Here’s one way to think about things. In about five years, people and companies will be spending about $4 trillion a year on artificial intelligence.[1] Give or take! Perhaps it will be $3 trillion, or $5 trillion. Perhaps it will take six years to get to that level, or four. Obviously there are weirder possibilities: Maybe in two years AI will cause the extinction of humanity, or at least render money obsolete; maybe we’ll all decide that AI is bad and stop building it, and in five years the annual spending on AI will be $0. But the center of the distribution might be something like “in five years, $4 trillion a year on AI, give or take.”
And then on crude simple assumptions a cash flow of $4 trillion a year is worth something like $40 trillion in present-value terms,[2] which is on the order of half of the value of the US stock market today.[3]
This is just a big deal relative to, you know, whatever else you’ve got going on, financially. In some sense the basic job of an investor in 2026 is:
You need a model of what the future cash flows of AI will actually be, and when they will arrive.
You need a model of who will get them: Okay, companies will spend $4 trillion on AI, but where will that money end up? Will a lot of it be captured as profits at frontier AI labs? As profits at application developers downstream of the AI labs? By chipmakers? By electric utilities? By the neoclouds and other data-center operators that sit between the chipmakers and AI labs? At each level, will the companies make enough to pay off the debt that they take on to build all this AI capacity? Will they make enough to make their shareholders rich? There is some complicated waterfall of cash flows from all of this AI money, one that is partially baked into the financing documents for the AI data center projects but that is largely based on your assumptions about industry structure, capacity, demand, bargaining power, etc., in the future. And this waterfall will be full of nonlinearities: There’s some neocloud that goes bankrupt and defaults on its bonds in the state of the world where AI generates $4 trillion of revenue, but that pays off those bonds and returns 1,000% for its shareholders in the state of the world where AI generates $4.1 trillion.
You need a model of everyone else’s model: If you think “$4 trillion of cash flow in five years,” is that optimistic or pessimistic relative to market consensus? (If the market thinks “$5 trillion in four years,” and your model is right, then AI stocks will go down.)
You need to adjust your model for news. If last week’s election results make data center moratoriums more likely, will that change the magnitude of future AI cash flows? Will it push back the timing of those cash flows? Will it mean that a neocloud with planned data centers in North Dakota will capture more value, while a hyperscaler with planned data centers in Texas will capture less? Etc. Ideally your model has some sliders for how big the cash flows will be and when they will arrive, and when news happens you can slide the sliders, and they will tell you whether your favorite neocloud will go bankrupt or whatever.
A few points about this. First: It’s a hard job. You could make it easier by having a very extreme model: “AI is just hype and none of these cash flows will happen” would make the rest of your decision process easier, as would “AI will cure all human problems and usher in an era of perfect abundance and harmony in two weeks,” or “AI will kill us all in two weeks” for that matter.
But most investors seem to be operating on the theory that big tech companies are going to plow trillions of dollars of capital expenditure into building AI capacity as quickly as possible, and that this capex will be more or less justified by future AI revenues, and that there will be some distribution of winners and losers and people who boringly earn their cost of capital from those future revenues. But there is a range of disagreement over the details, over the timing and magnitude and winners and losers.
Second: There’s nothing new or unusual here. This is the basic job of any investor any time, trying to figure out the future cash flows of companies and compare your estimates to the market consensus.[4] It’s just that, right now, a lot of the market-implied future cash flows are from AI, and those flows are relatively uncertain and speculative and have a somewhat science-fiction-y flavor to them. “We’ll be spending $4 trillion a year on the robots and cancer cures we haven’t invented yet”: Not a bad thesis, just one with a lot of variance.
A lot of variance. There is a wide range of assumptions you could make about the future of AI, and your model is going to be very sensitive to those assumptions. Today’s financial markets are very levered to very uncertain stories about the magnitude and timing and shape of future AI cash flows. If the consensus story changes — from, say, “hyperscalers and AI labs will work as fast as they can to build as much data-center capacity as possible” to “actually data centers will be built much more slowly because of political opposition” — then that has huge nonlinear implications for which stocks will go up and which data-center bonds won’t get paid back and who will win or lose from AI.
And so, if Jensen Huang has an affair with an alleged Chinese spy, could that affair reduce the present value of Nvidia Corp.’s future profits by $400 billion? Easily! Bloomberg’s Ian King reports:
Nvidia Corp.’s investigation into Chief Executive Officer Jensen Huang’s affair with an alleged Chinese spy raises a series of once-unthinkable questions, including whether the man at the center of the AI boom is at risk of losing his job.
Nvidia shares slid as much as 8.8% on Monday after Huang disclosed the romantic relationship in a memo to employees, at one point erasing almost $400 billion in market valuation. The company’s board is assembling a committee of independent directors who will work with outside counsel to investigate the matter.
With the stunning revelation, investors need to consider a range of possibilities that could affect not just Huang and Nvidia, but the entire artificial intelligence industry. Nvidia, the leading maker of AI chips, has used its financial strength to back dozens of major players, from frontier labs like OpenAI to cloud computing players like CoreWeave Inc. Huang has also been among the most forceful advocates for the benefits of AI, from Washington to Wall Street. ...
The latest revelations “highlight the considerable key man risk that Nvidia faces,” [Bernstein analyst Stacy] Rasgon said. “Nvidia is not an ordinary company and Jensen is not an ordinary CEO; he IS Nvidia, clearly sets the long-term strategy and direction, and (with supposedly 60 direct reports) pulls most of the company’s strings himself.”
All sorts of stories. Huang seems to be essential to Nvidia’s ability to capture a lot of AI revenue for itself; maybe the board will actually fire him even though he is irreplaceable, and Nvidia’s performance will suffer.[5] Or some lesser story. A portion of Nvidia’s current and future AI revenue comes from selling chips to Chinese companies, and Nvidia has pushed to be able to do more of those sales. The alleged spy was not just having an affair with Huang; she was also buying advanced Nvidia chips for a Singaporean neocloud that has been investigated for allegedly passing them to Chinese companies. Does this mess make it less likely that Nvidia will be allowed to sell chips to China in the future, or that the US will otherwise tighten export controls on chips in ways that lower Nvidia’s revenue?
Or, more broadly, Huang is one of the more prominent voices against “pacing the frontier”: Does he lose a bit of prominence and credibility from this mess? Does that make it more likely that AI development will be delayed? Maybe not by a ton: There is no crisp causal connection between “Jensen Huang has an affair with a spy” and “big US AI labs get an antitrust waiver to cooperate on slowing the pace of frontier model development” and “Nvidia sells fewer chips.” But a 1% increase in the probability of that happening really is worth tens of billions of dollars.
I guess my point here is that Jensen Huang’s affair with an alleged Chinese spy is plausibly the most expensive extramarital affair in human history? And it kind of should be?
Meanwhile: last week
Just to continue, for a minute, with this idea of a model of the timing and magnitude of the AI cash flows, and of who will get them. One way to think about Situational Awareness, Leopold Aschenbrenner’s AI-focused hedge fund, is that:
It was early to the business of building this sort of model;
It was unusually sophisticated in building this model: Aschenbrenner had a better picture than much of the market of exactly who would win and when;
It pushed all the sliders way way way toward “bigger and faster”: It made a huge bet that the AI boom would be bigger and happen more quickly than the market expected, so it got long all the companies with the most leverage to that thesis (and short all the companies that would lose out);
This was correct for a while, and made Situational Awareness huge piles of money;
And then the market concluded “actually it’s gonna be a bit slower and smaller than we thought last week” and Situational Awareness got smoked.
Do you even remember why the market concluded that, back when Situational Awareness got smoked in July? It’s not important. The point is not that Situational Awareness got the path of AI adoption wrong, permanently, and its thesis has now been disproven by actual events. The point is that Situational Awareness made bets on the path of AI adoption, and the market converged on its bets and gave it money, and then the market diverged from its bets and it lost money.
Back when Situational Awareness had its oopsie, I suggested that it was in many ways an unusual hedge fund. Most big multistrategy hedge funds are less volatile and more focused on risk, and they prefer to have lots of uncorrelated bets rather than one giant levered omnibet on the speed of AI adoption.
At some level, though, every hedge fund is kind of in this same business. Every tech equity long/short portfolio manager, and kind of every other hedge fund manager, has some model like this, some sense of the timing and magnitude of AI cash flows and the waterfall of who will get them. If you are a conservative long-term credit investor, you might have a model like this, and use it to find the most conservative possible AI investments, the data-center bonds that will get paid back even if the AI boom is slower and smaller than the market expects.
But if you are a tech equity long/short manager, your business is somewhat different. Your business is to know the market consensus, to have your own hopefully better-informed view of reality, and to make high-leverage bets that will pay off when the market converges on your view of reality. If you think that the market expects $X billion of hyperscaler capex announcements, but your own research suggests that the actual announcements will be $Y (Y > X), then you should buy whatever companies have the highest leverage to AI capex announcements, and short whatever stocks will go down the most on that news.
And you’ll make short-to-medium-term bets on whatever you think will go up the most on news, and you’ll do it with a fair amount of borrowed money, and the result is that you will be the most levered part of the whole gigantic AI capital structure at the center of the modern economy. And then if there’s bad news — if the market consensus is that AI cash flows will be a little lower and take longer to arrive — then, like, some data-center bonds will go down a tiny bit, and some hyperscaler stocks will go down a bit more, and some neocloud stocks will go down even more, and you will blow up and lose your job. You are living at the tip of the spear, and if you’re wrong you’ll know it before everyone else does.
Anyway, even before Jensen Huang’s own oopsie, last week was rough for hedge fund managers. “Engineers Gate, a systematic hedge fund, has cut the majority of a team that employed fundamental managers after a bruising October,” Bloomberg reported. And: “Walleye Capital plans to return about $2 billion to clients following its decision to cut back on equities trading.” And: “At least five portfolio managers focused on credit trading at Brevan Howard Asset Management have left the hedge fund firm, the latest sign of fallout from market turmoil.” The Financial Times reported on “How Wall Street’s all-in bet on AI blew up.” The Wall Street Journal reported that “Hedge Funds Dump Positions and Cut Staff as Selloff Deepens.” Bloomberg described the “Anatomy of a Selloff: Sudden Chain Reaction Rattles Wall Street.” And Bloomberg reports this morning that “hedge funds dumped technology stocks at one of the fastest paces on record last week, and the momentum crash that followed was the biggest in about 40 years.” Byrne Hobart writes that the market moves get “exacerbated further when levered investors in various contexts are drivers of both the main market-moving theme and, by extension, GDP-level fundamentals.” Bloomberg’s Tracy Alloway writes: “The financing and trading structures aren’t separate from the AI story — they help shape it. On the way up, leveraged products add exposure, gains attract more capital, and rising stock prices make buildouts with enormous price tags look more realistic.”
The economy is levered to the pace of the AI buildout, but some people are more levered to it than others, and when expectations change they get blown up.
But is it securities fraud?
I don’t make the rules! But:
Jensen Huang had an affair with an alleged Chinese spy.
Nvidia did not go around telling investors “hey FYI our CEO is having an affair with a Chinese spy.”
Actually the opposite: “Just in the past two weeks, despite reports of a close relationship between the two, Jensen defended himself, calling their interactions ‘completely normal.’”
When the news came out, it wiped hundreds of billions of dollars off of Nvidia’s market capitalization.
LOS ANGELES—(BUSINESS WIRE)—Schall, Brown & Schwartz LLP (“SBS”), a national shareholder rights litigation firm, announces that it is investigating claims on behalf of investors of NVIDIA Corporation (“Nvidia” or “the Company”) (NASDAQ: NVDA) for violations of the securities laws.
INVESTIGATION DETAILS: The investigation focuses on whether the Company issued false and/or misleading statements and/or failed to disclose information pertinent to investors. Nvidia is the subject of a New York Times report published on November 8, 2026, titled: “Nvidia Opens Investigation Into Jensen Huang’s Relationship With Suspected Chinese Spy.” According to the report, “Nvidia said on Sunday that it was opening an investigation into Jensen Huang, its chief executive and co-founder, after he acknowledged to the company’s board that he had been intimate with a business executive who U.S. intelligence determined was a Chinese spy.” Based on this news, shares of Nvidia fell nearly 5.5% in pre-market trading on November 9, 2026.
If a CEO has an affair that knocks hundreds of billions of dollars off the stock price, shareholder plaintiffs’ lawyers are going to demand a cut.
I’m having an affair with the CEO of a listed company and I load up on long-dated puts on his company. Would that be considered insider trading … ?
The information is obviously non-public, but how do you assess materiality here? I don’t think it necessarily has to be material. In most cases it probably wouldn’t be.
It’s definitely material if you’re a Chinese spy.
Goldman standup guy
Ioftenwritearoundherethat “what if people could sell stock in themselves” is “the great late-night dorm-room question of financial capitalism,” but that doesn’t mean it’s funny. Like, the first time it occurs to you, late at night in your dorm room, you might get a chuckle out of it. “Hey I just had a crazy idea, what if people could sell stock in themselves,” you say to your roommate, and your roommate chuckles too, because you are freshmen at the university of financial capitalism and you don’t know any better. But then you get older and wiser and you see, like, the 100th company pitching “what if people could sell stock in themselves,” and you react with a weary shrug. “Oh, right, that,” you say.
Still I cannot deny that, like, 70% of people with “what if people could sell stock in themselves” business ideas are basically serious and profit-seeking, but the other 30% are doing it as comedy. The first time it occurs to you, you might find it funny, and every day it occurs to hundreds of people for the first time, and some of them go and do it as a bit.
If a guy comes to you and is like “hello, I’m a former Goldman Sachs investment banker looking to get into standup comedy, and I’m raising capital to do standup by selling shares in my future standup earnings, would you like to buy some,” is that a good investment? Is it a good joke? I have my views, but I do not give investing or comedy advice around here. But here’s this guy:
The 57-year-old North Carolinian had tried his hand as a Goldman Sachs investment banker, Sports Illustrated columnist and startup co-founder. But a few years ago, on Christmas Day, he had a bit of an epiphany. “I was on a run, and I realized I wanted to take one more big swing before I got too old,” he tells Variety. He decided it was time to pursue a career in stand-up comedy. Pete McEntegart was no more; from that day forward, he’d be known as Pete Smack. ...
Smack decided he would raise money for his comedic venture and hold himself accountable to investors.
He turned to the startup crowdfunding service Wefunder and became the first entertainer to solicit investors on the platform. By early last year, he raised nearly $60,000 from 104 investors, the majority of whom committed between $100 and $150. Buying into Smack is a legitimate investment regulated by the SEC. To set up his seed round, so to speak, the comedian spent $5,000 on lawyers.
Smack pitched prospective investors in a humorous presentation posted to his Wefunder page. “You can’t have a truly diversified portfolio if you’re not directly invested in a standup comedian,” he began.
I bet that joke killed on his desk at Goldman!
Smack offered family, friends and strangers on the internet a chance to split 10% of his gross revenues through 2029. On Jan. 1, 2030, their investment converts to equity at a $1 million pre-money valuation. “While I can’t guarantee what your financial return will be — the lawyers were pretty adamant about that,” Smack wrote on his Wefunder page, “I’m confident you’ll have a fun and entertainingly documented ride.”
See, what would actually be funny is buying one share and then suing him for securities fraud.
You received this message because you are subscribed to Bloomberg’s Money Stuff newsletter. If a friend forwarded you this message, sign up here to get it in your inbox.
pacing the frontier
