Nvidia has said it expects tighter US export controls on its China-bound AI chips to wipe out $8 billion in sales in the coming quarter, even as the chip giant reported strong earnings and revenue for the first quarter of its fiscal year 2026, which ended on April 27.
The US-based chip designer recorded year-over-year (YoY) growth of 73 per cent driven by surging demand in other global markets. Nvidia shares rose by six per cent the day after it reported earnings on Wednesday, May 28. The company reported revenue of $44.06 billion for the quarter, up by 69 per cent YoY, beating analyst estimates of $43.3 billion. Nvidia further said it expects about $45 billion in sales in Q2 FY26.
Nvidia’s quarterly report has allayed fears among investors that demand for its Graphics Processing Units (GPUs), used to train and deploy AI models like OpenAI’s o3, is slowing down.
The company’s quarterly earnings were also closely watched by the tech industry for clues on how growing uncertainty from US President Donald Trump’s tariff policies and export restrictions on AI chips to China, will affect Nvidia’s international business.
China is a key market for Nvidia, accounting for 13 per cent of its sales in the past financial year. In a bid to maintain its lead in the high-stakes AI race, the US government has sought to prevent the most advanced chips from being sold in China by tightening export rules.
While Chinese competitors still lag behind Nvidia’s cutting-edge chip design technology, several analysts have warned that the gap is narrowing. They have pointed out that US efforts to curb China’s access to advanced semiconductors have not been successful due to loopholes and existing chip stockpiles in China. Some have even argued that the restrictions have counter-intuitively intensified competition against one of America’s most valuable companies while cutting off access to a key growth market.
Nvidia’s H20 chip
The H20 processor was specifically designed by Nvidia to comply with existing limits by the US government. Chinese AI startup DeepSeek’s R1 model is believed to be powered by a mixed stack of Nvidia’s H800, H100, and H20 GPUs. Following the debut of DeepSeek’s low-cost, compute-efficient AI model, several Chinese companies including Tencent, Alibaba, and TikTok parent ByteDance reportedly ramped up orders for H20 chips.
Nvidia H200 Tensor GPU, which has the same architecture as the H20. (Image: Nvidia)
The H20 is not Nvidia’s most advanced chip and was primarily designed to get as close as possible to US export limits. However, it reportedly offers the same compute as advanced chips for a crucial step in developing large language models (LLMs) known as inference. This process involves running an LLM on previously seen data to identify the patterns it has learnt after the training stage.
During the first quarter of fiscal 2026, the Trump administration reportedly told Nvidia that the H20 chip would require a separate export licence to be sold in China. Besides the expected $8 billion hit, Nvidia said it incurred $4.5 billion in charges related to excess inventory for the H20 chip.
If not for the China-related charge, Nvidia said its gross margin of 61 per cent for the quarter would have been 71.3 per cent. Extra sales worth $2.5 billion would have been recorded in the reported quarter too, the company added.
Notably, US export restrictions on chip sales to China show no signs of easing. The Trump administration has also ordered more companies, including firms that develop design software for semiconductors, to stop selling to China in the absence of a licence, according to a report by Reuters.
‘China is closed to US companies’
In an earnings call on Wednesday, Nvidia CEO Jensen Huang told investors that the $50 billion market in China for AI chips is “effectively closed to US industry.”
“The H20 export ban ended our Hopper data center business in China. We cannot reduce Hopper further to comply. As a result, we are taking a multibillion-dollar write-off on inventory that cannot be sold or repurposed. We are exploring limited ways to compete, but Hopper is no longer an option. China’s AI moves on with or without US chips,” Huang said.
Speaking at the annual Computex tech trade show held in Taiwan earlier this month, Huang said that Nvidia’s market share in China had fallen from 95 per cent to 50 per cent over the past four years.
However, on the earnings call, he also said that global demand for Nvidia’s AI infrastructure was still going strong. “Countries are racing to build national AI platforms to elevate their digital capabilities,” Huang added.
The Trump administration has also rescinded the AI diffusion rule which was introduced by former US President Joe Biden at the start of the year. This rule essentially limited the number of US-made AI chips that could be sold in international markets, including India, without special approval by the US government.
The moat
Nvidia and other US chip designers have long been lobbying against export controls to China as they fear losing out on a key market. Meanwhile, the success of DeepSeek and reports of Huawei’s chip progress have led many to question the effectiveness of US-imposed chip controls.
But even as these restrictions become more stringent, Nvidia’s business is not likely to struggle as it continues to sell to major hyperscalers like Meta and Microsoft. “On average, major hyperscalers are each deploying nearly 1,000 NBL72 racks, or 72,000 Blackwell GPUs per week, and are on track to further ramp output this quarter,” Colette Kress, the chief financial officer of Nvidia, said on the call.
It is also reducing its reliance on big tech buyers by supplying for national AI initiatives such as the recently announced 500-megawatt AI infrastructure project in Saudi Arabia and a 5-gigawatt AI campus in the UAE under the US-led Stargate Project. “Sovereign AI is a new growth engine for NVIDIA Corporation,” Huang said on the earnings call.
Nvidia’s net income rose by 26 per cent to $18.8 billion, up from $14.9 billion a year earlier. Sales from AI chips and related parts accounted for 88 per cent of the company’s total revenue. Nvidia further said that half of the revenue from its data center business was from large cloud providers while the company’s network products, commonly used to connect stacks of chips for AI research, saw $5 billion in sales.
Meanwhile, the company’s gaming division saw 42 per cent growth YoY to $3.8 billion. Besides being essential for AI development, Nvidia’s chips are also used for gaming applications. The upcoming Nintendo Switch 2 gaming console will also be powered by a processor designed by Nvidia.
Nvidia’s GB200 NVL72
Besides export controls, Nvidia shareholders were also focused on the company’s rollout of its new GB200 NVL72 ‘thinking machines’ that comprise 72 Blackwell GPUs and cost around $3 million, according to a report by TechCrunch. These systems are specially designed for reasoning.
NVIDIA GB200 NVL72 delivers trillion-parameter LLM training and real-time Inference. (Image: Nvidia)
The new hardware began shipping in February this year, soon after the disruption caused by the launch of DeepSeek’s R1. Many analysts had estimated that the chaos around DeepSeek would reduce the unit’s estimated shipments by half.
In its results, the company highlighted strong momentum in Blackwell-based systems as the GB200 NVL72 machines ramped to full-scale production during the quarter “across system makers and cloud service providers.” “Microsoft, for example, has already deployed tens of thousands of Blackwell GPUs and is expected to ramp to hundreds of thousands of GB200s with OpenAI as one of its key customers,” Nvidia CFO Colette Kress said on the call.
Jensen Huang, chief executive of Nvidia, speaks alongside President Donald Trump during the Investing in America event at the White House in Washington, on Wednesday, April 30, 2025. (Pete Marovich/The New York Times) 
Nvidia H200 Tensor GPU, which has the same architecture as the H20. (Image: Nvidia)
NVIDIA GB200 NVL72 delivers trillion-parameter LLM training and real-time Inference. (Image: Nvidia)
Share
