainty whether Beijing would permit any imports whatsoever.
Domestic Chinese Competitors Making Progress
Kress highlighted an additional challenge: multiple recent public offerings from Chinese semiconductor manufacturers that she stated “have the potential to disrupt the structure of the global AI industry over the long term.” Nvidia maintains its commitment to ongoing dialogue with both Washington and Beijing.
Regarding OpenAI developments, CEO Jensen Huang stated this week that Nvidia’s $30 billion contribution to OpenAI’s $110 billion funding round completed in late February “might be the last time” the semiconductor giant backs the AI firm, citing expectations of an imminent public offering. Huang also clarified that a previously discussed $100 billion investment arrangement with OpenAI is “not in the cards.”
