I've spent the past fortnight watching what has to be one of the most extraordinary corporate meltdowns in recent history. Nvidia, the company at the centre of the AI revolution, has lost its founder-CEO, shed an enormous amount of market value, and attracted an apparently inexhaustible supply of accusations ranging from accounting fraud to treason.
All because Jensen Huang had an affair with a woman whom American intelligence believes may have been working for China.
Or so the story goes.
I've been watching the coverage with increasing incredulity. Every day brings some new allegation, generally more sensational than the last, and remarkably few seem to require any supporting evidence. The internet has collectively decided that Huang is guilty of everything from deliberately circumventing export controls to handing over the keys to Nvidia's technological kingdom.
And the share price has obligingly followed the hysteria downwards.
But something about the whole affair has been bothering me. Not simply the media's apparent inability to distinguish an allegation from a fact, or the quite extraordinary willingness of otherwise intelligent people to abandon basic standards of evidence when presented with a sufficiently salacious story.
It's the timing.
There is a missing variable in this picture. Something that makes the otherwise rather bizarre behaviour of the Trump administration, Nvidia's board, and the various national-security officials involved fit together.
I think I may have found it.
What if the entire affair is about negotiating with China?
And what if Nvidia's apparent destruction is not an unfortunate side effect, but a calculated cost?
Start with what we know
American intelligence has reportedly known about Jensen Huang's relationship with Alice Huang since at least 2025.
Trump was reportedly briefed in late 2025, and again by Marco Rubio in May 2026, while travelling to meet Xi Jinping in China.
He did precisely nothing. Huang retained his access to the White House and continued advocating for American semiconductor exports to China.
Meanwhile, there was a longstanding disagreement within the administration about precisely those exports.
Huang's position was straightforward: sell China American chips, make Chinese businesses dependent on American technology, and preserve the dominance of the American software and hardware ecosystem.
The national-security faction took the opposite view. Every advanced American chip supplied to China was potentially strengthening a strategic competitor.
Two perfectly intelligible positions, with enormous economic and geopolitical consequences.
In spring 2026, Rubio reportedly oversaw a classified memorandum based on CIA intelligence about Alice Huang. According to the Washington Post, he recast the intelligence in more assertive language.
That is a remarkably interesting detail.
We haven't seen the original CIA assessment or Rubio's memorandum. Nor do we know who disclosed its contents to journalists. The reporting establishes that Rubio prepared and circulated the memorandum, not that he personally leaked it.
But the distinction between what the CIA assessed and what Rubio subsequently wrote could turn out to be very important indeed.
Now consider what happened next.
On 28 October, Nvidia announced that it would report quarterly earnings on 17 November.
On 29 October, the first major allegations became public.
On 3 November, Trump's Republicans suffered a midterm defeat.
On 11 November, Nvidia's board accepted Huang's resignation as president, CEO and director, effective immediately.
And on 16 November, three days before Trump was due to meet Xi in Shenzhen, Washington announced new restrictions on Chinese access to American AI computing through overseas data centres.
That last development is the one that really caught my attention.
The curious case of Bessent and Lutnick
Scott Bessent, the Treasury Secretary, has been publicly attacking Howard Lutnick, the Commerce Secretary, for failing to impose sufficiently restrictive controls on advanced semiconductor exports.
Lutnick's position has been that unnecessarily antagonising Beijing risks a wider confrontation.
And Trump?
Trump has been positively serene.
This is a man whose customary response to disagreement is not exactly characterised by studied equanimity. Yet when asked about the increasingly public dispute between his two Cabinet secretaries, he declined to condemn either.
He described them as representing two different approaches to China, both of which he understood.
Which is interesting enough.
But the new export-control rule contains an extraordinary detail.
Although formally issued by the Commerce Department's Bureau of Industry and Security, its own announcement says it was written by and at the direction of the Treasury Department and the National Security Council.
Read that again.
The department responsible for administering export controls has published a rule explicitly identifying another department and the White House security apparatus as its authors.
Bessent's faction has, apparently, got its policy.
And they've got it three days before Trump meets Xi.
Coincidence? Entirely possible.
But now consider the incentives.
Good cop, bad cop

Imagine you're preparing to negotiate with China.
You want access to rare earths and other strategic materials. China wants continued access to American technology. Both sides have things the other wants, and both have ways of making life unpleasant.
How do you improve your negotiating position?
One approach is to threaten restrictions on advanced computing unless China provides concessions.
The trouble with that is that the move is transparently coercive. Beijing knows exactly what you're doing, and might respond by escalating its own restrictions.
There is another approach.
Suppose an intelligence scandal erupts involving the CEO of America's most strategically important semiconductor company. Suddenly there are congressional investigations, national-security concerns, accusations of illicit exports, and demands for drastic action.
The administration announces tougher restrictions.
Not, ostensibly, because it's trying to extract concessions from China.
Heavens, no.
Because the scandal has revealed an urgent national-security problem that must be addressed.
Bessent plays the hawk. Lutnick warns about the consequences of confrontation. Trump, displaying an unusual degree of restraint, listens carefully to both.
And then he meets Xi.
Now Trump can say, in effect:
Look, my national-security people are demanding these restrictions. They're very serious about them. I'd like to maintain our good relationship, but I'll need something substantial if I'm going to relax them.
The negotiating position is no longer a threat to introduce new restrictions.
The restrictions already exist. Relaxing them becomes a concession.
And what might China offer in return?
Rare-earth access would be an obvious candidate.
China has already extended the suspension of certain export restrictions until January 2027, while retaining other controls. That is a useful reminder that access to strategic materials is not settled permanently.
There are any number of other possible concessions: trade, tariffs, market access, technology arrangements.
Trump returns home and announces a tremendous agreement.
A very good deal.
The best deal, probably. Nobody has ever seen such a deal.
And everyone gets to declare victory.
But why involve Nvidia's leadership?
This is where the hypothesis becomes rather more speculative, and considerably more interesting.
Huang has resigned, but Nvidia's SEC filing doesn't tell us who initiated the departure, whether he was pressured to resign, or what prompted the decision.
The board initially expressed confidence in him. Days later, he was gone.
Now, despite an extraordinary campaign by employees, shareholders and industry leaders calling for his reinstatement, the board reportedly refuses even to consider bringing him back while its investigation continues.
No detailed explanation of the departure has been published.
This is generally interpreted as evidence that the board has decided Huang's conduct was so reprehensible that Nvidia cannot possibly continue under his leadership.
But there is another possibility.
What if they've been told not to bring him back yet?
Not permanently. Not because the investigation has established some appalling crime. Simply because the political situation hasn't been resolved.
Keep the interim CEO. Continue the search for a replacement. Maintain the appearance that the leadership transition is proceeding normally.
Don't make any moves until Washington has finished negotiating.
If Huang subsequently returns, the board can describe the reinstatement as the result of a careful investigation, reflection, shareholder engagement, or whatever other corporate-governance language proves convenient.
This is not an allegation that Nvidia's directors have received such instructions. I have seen no evidence that they have.
It is a hypothesis that would explain several otherwise rather awkward features of the sequence.
And it comes with an important consequence: if the separation was always intended to be temporary, the staggering market reaction could be a cost somebody anticipated rather than a consequence nobody foresaw.
That would be an extraordinary gamble.
Which brings us to the obvious objection.
Surely nobody would deliberately do this to Nvidia?
Quite.
Why would an American administration deliberately destabilise one of the most valuable companies in the world, the leading supplier of the computing infrastructure underpinning its AI ambitions, just before an important diplomatic summit?
Why risk damaging investor confidence, disrupting an enormous financing ecosystem, and potentially compromising America's technological leadership?
It sounds insane.
And that is precisely what makes the possibility interesting.
In strategic terms, an apparent act of self-harm can be an effective signal if your opponent cannot readily imagine that you would undertake it merely as a negotiating tactic.
Of course, there are limits to this reasoning. An action being surprising does not make it clever, and economic damage does not become imaginary merely because someone hopes to reverse it.
Nvidia's share-price decline is real. Its leadership disruption is real. Financing arrangements, customer confidence and technical execution cannot necessarily be restored by issuing a cheerful press release after a summit.
For this strategy to make sense, the people undertaking it would have to believe that the eventual diplomatic gains justified the disruption, and that the damage could be contained.
Which brings us back to the earnings announcement.
The conveniently timed earnings report
Nvidia reports tomorrow, 17 November.
Management has previously indicated that demand continues to exceed its ability to supply hardware, with substantial expansion expected across successive generations.
Suppose the company produces another exceptional quarter and issues robust guidance.
That could provide a useful stabilising influence precisely when the market is questioning its future.

The administration gets its diplomatic confrontation on Thursday, while investors have fresh operating results against which to judge the company's prospects.
If the negotiations produce a breakthrough, export restrictions can potentially be relaxed and the political temperature reduced.
Perhaps the board's attitude towards Huang's return changes soon afterwards.
Perhaps the independent investigation concludes that the affair was deeply embarrassing but uncovered no material compromise of Nvidia's technology.
Perhaps the scandal simply fades as abruptly as it appeared.
I stress: none of those outcomes is established, and none would prove the events had been coordinated in advance.
But the sequence would be consistent with the mechanism.
And its timing is remarkably convenient.
So what would convince me?
The next few days give us a chance to see whether that timing mattered.
If Thursday's negotiations produce Chinese concessions on rare earths or other strategic materials, accompanied by American flexibility on advanced computing access, that would support the bargaining-leverage component.
If Nvidia's board subsequently reopens the question of Huang's reinstatement, particularly without any dramatic intervening investigative discovery, that would add another interesting piece.
If the intelligence investigation winds down shortly afterwards without finding material wrongdoing, the question of why Huang departed so abruptly would become more pressing.
Conversely, if the chip restrictions remain in place regardless of Chinese concessions, the investigation produces substantive findings explaining Huang's departure, and the board proceeds with a permanent successor, the integrated hypothesis becomes substantially less convincing.
And even a perfectly aligned sequence of favourable developments would not prove prior coordination.
For that, we'd need something much more direct: evidence about the original disclosure, communications between Washington and Nvidia, or documentary proof that someone deliberately sequenced these events.
Until then, this remains a hypothesis.
A rather intriguing one.
Cui bono?
I've watched a great many people examine this scandal as though it were fundamentally a question of whether Jensen Huang should have kept his trousers on.
That may make for excellent television. It doesn't strike me as a particularly satisfactory explanation for the broader geopolitical developments.
The questions I find interesting are rather different.
Who benefits from weakening Huang's influence over American export policy?
Who benefits from imposing additional restrictions on Chinese computing access immediately before major bilateral negotiations?
Why has Trump declined repeated invitations to publicly choose between his feuding Cabinet secretaries?
Why has the administration suddenly found a way to impose restrictions that were previously stalled?
And why has Nvidia's leadership crisis unfolded in such close proximity to both its earnings announcement and the summit?
None of those questions proves anything by itself.
Together, however, they form a pattern worth examining.
My suspicion is that the Nvidia scandal has become a bargaining instrument in a much larger confrontation over technology, rare earths and the terms of America's relationship with China.
Whether that was planned from the outset or opportunistically exploited remains the critical uncertainty.
The distinction matters. A skilful exploitation of unexpected circumstances is very different from deliberately engineering an extraordinary corporate crisis.
But if the full, coordinated version turns out to be correct, it would be an astonishing piece of realpolitik: publicly destabilising America's most important AI supplier, using the resulting outrage to justify restrictions, extracting concessions from Beijing, and then allowing the crisis to resolve.
All while maintaining the appearance that each event emerged organically.
And if the crisis was not coordinated, it is still worth understanding how an affair, an intelligence assessment, a corporate resignation and an interdepartmental policy battle have converged at such a consequential diplomatic moment.
I have no intention of mistaking a pleasing pattern for proof. That's precisely the mistake I've spent the past fortnight watching other people make about Huang.
But equally, I see no reason to ignore a hypothesis simply because it suggests that the public explanation may not be the whole explanation.
Thursday should be interesting.