I went out for tacos with my wife a few days ago and as I updated her on the latest gorpmans and bleemers of the Jensen Huang scandal, a certain neutral look in her eyes led me to suspect that I had once again fallen victim to the Geochemist’s Folly, always a danger to those of us selected at birth by the capering ‘tism goblin. I had implicitly assumed that she was walking around with a full background briefing on yet another of my baffling special interests, but to her credit my wife is a normal person and the current economic state and geopolitical importance of the G.P.U. market, Nvidia’s place within that market, and ex-C.E.O. Jensen Huang’s role in both Nvidia and the current e. s. & g. i. of same are not perpetually top of mind for her.
I thought that if an intelligent and well-informed but fundamentally normal person such as my wife wasn’t entirely clear on why we all care so much about who some computer executive is banging, perhaps many of you find yourselves in the same position. So in the hopes of once again turning my lifelong burden into your benefit, I give you:
The Normal Person’s Guide to the Jensen Huang Scandal
Who is Jensen Huang?
Jensen Huang is the centi-billionaire co-founder, C.E.O., and until very recently unchallenged cult leader of G.P.U. company Nvidia. Like all these tech guys he has selected a uniform so his appearance can be properly branded, and his uniform is a leather jacket. He looks like this:
That’s a still from a video of Huang signing a woman’s boob at the 2024 Computex trade show in Taiwan, which even at the time garnered a bit of a 🤨 from the tech press, but in hindsight was a real “do you think someone like this could get honeytrapped?” moment.
Ok but what’s his, like, deal?
Jensen Huang was born in Taiwan but sent by his parents to live in America at age 9. He “became a nationally ranked table-tennis player” apparently, among other colorful and probably traumatizing childhood experiences, but eventually he got an undergraduate electrical engineering degree and went to work for chip company AMD while he got his Master’s in electrical engineering at Stanford. As we all know, nothing good ever came from Stanford but unusually for rich tech guys, Huang is at least an actual engineer who knew how to design chips at some point.
In 1993, Huang and two co-workers founded a new company to make graphics hardware for personal computers and named it Nvidia, in a reference to the Latin word for envy. Nvidia was reasonably successful selling graphics cards to gamers for 20 years or so, and then crypto happened, and then A.I. happened, and then Nvidia suddenly became the literal most valuable company that has ever existed.
This might be the first moment where a normal person, such as yourself perhaps, may think “wait, the literal who that has ever what now?” So let’s first look at what a G.P.U. is.
Do we have to?
Yes.
Fine. So what’s a G.P.U. then?
G.P.U. stands for graphics processing unit. When you’re using your computer, it has to continuously figure out what color and brightness each pixel of your screen should be. Most of the pixels of your regular desktop screen don’t change most of the time, even if you really feel jacked in to the Matrix while you’re updating a spreadsheet or writing emails. But a graphics intensive game has to calculate a whole screenful of 3D geometry, and update all of it sixty or more times per second. It’s a lot of math, and your C.P.U. is very fast at doing math but it’s also tuned for the kind of math where everything has to be done sequentially—where operation 1 feeds into operation 2, which feeds into operation 3, and so on.
Figuring out what color each pixel on the screen should be, on the other hand, is a kind of problem that computer science whimsically calls “embarrassingly parallel.” Each pixel’s value can be calculated with no reference at all to any other pixel on the same screen, which means that in principle, you can calculate them all at the same time if you have enough dedicated processors to do it. So when games started to overwhelm the one-at-a-time capabilities of the C.P.U., companies like Nvidia offered a way to offload all those tedious but independent calculations onto a bank of individually slower and less capable but much more numerous chips which were specifically designed to do exactly the kind of vector calculations that 3D graphics demanded.
Here’s a fun visual demonstration of this difference, from Mythbusters’ Adam Savage and Jamie Hyneman at, appropriately enough, a 2008 Nvidia developers conference:
Your high end modern C.P.U. has maybe 8 or 12 cores, and can do approximately that many calculations at the same time. The best G.P.U.s right now have more than 21,000 cores—each only good for a limited set of operations, but all able to work at the same time.
That’s neat, but it doesn’t really scream “most valuable company ever to exist?”
No, I hear you. We’re getting there, trust me.
In 2008, Satoshi Nakamoto invented Bitcoin—a form of money that’s only good for doing crime—and for a little while between 2017 and 2022 much of the world’s excess wealth was tied up in guessing numbers. Because the way you make cryptocurrency is to take a bunch of transactions you want to add to the blockchain, prepend a random number to them, hash the whole thing, and see if your hash has an arbitrary property like “ends in five zeros” or whatever. If it does, you just generated a valid block, and you get to submit it to the global ledger and reap your reward in the form of a brand new number that no one else is allowed to own. Usually it doesn’t, and you pick a new random value to prepend and try again.
The more systems-minded among you will have noticed that this calculation doesn’t depend on anything else. Every iteration starts with the same set of transactions, and we’re not searching sequentially for the magic number that makes a block valid, we’re just guessing and checking. So this is another embarrassingly parallel problem, and crypto miners rapidly figured out how to shunt the actual guessing and checking off to G.P.U.s, which could crank out simple hash calculations thousands at a time. Crypto miners built out great energy-sucking data centers full of heaps of Nvidia hardware to guess numbers at a dizzying clip, and Nvidia briefly became what we thought at the time was ludicrously valuable, with a peak market cap close to $800 billion in early December, 2021.
Then Sam Bankman-Fried got pinched, everyone suddenly realized that it had been almost fifteen years and no one had yet come up with a non-crime use for cryptocurrency that wasn’t “gambling on cryptocurrency,” and Nvidia’s value gradually drifted downward until October, 2022.
And then A.I. happened.
Ugh.
I know. But I’m just telling you what is, not what should be. And frankly I don’t know enough about how A.I. model training works to explain it, so I’m not going to try. But what matters to us is that training Claude to write mediocre code involves an ungodly number of the same style of parallel calculations that first graphics and then crypto did, and Nvidia has also designed special hardware even more tuned to the specific needs of A.I. training. For the last few years the whole U.S. economy has been three A.I. data centers in a trench coat. Nvidia’s A.I. business is like selling shovels in a gold rush, if each shovel were guaranteed to dig up a set amount of gold per millisecond.
In summary, by October of this year—just one short but eventful month ago—Nvidia’s market cap peaked at more than five and a half trillion imperial United States wing-wangs.
And that was… a lot?
Yes.
Compared to, like, Apple?
Yes, half a trillion dollars bigger.
Alphabet? Meta? Exxon?
Look, I don’t keep a running tally of the biggest companies in the world in my head either, so I was as surprised as you might be to learn that Nvidia was bigger than all of them. Alphabet (Google) was like $4 trillion. Meta was less than $2 trillion. Exxon Mobil isn’t even in the top 20. This was the list as of October 1st, 2026.
Wow that’s crazy.
Yeah. So that’s the first plank of “why do we care what some computer guy did?” He was, at the time, the unquestioned head of the literal biggest corporation in human history. The second plank is: this isn’t a sex scandal, it‘s a trade scandal that was revealed by the sex scandal.
The U.S. banned A.I. chip exports to China. But China is a huge current and presumably future market for A.I. chips, and what do you think happened when the Trump administration tried to block the movement of these extremely valuable, extremely portable commodities? As Ed Conway wrote in Material World:
…we’re looking at flows of computer servers from Malaysia to China. And look: up until late 2023 there were very few of them crossing the border. And no surprise, since China could just buy those chips from the US. But then, after the ban, there’s a sudden rise in Chinese imports of servers from Malaysia. All the way until Malaysia puts its own restrictions on transshipments of American chips to China.
Nvidia is an extremely valuable company partly because the business world currently believes the entire economy might be nothing but an A.I. gold rush ever again, and everyone would like to have a stake in the world’s main shovel maker. But also because, theoretically, Nvidia isn’t even allowed to sell its products in one of the biggest markets—if not potentially the biggest market—for them: China.
But what if that potential market is already pretty much saturated, via a combination of deals like the one Jensen Huang and his paramour Alice “No Relation” Huang hammered out in between bouts of leather-jacket-off time, and deals where Chinese A.I. labs rent compute time from data centers in third countries where Nvidia is allowed to sell its chips?
As it happens, Financial Times’ Owen Walker and Antoine Gara just reported that:
Nvidia stopped disclosing the scale of its business in Singapore last year, at a time the city-state accounted for almost half of its worldwide revenues.
Sorry I forgot I was your audience stand-in for a minute there, but did they just say half of its worldwide revenues came from Singapore? 👀 👀 👀 👀
Yes. Now that’s not to say that half of Ndivia’s worldwide revenue is directly from illicit or semi-licit Chinese sales, but it is a little weird that (a) HALF of the company’s revenue was being booked through its Singapore subsidiary, and (b) it decided to stop reporting this without really saying why. But it was “a 72 per cent increase in 12 months” of 2025. Weird!
Weird!
Weird.
So where does that leave us? Because this is getting long for email and I have to… go. Do other things.
Right, of course, I think that’s most of it. Just one more point.
Yesterday, about the Anthropic I.P.O., Matt Levine wrote:
The market is largely the AI trade... If you are an investor, you have to have some mental model of the AI trade, because the AI trade is virtually all there is.
Anthropic is a public company now, and allegedly worth a lot of money, but Anthropic is also a software company. Anyone who has paid any attention at all to the world of software in the last 30 years has watched as expensive proprietary software has been copied, undercut, and ultimately replaced by open source or free software in nearly every market. Open-weight models already exist—free, pre-trained A.I. models you can download and run on your own hardware, for your own purposes—in fact many of them are Chinese in origin. And a huge part of the value proposition of A.I. software is that it is very good at rapidly copying and replacing existing software. There’s no obvious reason to assume that what has been true of software generally will suddenly not be true of A.I. models. Which means that if the whole economy is A.I., and A.I. software tends to a value of zero over time, then the whole economy is specifically A.I. hardware.
And if that’s your mental model of the A.I. trade, maybe you’re not that worried whether Nvidia is valued as if it can’t sell to 50% of its total market when it is in fact already selling to 100% of its total market. But maybe… you are a little bit worried?
Huh.
Anyhoo, since early October Nvidia has lost more than two and a half trillion dollars in market cap. Nvidia’s Board of Directors has announced a “National Security and Export Control Oversight Committee,” which I’m sure they were just about to do anyway before all this went down. The Securities and Exchange Commission is warming up its probes. Leaks suddenly abound about Huang’s abusive managerial style and the questionable roles his children play as executives of the company. The Trump administration is investigating what they definitely already knew was going on more than a year ago and the scandal is pouring more blood in the MAGA shark tank. Nvidia’s weirdest dorks, the “Branch Nvidians” as Liz Lopatto reported they call themselves, are begging for Huang to return, as are the flock of dipshits and losers running Silicon Valley. Nvidia’s board has done its best to shut that nonsense down cold, presumably because they are too busy trying to figure out whether they’re eventually going to jail along with Huang or not.
And finally, all the way back in the quaint world of video games, Grand Theft Auto 6 came out, and apparently it’s real good.
Please buy me one (1) mediocre dinner per year in exchange for this. That’s my model of the economy.


![A screenshot from companiesmarketcap.com of Nvidia’s market cap from 1999 to the present. It’s a flat line until 2020, then a small peak at almost $1 trillion with a label reading “Crypto,” then a much, much higher peak at nearly the edge of the chart, close to $6 trillion, labeled “A.I. lmao [melting face emoji].”](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,quality=80,format=auto,onerror=redirect/uploads/asset/file/8142a2fb-410d-473e-b471-9dce18e5cda2/nvidia_chart.png?t=1795204811)

![Megane Top Gun posted “Yeah they still got it tbh” above a screenshot of the in-game GTA6 instagram with a post that reads “yeah I like ASS—[acrostic] ASS-S-S” and a caption that says “the booty hath got me acting diabolic”](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,quality=80,format=auto,onerror=redirect/uploads/asset/file/e8d34977-d277-4484-8102-7b405dd628e8/image.png?t=1795210816)
