Nvidia could be facing significantly steeper financial losses than initially anticipated due to the latest U.S. export restrictions on artificial intelligence (AI) chips. While the company first estimated a $5.5 billion hit from the blocked shipments of its China-specific H20 chips, industry insiders now expect the losses to exceed $10 billion — nearly double the original figure.
The revised estimate stems from growing concerns that Nvidia, along with its Chinese clients, were caught off guard by the scope and timing of Washington’s new export controls. These rules came into effect earlier this week and are designed to limit the transfer of high-performance AI chips to China.
The Financial Times reports that Nvidia had hoped its H20 chips, engineered with reduced capabilities to comply with earlier restrictions, would not fall under the updated export rules. However, that hope appears to have been misplaced. The H20 has been a central product for the Chinese market, where Nvidia generated between $12 billion and $15 billion in revenue last year alone.
Adding to the confusion, Nvidia executives had met with former U.S. President Donald Trump in Florida earlier this month. Following the meeting, the company believed it might be shielded from any new export crackdowns, especially after presenting its ambitious plan to invest $500 billion in the U.S. economy. That optimism proved short-lived.
Nvidia had already assured major Chinese customers — including Alibaba, ByteDance, and Tencent — that their H20 orders would not be disrupted. But under the new regulations, the U.S. government now requires licenses for the export of certain AI chips. The problem is, it remains unclear how and where buyers can apply for these licenses, or what criteria will be used to issue them.
Typically, Nvidia’s AI chips require more than six months of lead time before delivery. As a result, most of this year’s orders haven’t yet been fulfilled and are now subject to the tightened controls. Demand from China has spiked in recent months, particularly after the launch of the local DeepSeek AI model, prompting a surge in H20 orders.
Intel, too, has not been spared. The company’s Gaudi 3 AI accelerator, designed for data centres, is also caught in the crosshairs of the new regulations. While Nvidia has yet to provide specific details on how the export rules will affect its chips, Intel has already briefed its Chinese partners.
According to Intel, any chip with a memory bandwidth of 1400 GB/s, I/O bandwidth of 1100 GB/s, and a total bandwidth exceeding 1700 GB/s now requires a U.S. export license. Both Nvidia’s H20 and Intel’s Gaudi chips surpass these thresholds, meaning future shipments to countries like China must receive explicit approval from Washington.
The implications for both companies — and the broader global semiconductor market — are significant. With China being one of the largest buyers of AI hardware, U.S. tech giants may face growing revenue uncertainties, tighter operational constraints, and heightened geopolitical risks as Washington ramps up its technology-related sanctions.