Nvidia Corporation has reportedly ceased production of its H200 artificial intelligence chips destined for the Chinese market, according to a Thursday report from the Financial Times. The semiconductor giant is reallocating its manufacturing capacity at Taiwan Semiconductor Manufacturing Co. to focus on its upcoming Vera Rubin AI hardware platform instead.
Despite the production halt, Nvidia confirmed it possesses valid U.S. government licenses that permit the shipment of "small amounts" of H200 processors to certain customers in China. However, company officials and U.S. Commerce Department export enforcement chief David Peters have indicated that no H200 chips have actually been delivered to Chinese buyers to date. Peters told lawmakers last month that "my understanding is that none so far," citing ongoing delays due to regulatory "guardrails"—built-in sales conditions and compliance checks.
Regulatory Environment Shapes Supply Decisions
The strategic shift underscores how U.S. export controls and approval processes are increasingly dictating Nvidia's supply chain movements, even amid sustained strong global demand for its data-center processors. The H200 sits among Nvidia's premier data-center offerings, designed to power server farms that train and operate large-scale AI models. Shipments to China have faced significant obstacles, with regulatory restrictions from both Washington and Beijing slowing the flow of advanced computing semiconductors.
In early trading on Thursday, Nvidia shares declined approximately 0.3% following the report. The company recently noted that its current-quarter financial outlook excludes any potential revenue from data-center chip sales to China, highlighting the market's uncertainty. Back in January, Washington granted approval for H200 sales to China, but analysts like Seaport Research's Jay Goldberg described the restrictions as a difficult-to-enforce compromise, remarking that "put another way, this looks like a Band-Aid."
Transition to Next-Generation Platform
Nvidia's manufacturing reallocation signals the beginning of its next product phase. In January, the company announced that its Rubin platform—named for astronomer Vera Rubin—had reached "full production," with Rubin-based products scheduled to launch through partners in the second half of 2026. Chief Executive Jensen Huang stated that "Rubin arrives at exactly the right moment," emphasizing the platform's strategic timing.
During a Wednesday appearance, Huang also addressed Nvidia's investment strategy, suggesting the company's recent equity moves in top clients might be tapering off as several prepare for initial public offerings. When questioned about a potential $100 billion investment in OpenAI, Huang responded "not in the cards." He indicated that Nvidia's latest investment rounds in both OpenAI and Anthropic could represent its final private investments in those companies before they enter public markets.
The situation in China remains fluid. Should U.S. policy shift or Beijing alter its approach, demand could potentially resurge. Conversely, stricter enforcement or new limitations would likely freeze orders, prompting Chinese buyers to seek alternative suppliers. At present, reallocating foundry capacity demonstrates Nvidia's focus on products it can realistically deliver under current geopolitical constraints.
This decision leaves the H200 in a state of limbo—a processor designed for regulatory compliance yet still entangled in political complexities. Neither Nvidia nor TSMC has responded to requests for comment regarding the Financial Times report, and independent verification wasn't immediately possible. The development highlights the ongoing challenges facing semiconductor companies navigating the increasingly complex intersection of technology, commerce, and international relations.
