ada könig
← all pieces
Urtext · 2026.11.10

Love and Computers

The marital infidelity of Nvidia’s chief executive exposes the fragility of the financial structure underpinning artificial intelligence.

On Sunday 8 November, Jensen Huang, founder and chief executive of Nvidia, the California company whose chips power much of American artificial intelligence, wrote a letter to his employees: “I was unfaithful to my wife.” It would be celebrity gossip, were it not for the woman: Alice Huang (no relation), an executive at Megaspeed, a Singapore data-centre company and an Nvidia customer. According to anonymous US intelligence officials quoted by the New York Times, she had been recruited by China’s security services. Beijing dismisses the allegations as slander, Nvidia’s board says it has no reason to believe national security or company secrets were compromised, and no one has been charged.

The political effect, by contrast, is a fact. Three Republican chairs of House of Representatives committees have asked Huang to meet them behind closed doors on 18 November. Asked, not compelled: in June Huang had already declined an invitation from the Democratic senator Elizabeth Warren to testify about chip sales to China. That too is how power works in this story: those who can ask cannot compel.

Why does a private affair matter to the economy? Because Nvidia does more than sell chips. It also stands behind its customers’ obligations: up to $105 billion in guaranteed payments linked to OpenAI’s campus in Ohio, according to its filing with the SEC. The capital itself comes from private credit funds, which lend because Nvidia’s guarantee is in place. The structure rests on the credibility of the guarantor.

My suspicion is that this credibility is now being tested on several fronts at once: by the board’s inquiry, the House chairs’ request and the possibility of a divorce. Neither spouse has announced plans to divorce, so I am speculating. I see two scenarios. In the first scenario there is no prenuptial agreement: that is plausible, since they married in 1985, eight years before Nvidia was founded. In California, assets acquired during marriage are generally subject to equal division, and Lori Huang is co-trustee of a family trust holding more than 528 million Nvidia shares. To pay her, Huang would have to sell shares or borrow against them: another layer of leverage resting on the same foundation. In the second scenario there is an agreement. The pressure to sell shares or borrow against them largely disappears; what remains is credibility and political risk, neither of which depends on the contract. In either scenario, a divorce would today merely amplify a shock already under way, but it could become a factor in its own right if it struck at the point on which the structure depends.

Here comes the word many will use: bubble. Nvidia finances the customers who then buy its chips, and part of its revenue is its own money coming back. Those who dispute the word have arguments that deserve to be taken seriously: the chips and the models are real; so is the revenue; the guarantee has a ceiling; and financing customers is not unprecedented, since it helped build railways and telecommunications networks.

The Trump administration justifies its commitment to AI by pointing to the challenge from China; I suspect it has more to do with who stands behind it. The money is documented: one million dollars each from Altman, Bezos and Zuckerberg to Trump’s inaugural fund, and hundreds of millions from Musk for Republican candidates. The test lies where the two justifications diverge. In January Huang persuaded Trump to allow chip sales to China, with the US government taking a 25% cut. The Times reports that Trump was briefed on the suspicions surrounding Alice Huang but took no action, apparently reluctant to jeopardise Nvidia’s sales and trade relations with Beijing. On 2 November Trump said he did not know “the specific thing” and promised to look into it closely. Back in September, Huang was one of the three men whose phone calls helped stop a proposal for an AI regulator. Where the defence of AI and the defence of AI against China diverge, I suspect the former amounts to a defence of Nvidia. And silence protected nothing: Huang’s admission came just days after the US midterm elections had already sent the markets tumbling.

The 18 November meeting is not what matters most. Three things do, and they can be observed without watching the markets: the findings of Nvidia’s board inquiry; whether Huang agrees to speak behind closed doors to those who are asking; and whether lenders to the companies relying on the guarantee change their terms, demanding tougher clauses or an additional signature. These are signals, not proofs: my suspicion may be wrong, and the structure may hold this time too.

What remains is the question nobody has yet asked clearly: if the guarantee wavers, whose responsibility is it? Nvidia’s, which signed it; the funds that lent against that guarantee and turn it into a sales channel; or the savers and insurers who hold stakes in those funds without having signed anything? I do not know how much of this risk is already embedded in European portfolios. It is a question investors should put to those who manage them. Unless responsibility is defined in advance, in a binding document rather than a statement, it will be settled only after the loss has occurred.