Nvidia counts the cost of China export curbs

Nvidia reported results for the first quarter of its fiscal year, which ended on 27 April. The figures topped expectations, sending the shares up 5% after hours.

Key figures:

• EPS: $0.96 versus $0.93 expected;
• revenue: $44.06 billion versus $43.31 billion forecast;
• net income: $18.8 billion.

Revenue rose 69% from a year earlier. The company expects about $45 billion in the second quarter. If not for new export restrictions on the H20, the figure could have been $8 billion higher, the chipmaker said.

In the first quarter Nvidia recorded $4.5 billion in charges for excess H20 inventory and missed $2.5 billion in potential additional sales.

CEO Jensen Huang told investors on a call that China’s $50 billion AI‑chip market is “effectively closed to American industry.”

“The export ban on the H20 put an end to our Hopper data-center business in China,” he said.

Huang said the company is studying ways to continue competing in China’s AI market.

In May, media learned of the firm’s plans to release a new AI chipset for the Chinese market at a significantly lower price than the H20 models.

“China is one of the world’s biggest AI markets and a springboard to global success: half of the world’s AI researchers are based there. The platform that wins in China can today lead at the global level,” Huang said.

He argued that shielding Chinese chipmakers from American competition only strengthens them and weakens the United States. Other players in China are filling the void left by American firms, and their technologies are becoming more powerful.

Huang warned that the gap with Chinese alternatives is narrowing. Huawei’s newest AI chip is comparable in performance to Nvidia’s own H200.

At the same time, the chipmaker faces challenges in China because of export restrictions.

“Regulators in the PRC are examining whether compliance with applicable U.S. export-control measures constitutes unfair discrimination against customers in the local market. If the supervisory authorities conclude that we have failed to fulfill our obligations or violated any local law, we will be fined and restricted in our ability to do business,” the company said.

New hurdles

Restrictions on technology sales have hit more than Nvidia. Shares of EDA software makers Cadence and Synopsys fell 11% and 10%, respectively, after the Financial Times reported a White House ban on selling their products to customers in China.

The Commerce Department’s Bureau of Industry and Security (BIS) sent a letter to both companies, as well as Siemens, the paper said.

“We are aware of the reports and speculation, but Synopsys has not received notice from BIS,” said Synopsys CEO Sassine Ghazi.

HP has also been affected, warning of “trade-related” costs amid existing tariffs and a slowing economy.

Tariffs — blocked

Meanwhile, a US trade court blocked Donald Trump’s tariffs from taking effect. It ruled that the president exceeded his authority by imposing sweeping import duties.

The US Constitution grants Congress exclusive authority over foreign commerce, and those powers cannot be overridden by the president even under the pretext of protecting the American economy, the notice said.

“The court does not decide on the reasonableness or likely effectiveness of the president’s use of tariffs as leverage. Such use is impermissible not because it is unreasonable or ineffective, but because [federal law] does not permit it,” a three-judge panel said.

The court invalidated all of Trump’s tariff orders since January that were based on the International Emergency Economic Powers Act. They do not concern certain sectoral tariffs on cars, steel and aluminium, which rely on a different statute.

The Trump administration has appealed.

Export curbs are not the main issue

Many awaited Nvidia’s report to gauge the impact of export restrictions on chips. However, in the view of Kevin Cook, senior equity strategist at Zacks Investment Research, that is far from the central question.

The expert said a much more important area is the company’s rollout of new equipment, GB200 NVL72 — an exascale supercomputer in a single rack. Shipments began in February.

Exascale is a new frontier in supercomputing, enabling:

• training AI models in days rather than months;
• accurate climate, nuclear, biomedical and other models;
• ultra‑large‑scale simulations.

The GB200 NVL72 includes 72 GPUs and costs about $3 million. As shipments have only just started, there is no clear read on progress, Cook noted.

“If Jensen [Huang] says they will ship 10,000 units in the second quarter, the Street [Wall Street] will be very impressed. 10,000 is $30 billion for a $3 million product. I think they will do fewer than 5,000,” Cook said.

Those results will offer a read on companies’ appetite for the newest AI technologies, the expert said.

In April, Trump announced new tariffs for trading partners.