The United States has established policies based on the assumption that China cannot produce AI (artificial intelligence) chips, but this is clearly incorrect. China already possesses significant AI and manufacturing capabilities, and half of the world's AI researchers are based in China. The question is not whether China will have AI, but who will secure the platforms in the world's largest AI market. Ultimately, the platforms that succeed in China will lead the global AI competition.
Amid the blow from the semiconductor export restrictions imposed by the Trump administration against China, Jensen Huang, Chief Executive Officer (CEO) of Nvidia, expressed criticism regarding the containment of China. On the 28th (local time), during the fiscal year 2026 first-quarter (February-April) earnings conference call, Huang CEO said the company is at risk of losing the 50 billion dollar (approximately 70 trillion won) Chinese AI market due to the export restrictions. Nvidia predicted that losses due to sales restrictions in China for the second quarter (May-July) would significantly rise to over 10 trillion won and stated that they are exploring the development of other products to compete in the Chinese market.
◇ Impact of export restrictions materializes… “Abandoning the Chinese AI market would result in a 70 trillion won loss”
The direct damage from export restrictions on AI chips to China has reached several trillion won in Nvidia's first quarter. The Trump administration implemented restrictions on Nvidia's export of AI chips (H20) to China on April 9 to curb China's AI technological advancement. Nvidia processed 4.5 billion dollars (approximately 6 trillion won) as expense for the disposal of H20 chip inventory, and reported a loss of 2.5 billion dollars (approximately 3 trillion won) in revenue. As a result, the gross profit margin (the ratio of profit remaining after deducting expenses from revenue) declined from an initially expected 71.3% to 60.5%. Colette Kress, Chief Financial Officer (CFO) of Nvidia, explained during the conference call that the company has been selling H20 chips to the Chinese market with approval from the previous (Biden) administration, but the new export controls related to H20 did not provide a grace period for the company to deplete its inventory.
The damage is expected to worsen in the second quarter. CFO Kress stated, “The share of China in our first quarter data center revenue fell short of expectations due to the H20 export controls and declined compared to the previous quarter,” and noted that “the Chinese share will significantly decrease in the second quarter.” Last year, the Chinese share in Nvidia's revenue was 12%. As a result, the company anticipates a revenue loss of 8 billion dollars (approximately 11 trillion won) in the second quarter. Consequently, Nvidia's expected total revenue for the second quarter is 45 billion dollars (approximately 62 trillion won), falling short of the market estimate of 45.9 billion dollars (approximately 6 trillion won).
Nvidia has made it clear that it will not abandon the Chinese market, stating that it is exploring new products for export to China. CFO Kress remarked, “We are still considering limited solutions to supply data center computing products that comply with regulations in China,” adding, “Losing the 50 billion dollar (approximately 70 trillion won) Chinese AI accelerator market would negatively impact the company's business and only benefit our competitors in China and abroad.” Market research firm TrendForce reported that Nvidia plans to launch low-power and low-spec GPUs (graphics processing units) for the Chinese market as early as the second half of this year.
◇ Explosive demand for AI accelerators partially offsets losses in China
Despite the unavoidable impact from China, Nvidia emphasized that the latest "Blackwell" AI accelerators are seeing explosive sales due to a surge in demand for AI inference. The high demand for the high-value product Blackwell is expected to partially offset revenue losses in the Chinese market. In fact, 88% of Nvidia's revenue in the first quarter came from the data center business, which includes AI chips, of which 70% was driven by Blackwell demand. Revenue from the data center sector increased by 73% year-on-year to 39.1 billion dollars (approximately 53.8 trillion won), coming close to the market forecast of 39.2 billion dollars.
CFO Kress stated, “The demand for inference is soaring; for example, Microsoft processed over 100 trillion tokens in the first quarter, five times more than the same period last year,” and noted, “In this quarter, major hyperscalers have already deployed tens of thousands of Blackwell GPUs, and we expect to expand the GB200 products (equipped with Blackwell) to hundreds of thousands with one of our key customers, OpenAI.” Additionally, the high-value GB300 GPU, equipped with 50% more high-bandwidth memory (HBM) than GB200, is expected to start shipping in the latter half of the second quarter.
Nvidia stated that the construction of large-scale AI data centers, so-called "AI factories," is accelerating globally, adding that the demand for AI infrastructure, which is still in its early stages, will continue to grow. CFO Kress disclosed that “Currently, about 100 Nvidia-based AI factories are under construction, doubling the number from the same period last year and the average number of GPUs supplied to each factory has also doubled.” Huang CEO diagnosed that “Almost every country is attempting to build AI infrastructure as a necessity, and we are now at that starting point.”
In this context, it is essential to encourage China to utilize the AI platform of the United States, Huang CEO emphasized. He pointed out that "China's AI continues to develop regardless of whether it has U.S. chips, and rather, the export restrictions have accelerated innovation and scaling in China." He further noted that "AI export controls should be implemented in a way that strengthens the U.S. platform, and we must not push half of the world's AI talent in China to competitors like DeepSeek." The Chinese AI startup DeepSeek caused a global shock earlier this year by offering low-cost, high-performance AI models for free, circumventing U.S. sanctions. Huang CEO reiterated his criticism of the strategic flaws in the current containment measures against China, stating that "the U.S. wins when popular models like DeepSeek's perform best on U.S. infrastructure."