NVIDIA's major production adjustment: H200 gives way to next-generation chip Vera Rubin
Nvidia is shifting its manufacturing capacity at TSMC from the H200 chip to the next-generation Vera Rubin architecture, a strategic change reflecting the company’s latest assessment of the current regulatory environment.
On Thursday, according to the Financial Times citing two sources familiar with the matter, Nvidia has suspended H200 chip production for certain markets and has reassigned TSMC’s H200-related production capacity to Vera Rubin.
Nvidia CFO Colette Kress said in last week’s financial call that although the US government has approved limited exports of H200 products, “we have not generated any revenue so far, nor do we know whether any exports will be approved.”
This adjustment means Nvidia is no longer counting on large-scale sales of the H200 in the near term. Sources say this move could, to some degree, accelerate the delivery and launch process for Vera Rubin, which is currently in strong demand from tech giants like OpenAI and Google.
H200 capacity shelved, Vera Rubin takes over
H200 is Nvidia’s earlier-generation AI processor, previously positioned as a compliant product meeting US export controls. Vera Rubin is Nvidia’s latest chip architecture released earlier this year, designed for more complex AI systems and the booming demand in the high-end computing market.
According to a source inside Nvidia speaking to the Financial Times, Nvidia has now manufactured about 250,000 H200 chips. If only limited orders are ultimately approved, the existing inventory is sufficient to meet corresponding demand.
“Rather than waiting amidst uncertainty, Nvidia must turn to those goals it can be sure of achieving, especially when advanced products are in short supply,” said one insider.
Approval process stalled, supply chain ramp up in vain
Nvidia had previously pushed hard for permission to sell the H200. After signals of sales permission emerged from President Trump in December last year, Nvidia quickly ramped up production. According to earlier Financial Times reports, the company expected customer orders to exceed one million units, suppliers rushed to stock up around the clock, and deliveries were planned to start as early as March this year.
Nvidia CEO Jensen Huang said at the beginning of January this year that demand was “very strong, we have activated the supply chain, and H200 is being produced on the assembly line.” However, the approval process subsequently stalled.
Despite the shift in capacity, Nvidia has not completely ruled out restarting the H200. According to sources, if there’s a policy breakthrough, it would take no more than three months for Nvidia to reallocate or increase H200 supply chain capacity, and current inventory is sufficient to cover demand and deliveries during this period.
Risk warning and disclaimerThe market carries risks; investment should be made cautiously. This article does not constitute personal investment advice, nor does it consider the special investment goals, financial situation, or needs of individual users. Users should consider whether any opinion, viewpoint, or conclusion in this article suits their own particular circumstances. Invest accordingly and at your own risk.