President Trump announced Monday he will allow California-based Nvidia to sell its advanced H200 computer chips to “approved customers” in China, a boost to the semiconductor giant whose chips are widely used for artificial intelligence. 

The United States will take a 25% cut of chip sales to China, the president said in a Truth Social post. 

The deal does not include Nvidia’s more advanced Blackwell system or its soon-to-be-released Rubin system, Mr. Trump said.

He said he informed Chinese President Xi Jinping of his decision, and the Chinese leader “responded positively.” The sales will be carried out “under conditions that allow for continued strong National Security,” Mr. Trump added.

Details are being finalized by the Department of Commerce, according to Mr. Trump, who said “the same approach” will apply to other U.S.-based chip companies, including AMD and Intel.

A Nvidia spokesperson lauded the president’s decision Monday, saying it “strikes a thoughtful balance that is great for America.”

Nvidia specializes in designing advanced computer processors that are used to power artificial intelligence. In recent years, the AI boom has turned the firm into the U.S.’s most valuable publicly traded company, with a market capitalization of around $4.5 trillion as of Monday.

The Biden and Trump administrations restricted exports of advanced chip technology to some Chinese entities, warning the chips could be used for military purposes or could help Beijing build up its AI industry in ways that could be contrary to U.S. interests. Members of both parties have supported limits to AI chip exports, both on national security grounds and to give the U.S. artificial intelligence industry an edge over China.

Nvidia has pressed for greater access to the Chinese market. CEO Jensen Huang argues the U.S.’s export rules have backfired by compelling China to build up its own homegrown AI chip industry.

Huang has met with Mr. Trump several times to discuss export controls, including last week.

Meanwhile, Mr. Trump has sought to strike deals with U.S. chipmakers. In August, the Trump administration planned to grant export licenses to Nvidia and AMD in exchange for the U.S. government taking a 15% cut of revenue from sales to China. The administration also said over the summer that the U.S. government would take a 10% stake in Intel in exchange for federal grants to boost semiconductor manufacturing.

Nvidia plans to invest $100 billion in OpenAI, giving the ChatGPT maker a huge war chest to build data centers with the chip maker’s chips.

OpenAI will gain access to millions of Nvidia chips in the coming years.

“Everything starts with compute,” Sam Altman, chief executive of OpenAI, said in a joint news release. “Compute infrastructure will be the basis for the economy of the future, and we will utilize what we’re building with Nvidia to both create new AI breakthroughs and empower people and businesses.”

For OpenAI, the $100-billion investment will trickle in progressively as each gigawatt of the 10 planned gigawatts of power needed for the data centers is deployed. The first gigawatt data center is scheduled to go online in the second half of 2026.

“This investment and infrastructure partnership mark the next leap forward — deploying 10 gigawatts to power the next era of intelligence,” Jensen Huang, founder and CEO of Nvidia, said in a news release.

Globally, companies are expected to spend $375 billion on artificial intelligence infrastructure in 2025. Nvidia’s graphics processing units underlie most modern AI data centers. This makes its chips a critical commodity in the AI boom and has propelled it to the forefront of global geopolitics.

Nvidia has become a puppet master of the AI ecosystem. In the last year, the company has been at the center of several big-ticket deals, including a $250-million investment in CoreWeave, a cloud service provider that rents Nvidia GPUs. Soon after the investment, CoreWeave signed a $6.3-billion deal with the chip maker, ensuring any unsold cloud computing capacity will be purchased by Nvidia.

Last week, Nvidia picked up a $5-billion stake in its ailing rival Intel to develop data center and PC products. The same week, Nscale, another London-based cloud service provider that uses Nvidia chips, received a $500–million investment from Nvidia, which was followed by a $2.7–billion commitment to develop AI infrastructure in the U.K.

Critics are concerned that this could be an example of “circular financing,” in which a company invests in its biggest customers who then use the money to buy the investor’s products.

The concern is that this could lead to an overestimation of demand for Nvidia’s chips. Nvidia bulls see the $100-billion investment in OpenAI as another sign of the massive unmet demand for AI as companies and countries race to deploy computing infrastructure.

“We believe the Street is underestimating demand in this 4th Industrial Revolution,” said Dan Ives, senior equity analyst at Wedbush Securities. “We believe this is a multiplier impact as Nvidia is investing in Big Tech’s future and for every $1 Jensen puts into Nvidia, they will get $8 to $10 of revenue over the coming years.”

In 2016, Huang donated one of its first supercomputers, powered by eight GPUs, to OpenAI, which would go on to launch ChatGPT in 2022, marking the beginning of the generative AI boom.

OpenAI is currently valued at $500 billion, with 700 million weakly active users.

In January, alongside Japanese investor Softbank and database provider Oracle, OpenAI committed $100 billion to build AI infrastructure in the United States. As part of the “Stargate project,” the companies have started building out a 4.5-gigawatt data center in Texas. They plan to invest $500 billion over the next four years.

OpenAI signed a deal in May to build out a one-gigawatt data center complex in the United Arab Emirates, a deal that was brokered by the Trump administration. Oracle, Nvidia, Softbank, Cisco and G42, a royals-backed Emirati AI company, are set to back the project.

Nvidia, the world’s leading chipmaker, announced on Thursday that it’s investing $5 billion in Intel and will collaborate with the struggling semiconductor company.

The two companies will team up to work on custom data centers that form the backbone of artificial intelligence infrastructure as well as personal computer products, Nvidia said in a press release.

Nvidia said it will spend $5 billion to buy Intel common stock at $23.28 a share. The investment, which is subject to regulatory approvals, comes a month after the U.S. government took a 10% stake in Intel.

“This historic collaboration tightly couples NVIDIA’s AI and accelerated computing stack with Intel’s CPUs and the vast x86 ecosystem — a fusion of two world-class platforms,” Nvidia CEO Jensen Huang said. “Together, we will expand our ecosystems and lay the foundation for the next era of computing.”

The two companies said they will work on “seamlessly connecting” their architectures.

In morning trading, Intel shares jumped 25%, its biggest one-day percentage gain in decades. Nvidia shares added 2%.

For data centers, Intel will make custom chips that Nvidia will use in its AI infrastructure platforms. While for PC products, Intel will build chips that integrate Nvidia technology.

The agreement provides a lifeline for Intel, which was a Silicon Valley pioneer that enjoyed decades of growth as its processors powered the personal computer boom, but fell into a slump after missing the shift to the mobile computing era unleashed by the iPhone’s 2007 debut.

Intel fell even farther behind in recent years amid the artificial intelligence boom that’s propelled Nvidia into the world’s most valuable company. Intel lost nearly $19 billion last year and another $3.7 billion in the first six months of this year, and expects to slash its workforce by a quarter by the end of 2025.

The U.S. government stepped in last month to secure a 10% stake, making it one of Intel’s biggest shareholders. Federal officials said they invested in Intel in order to bolster U.S. technology and domestic manufacturing.

The deal is “bullish for U.S. tech,” Wedbush Securities analyst Daniel Ives said in a client note.

“This is a game-changer deal for Intel as it now brings them front and center into the AI game,” Ives said. “Along with the recent U.S. government investment for 10% (equity stake in Intel) this has been a golden few weeks for Intel after years of pain and frustration for investors.”

Nvidia, meanwhile, has soared because its specialized chips are underpinning the artificial intelligence boom. The chips, known as graphics processing units, or GPUs, are highly effective at developing powerful AI systems.

The deal between the two chipmakers comes as China moves to be less dependent on U.S. semiconductor technology. This week, Chinese officials reportedly forbade several large domestic technology companies from purchasing Nvidia chips, and Huawei announced that it was expanding its development of AI chips and manufacturing.

While Nvidia and Intel will work together to develop new chips, a manufacturing deal has yet to be struck between the two. The potential access to Intel’s chip foundries by Nvidia poses a risk to Taiwan Semiconductor Manufacturing Company, which currently manufactures the tech giant’s flagship processors.

One thing that can be said about Donald Trump’s transactional approach to policy-making is that, as destructive as it might be to our economic health, it gives business leaders clear options to get what they want out of the White House.

The latest case-in-point are the deals struck by chipmakers Nvidia and AMD to secure licenses to export their products to China. The White House named the price — 15% of their revenues from such sales — and the companies assented willingly.

Never mind that the exports originally had been banned — by the Biden administration and Trump himself — because of national security concerns. Never mind that the U.S. constitution explicitly prohibits charging any tax or duty on exports.

Never mind that a stack of U.S. laws, including the Export Control Reform Act of 2018, which Trump signed, don’t provide a pay-to-play escape clause from export restrictions.

Never mind that the exports may strengthen the domestic industry and even its military of China, a country that has been the consistent target of Trump trade policies.

Despite all that, Trump treated the deals as a win for the U.S. Explaining his side of the conversation when Nvidia asked for relief from the export ban, he related, “I said, ‘If I’m going to do that, I want you to pay us as a country something, because I’m giving you a release.’”

Under the circumstances, it shouldn’t be surprising that some trade professionals and investors see something corrupt in the arrangements. Among them is Christopher Padilla, an export control official under George W. Bush, who told the Washington Post: “Export controls are in place to protect national security, not raise revenue for the government. This arrangement seems like bribery or blackmail, or both.’’

Nvidia, as it happens, has a written anti-corruption policy stating, “We do not tolerate bribery or corruption in our business.” The policy bars “promising, offering, providing, or authorizing the provision of money or anything of value … to obtain, retain, or direct regulatory approvals, contracts, business, or other benefits.”

When I asked Nvidia about the deal, the company referred me to the sole comment it has made in response to questions about it: “We follow rules the U.S. government sets for our participation in worldwide markets.” AMD didn’t respond to my request for comment.

These deals are unprecedented; as the Financial Times observed, citing trade experts, “no US company has ever agreed to pay a portion of their revenues to obtain export licenses.”

There’s no question that Nvidia lobbied ferociously for a lifting of the export ban. The company made a $1-million contribution to Trump’s inaugural committee. Its CEO, Jensen Huang, met directly with Trump to discuss the ban; media reports say that Trump initially demanded a 20% fee, but Huang negotiated it down to 15%.

As has been the case with other Trump-negotiated trade deals, the details of this one are murky in the extreme. The terms haven’t been reduced to writing.

Indeed, White House spokeswoman Karoline Leavitt said Tuesday that its “legality … is still being ironed out by the Department of Commerce,” with an eye toward replicating it with other companies.

Among the questions is how the fee would be paid, and how the money would be spent.

Still, what’s known has caused concern for export regulators, experts and legislators.

“Export controls are a frontline defense in protecting our national security,” tweeted Rep. John Moolenaar (R-Mich.), chair of the House select committee on the Chinese Communist Party, “and we should not set a precedent that incentivizes the Government to grant licenses to sell China technology that will enhance its AI capabilities.”

The effect of this deal on other companies also raises the hackles of economists and trade experts.

“Other American semiconductor companies like Qualcomm and Intel may say, ‘If we develop this cutting-edge chip and the government decides that it has some national security interest in that chip, we might have some of our revenue taken away as well,’” says Kyle Handley, a trade economist at UC San Diego, “so they may decide not to do the R&D investments and the innovation and hire the workers to develop those things.”

The revenue payback “will certainly have a chilling effect,” Handley told me, because the government fee “might make the initial investment appear uneconomical.” That’s especially so if the administration tries to apply the arrangement to industries such as software or pharmaceuticals. The export charge could become a particular burden on startups — Nvidia plainly has enough money to pay the fee, but many other innovative companies wouldn’t.

Whether or how the export tax can be stopped is an open question. For one thing, it’s unclear who would have standing to bring a lawsuit to stop it. Nvidia and AMD have accepted the deal, so they presumably wouldn’t file a case. Companies that fear the imposition of export fees on their own products might have to wait until they could show concrete damage to their own interests in order to bring a case in federal court.

As long as manufacturers such as Nvidia are willing to bow to Trump’s demands, he may have a clear field.

If a legal challenge does emerge, the administration has tried to characterize the export fee as something other than a tax in order to circumvent the constitutional prohibition.

Nvidia developed the H20 chip at the heart of its deal specifically to address an export ban the Biden administration imposed on the company’s sales to China in 2023.

The Chinese government, however, isn’t enamored of the chip. Chinese regulators have been pressuring domestic companies to avoid the chips out of cybersecurity concerns, including suspicion that the chips could contain hidden code that could subject them to outside control. (Nvidia has denied that the chips contain any such back-door exposure.)

The chips also are outmoded for some AI applications compared to the company’s top-of-the-line Blackwell series, which are still subject to a U.S. export ban.

As my colleague Queenie Wong reported, Trump seems to think that he and Nvidia’s Huang had put something over on the Chinese. Trump called the H20 chip “obsolete” and said Huang was “selling a essentially old chip.”

But others say the H20s may yet be preferable to Chinese-designed chips for Chinese firms, although Chinese products are consistently improving.

“The H20 is a potent accelerator of China’s frontier AI capabilities, not an outdated AI chip,” as 20 former government trade officials told Commerce Secretary Howard Lutnick, who oversees trade policy, in a July 28 letter.

“If the U.S. backs off of export controls to China, we believe that China’s next generation of frontier AI will be built on the backs of the H20,” the officials wrote. The chip “will not simply power consumer products or factory logistics; they will enable autonomous weapons systems, intelligence surveillance platforms, and rapid advances in battlefield decision-making.”

How the export deal may affect Nvidia’s top or bottom lines is murky, though its immediate effect doesn’t seem all that significant.

In May, the company announced a $4.5-billion writedown of unsold H20 inventories in the first quarter ended April 27 because of the U.S. ban on H20 sales. But it still recorded $23.3 billion in operating profit for the quarter on sale of $44.1 billion. For its last full fiscal year ended Jan. 26, Nvidia reported a pre-tax profit of $84 billion profit on sales of $130.5 billion.

In stock market terms, Nvidia is the world’s most valuable company, with a market value of $4.4 trillion; its price-earnings multiple is a robust 58.4. With a gain in share price so far this year of nearly 35%, it’s one of a handful of AI-related companies that has kept the market buoyant despite investor concerns about a developing economic slowdown due in part to Trump’s trade policies.

But the company is looking ahead to further incursions into the China market over the long term. “The China market is about $50 billion a year,” Huang told Taiwan-based technology strategist Ben Thompson in May, bemoaning the need to leave behind “the profits that go with that, the scale that goes with that, the ecosystem building that goes with that” while the export ban was in place. So it made sense to allow Nvidia to extract revenue and profits from the cross-border trade.

China is sure to power ahead on AI technology with or without Nvidia’s chips, Huang said — “anybody who thought that one chess move to somehow ban China from H20s would somehow cut off their ability to do AI is deeply uninformed.”