Nvidia's board has reportedly and wisely closed the door on Jensen Huang’s return, barely a week after boardroom pressure pushed him to resign as chief executive. A dubious outside campaign is still trying to pry it open. Tech leaders like SpaceX’s Elon Musk and Y Combinator’s Garry Tan have signed a petition demanding his reinstatement, while financier shareholders Brad Gerstner and Gavin Baker have urged Nvidia's board via letter to “avoid a rush to replace Jensen.”That door should stay shut. Over 33 years, Huang built the company at the center of the AI era, but in our view, his conduct nevertheless made his position untenable. He misled directors about undisclosed relationships and oversaw a firm no one could check—a breakdown in trust that was magnified in China, where he advocated loosening export controls even as restricted chips reached Chinese military units and a Shanghai facility tied to Alice Huang’s network. The political push to restore him rests on the billions his most prominent advocates have riding on his customer-financing web and on misleading geopolitical spin, at the expense of genuine shareholder concerns and national defense.In the spirit of the transparency we ask of Huang’s advocates, we note that our upcoming book, When Machines Act: The Promise and Peril of Our Agentic AI Future, includes Huang’s own words, drawn from his commentary at our Yale Chief Executive Leadership Institute CEO forums.A breach of trust For 13 months, Huang maintained through spokespeople that his ties to Alice Huang, head of major customer Megaspeed, were strictly business. Only after the Financial Times reported Washington’s concerns about her did the Nvidia CEO confess the affair to general counsel Tim Teter. Even then, he assured Teter, who spoke with the board, that there had been no others. Four days later, the New York Times uncovered a second affair.The Megaspeed entanglement posed a textbook conflict of interest. Megaspeed bought billions of dollars of scarce processors and is under federal investigation, and Nvidia's own code of conduct flags personal relationships as a conflict risk. Nothing currently in the public record indicates Huang ever disclosed the affair or stepped aside from a single decision involving Megaspeed.Huang’s second relationship crossed a similarly bright governance line in that it involved an executive who reported directly to him. In the past, corporate boards at McDonald’s and Intel ousted chief executives over relationships with employees, and an ally working on Huang’s return conceded, “Having an affair with a subordinate is a fireable offense.”Those lapses unfolded inside an organization seemingly engineered to leave its leader unchecked, with roughly 60 direct reports, family in senior posts, and no clear successor, board chair, or chief compliance officer. When a board of directors can no longer trust a chief executive, removal is their duty. We would argue that Huang’s conduct would be unacceptable of any chief executive, founder or not. China’s recordNowhere did that breakdown in trust carry steeper stakes than in China, where Huang was Washington’s foremost advocate for looser export rules. Documents reviewed by Bloomberg place servers holding Nvidia's H800, H100, and H200 chips inside a Shanghai facility tied to Alice Huang’s network. Some of those units reportedly went first to Aperia Cloud Services, an official Nvidia Cloud Partner in Singapore, before being smuggled into China. The servers were pledged as collateral in January 2025, when all three chips were barred from sale to China, nearly a year before President Donald Trump’s December 2025 decision—lobbied for by Huang and backed by Commerce Secretary Howard Lutnick—to let Nvidia sell the H200 to approved Chinese buyers for a 25% U.S. government cut.Far more alarming, other top-tier Nvidia hardware reached a blacklisted institution with ties to China’s military, where American technology could be turned against the United States. Procurement records document A100 and H100 purchases by People’s Liberation Army units, and Reuters reports that a blacklisted defense university bought two H200s through a reseller in September. Neither source says Nvidia made the sales. Exporting controlled silicon to China’s military or blacklisted entities without a license is illegal regardless.China’s premier AI labs built their newest models on export-controlled hardware too. Their “innovation” rests partly on American chips they were banned from buying and American models they were barred from copying. A forensic study found the fingerprint of Nvidia's Blackwell chips in six Chinese frontier models, and Anthropic says it caught Chinese labs distilling its models at massive scale—raising the question of whether their “innovation” rests partly on American chips they were banned from buying and models they were barred from copying. The study neither pinpoints where the Blackwell hardware sat nor establishes whether its use violated any law.Even without smuggling, Chinese companies tap advanced chips in Southeast Asian data centers built expressly for them, a loophole that was legal until the Commerce Department’s recent rule. The Wall Street Journal reports that Nvidia told many overseas cloud partners to keep Chinese customers below a ceiling of 30% to 49%. A supplier policing that ratio knows the figure, and disclosing it would be an elementary first step toward accountability.In May 2025, three months after Dell cut off Aperia over diversion fears, Huang insisted there was “no evidence of any AI chip diversion.” Yet short seller Culper Research counts billings routed through Singapore swelling from 8.5% of Nvidia's revenue to roughly 20% over three years, hitting 22% in the quarter ended July 2025. The very next quarter, the standalone Singapore line vanished from company filings when Nvidia switched to reporting sales by customer headquarters.In financial statements, Nvidia has said that over 99% of controlled data-center revenue billed in Singapore came from US-based customers. But a buyer that books orders through a U.S. entity is not the same as an American end user. Megaspeed bought most of its hardware through Aivres, the California arm of Inspur, a Chinese server maker which has been blacklisted for supporting China’s military. Suggesting that Singapore’s surging billings reflected nothing beyond American demand strains credulity.Neither Huang nor Nvidia has been formally charged with a crime, and the company maintains it complies with the law. Where the board’s unfinished inquiry would ordinarily counsel patience, it seems clear to us that the push to restore him anyway points to the financial self-interest driving his backers. The money behind the campaignUnder Huang, Nvidia remade itself into its customers’ chief financier, “a central banker for start-ups that needed chips and funding,” as The New York Times phrased it. Through July, its equity stakes ballooned from about $7 billion to $99 billion. It has agreed to conditionally guarantee up to $105 billion in leases for OpenAI’s Ohio campus, and one economist calculates it backstops up to $125 billion of a $500 billion financing program with Wall Street firms.Huang may have personally assembled this financial edifice, but the fractures were showing before the scandal broke. By August, Nvidia was negotiating billions in guarantees for buyers unable to afford its hardware outright. Colette Kress, Nvidia former finance chief and now interim CEO, halved Nvidia's initial commitment to OpenAI, and late that month the firm paused certain credit backstops for cloud operators.Without Huang, smaller “neocloud” operators could face a painful reckoning. Nvidia's backing was widely seen as a hedge against capture by its largest customers—hyperscalers such as Microsoft, Amazon, and Google, which together account for about half its revenue. Heavily indebted neoclouds such as CoreWeave, in which Nvidia holds a 7% stake, depend on a patron, and Huang was among their biggest champions.Now, Huang’s loudest defenders sit in that web. Public filings as of June 30 show Nvidia as the largest holding at Gerstner’s Altimeter, paired with roughly $600 million of CoreWeave. Baker’s Atreides owned both names too, alongside a dominant stake—roughly a third of its reported portfolio—in SpaceX. Nvidia, in turn, held a $21 billion stake in SpaceX in June, and Musk’s company is a major buyer of Nvidia's chips.Of course, financial exposure does not automatically make these investors wrong, but readers deserve to weigh their appeals to due process beside their stake in keeping the financing spigot open. “Many of those deals were a handshake with Jensen,” Musk warned over social media. Supply deals that hang on one man’s handshake leave an enterprise only as sound as his judgment.Let Nvidia's board do its jobThen, there is the claim that Huang fell victim to a political hit. President Trump added to the confusion on Monday, telling reporters that “nobody’s proven that she’s a spy,” in reference to Alice Huang, and that national security “was not put at risk,” despite his own secretary of state, Marco Rubio, reportedly overseeing a memo warning that she was one. Regardless, Nvidia's directors acted on firmer ground—Huang’s assurances to the board clearly seem to have been false.No more persuasive is the rallying cry offered by Anduril founder Palmer Luckey: “China wants Jensen out, so I want him in.” In reality, China wanted Huang to stay in. True, Beijing’s security agencies curbed Nvidia purchases to favor Huawei, but whatever Beijing’s industrial planners may have wanted, China’s AI companies wanted Huang’s chips, and got them. Administration hawks, including Rubio and Scott Bessent, pressed to enforce the very restrictions Huang lobbied to loosen.Financial disclosures reviewed by the Washington Post suggest Trump held between $13 million and $62 million in Nvidia stock at the end of August, up from $615,000 to $1.3 million at the end of 2024, a period when his administration let the company sell advanced chips in China. The White House told The Post that independent managers handle his portfolio. Still, the people setting America’s chip policy and those profiting from it overlap uncomfortably.Huang’s defenders fall back on a related fatalism, arguing Washington had no choice except to keep selling advanced processors to China in exchange for rare earths. However, America depends less on China for these minerals than commonly imagined. What it lacks is sustained investment to extract and process them. One industry estimate puts a domestic mine-to-magnet supply chain at tens of billions of dollars, rising toward $100 billion over time. Against the $1.9 trillion erased from Nvidia's market value since its October peak, roughly 20 times that top figure, building domestic capacity looks cheap.A board that can be overruled by a petition has no function. Nvidia's directors deferred to Huang for years and intervened late, and reversing course under outside pressure now would merely extend a habit of deference. Instead, the board should let its investigation finish, restore standalone Singapore revenue disclosures, and select a successor on the merits. Huang spent 33 years building an enterprise the AI era cannot do without. Now comes the defining test of a great company—whether it can outlast the person who built it.Economists across the political spectrum, from Karl Marx to Joseph Schumpeter, have argued that the rapacious nature of capitalism leads to its own demise. History has proven them wrong so far.
Why Jensen Huang Should Not Be Reinstated as Nvidia's CEO
Full Article
Original Source
Read the full article at Time →KhanList aggregates and links to publicly available news content. We do not host full articles from third-party sources. Always verify important information with original sources.