Universal basic AI stocks |
One view of the artificial intelligence boom is that AI will replace all of human economic endeavor. Robots will make the stuff, and robots will manage the factories that make the stuff, and robots will set corporate strategy about what stuff to make, and robots will run the hedge funds that allocate capital to the companies whose robots make the stuff, etc., just robots all the way down. All of the corporate profits that used to go to companies that did stuff — manufacturing, retailing, consulting, whatever — will in the future go to the meta-companies that own the robots, and all of the labor income that used to go to people who had jobs will in the future also go to the robot companies, because the robots will take all the jobs. I’m sort of kidding but also sort of not. This is, for one thing, the way the leaders of the big AI labs talk, possibly because they believe it. The reason that the big AI labs have gone from nothing to trillion-dollar valuations in the blink of an eye is that investors believe some toned-down version of this story. They believe it about the AI labs (thus the valuations) and also about the victims of the AI labs; thus the frequent occasions earlier this year when an AI lab would announce some product that might disrupt some industry, and the stocks of all the companies in that industry would drop. SpaceX says that the total addressable market for AI “enterprise applications” is $22.7 trillion, perhaps 20% of the world’s economic output and perhaps 40% of the revenue of the world’s corporations. That is not quite “we will replace all human economic endeavor,” but it is closer than any non-AI company has ever gotten. Another view is “lol this is all dumb, of course the people leading AI labs talk like this, they need to raise umpteen bazillion dollars to build data centers and pay themselves, and the AI-will-take-over-everything thesis is vastly overstated.” But let us just suspend disbelief for a moment and assume that the maximalist view is correct and soon everything will be AI. What … uh … what should one do about that? There was a brief moment when a semi-plausible answer might have been “have faith in our benevolent AI overlord, Sam Altman”: For a while, the clear leader in AI was OpenAI, which was run as a nonprofit organization with the goal of producing AI for the benefit of humanity, not for profit. If that had, you know, worked out, then AI would take all the jobs and businesses and produce all the goods and services, but maybe it would give away all the goods and services to whoever needed them, a socialist paradise overseen by Altman and his nonprofit board of trustees and his helpful robots. But more recently OpenAI (1) has a lot more competition for the lead in building AI and (2) has become rapaciously capitalist, so never mind. The more self-help answer is that you had better own some robots yourself. The ideal way to do that is to start an AI company right this minute, if that’s at all plausible, but for most of us it isn’t. Failing that, maybe:
- Sell all your investments and put the money in a portfolio of the big AI companies. This takes care of the risk that the AI companies will render all of the other companies worthless.
- That’s still not enough, though! AI will also render your human capital worthless; you need to offset your loss of labor income.
- Maybe borrow a bunch more money to invest in AI stocks?
I emphasize that this is a goofy hypothetical and not at all investment advice. But I think it is the logical consequence of believing the AI maximalist thesis? In the current world, humans consume goods and services and pay for them with some combination of labor income and rents from capital. In the AI maximalist future world, you will still want goods and services, but you will have to pay for them with rents from the capital you have invested in AI stocks, because AI stocks will be the only source of value. Again, the counterargument is: “This is all dumb, and the AI people only talk like this to scare us into giving them money.” I can see both sides. I feel like a lot of stories in financial markets these days are about essentially this tension, “everyone must prepare for apocalypse by owning AI stocks” vs. “you are trying to trick us into owning AI stocks.” The most literal version might be this story from Notus last week:
Senior U.S. officials have held preliminary discussions with major artificial intelligence companies about the potential for the federal government to acquire some shares in their firms, according to three people familiar with the matter.
Sam Altman, the CEO of OpenAI, has discussed the idea with senior Trump administration officials periodically since the president began his second term, said two of the sources, all of whom spoke on the condition of anonymity to reflect private deliberations. Altman first pitched the concept directly to President Donald Trump in a conversation in early 2025, and has discussed it again with senior administration officials in recent weeks as a way to more broadly distribute the economic benefits of AI to the public, they said.
While planning is ongoing and details are in flux, discussions have centered on having the firms voluntarily cede the shares to the government, the people said. The returns on the investment could then be directed to public purposes, one of the people said, such as distributing a dividend payment to all American households. …
This instinct is bipartisan: Sen. Bernie Sanders called this week for the U.S. government to acquire 50 percent equity stakes in the AI companies, effectively bringing the companies under federal control. The Vermont independent’s forthcoming bill also would tax at 50% the stock of OpenAI, Anthropic, xAI and other AI firms, with the proceeds being placed in a sovereign wealth fund for the public.
Arguably what Altman, Trump and Sanders agree on is something like “AI will displace a lot of human economic activity so we’d better make sure everyone gets a piece of it.” They share, broadly, the AI optimist view. Meanwhile a surprising proponent of the AI pessimist view is S&P Dow Jones Indices. We have talked a lot in recent weeks about how index providers are falling over themselves to add the big upcoming AI initial public offerings — SpaceX, OpenAI and Anthropic — to their indexes. The most important index is the S&P 500, which normally waits to add companies until they (1) have been public for at least a year and (2) are profitable. S&P published a consultation about waiving some of those requirements so that the big AI companies could join the index in six months and before they are profitable. It was widely assumed — certainly by me — that this is what S&P’s customers wanted, and for fairly obvious reasons:
- Index investors want to own all the big companies, and SpaceX, OpenAI and Anthropic will certainly be big companies, so index investors want to own them.
- If you give any credit to the AI maximalist view, then you might worry about the rest of the S&P 500. Like all those widget companies and financial services companies and so forth are gonna get eaten by the robots, if the maximalist view is right. Excluding the big AI companies from the S&P 500 makes that index, to some extent, a bet against the AI maximalist thesis. (Not entirely — Alphabet and Nvidia are in the index — but to a meaningful extent.) If you are a passive investor, your goal is probably not to make a big high-stakes bet against a popular thesis of the current stock market. You just want the market. Excluding the AI portion of the market is maybe a stronger bet than you want to make.
I wrote once that a natural hedge to the AI maximalist thesis was to make sure that everyone owned a lot of index funds. “Universal basic index fund ownership,” I called it: “You can buy index funds that include all of the big US public companies; probably some of them will capture the profits from AI.” If the index funds mostly exclude the big AI companies, that won’t quite work. On the other hand, a lot of people were really mad at the idea of S&P waiving its rules for the big AI IPOs. That was partly a matter of fairness, the look on Elon Musk’s face, etc., but it was also partly a matter of valuations. Some people just think the AI maximalist thesis is silly and don’t want 10%, or 100%, or 500% of their money to be allocated to AI companies at their current huge valuations, and they don’t want index providers to force their hands. I guess the skeptics won that one. Last Thursday, Bloomberg’s Isabelle Lee and Neil Campling reported:
S&P Dow Jones Indices will keep its existing eligibility requirements for benchmarks including the S&P 500, closing the door to fast entry for big tech IPOs like SpaceX and delaying billions of dollars in flows from passive funds.
The index provider in a press release Thursday said it will not shorten the 12-month seasoning period for newly public companies it currently has or waive existing profitability and public-float requirements based on a company’s size, diverging from a broader industry shift embraced by rivals Nasdaq Inc. and FTSE Russell. …
“I am genuinely surprised,” said James Seyffart, ETF analyst at Bloomberg Intelligence. “But S&P is the market leader and they can buck the trend.”
And on Friday, Lu Wang, Justina Lee and Isabelle Lee added:
“It’s a calculated gamble,” said Campbell Harvey, a finance professor at Duke University and partner at Research Affiliates. “If they’re in late and their investors do not participate in the explosive upside that happens after June 12th, that’s the downside. And the upside is they’re late because SpaceX goes down like 50% over the next 12 months. So it is an active decision.” ..
[It will] leave investors confronting a dilemma passive investing was designed to avoid: buy the stocks and risk chasing hype at stretched valuations, or stick with the benchmark and risk missing some of the market’s most consequential new companies.
It could be years before the big AI companies are profitable enough to get into the S&P. And then a few years after that, maybe they’ll take over the world. In a sense, S&P is betting they won’t.
Public markets are the new private markets |
That said, here is a Substack post from finance professor Baolian Wang noting that 12% of the profits of the S&P 500 last quarter came from marking up shares in private startups:
In Q1 2026 alone, just three companies—Alphabet, Amazon, and Nvidia—reported a staggering $69.2 billion in non-operating windfall under their Other Income and Expenses (OI&E) lines. When you run the macro numbers, this single accounting phenomenon artificially inflated the entire S&P 500’s quarterly earnings by about 12%. …
Under GAAP rules, companies are required to measure their equity investments in both private and public companies at “fair value” at the end of every single quarter. … If a startup they invested in raises a new private funding round at a higher valuation, the corporate backer must mark up the value of their shares on their own balance sheet. … Crucially, that imaginary, unrealized “paper profit” must be funneled directly through the income statement under the OI&E line.
Because we are currently riding an unprecedented wave of capital concentration into private AI infrastructure, late-stage tech startups, and secondary markets, these private valuations have gone completely parabolic.
The $69.2 billion of OI&E isn’t entirely from marking up startups, but that’s plausibly most of it. [1] One intuition here is that Alphabet’s, Amazon’s and Nvidia’s first-quarter earnings include (1) their actual earnings for the quarter plus (2) the increase in the total capitalized future earnings, for the rest of time, for their startup investments. In an AI boom, of course the total future earnings of their AI startup portfolios would be worth a lot relative to their actual earnings this quarter. Anyway, people constantly go around complaining that ordinary investors can’t get access to the hot private startups that are creating all the wealth, but that’s not true, is it? Some big public tech companies own stakes in some of those startups. You can buy an S&P 500 index fund, and then 12% of your profits will come from the increasing values of hot private startups. Even before those startups get into the S&P themselves.
The way stock investing works in the US is that you buy a stock for $100, and then you hold it for 10 years, and then you sell it for $300 and have $200 of capital gains, which you pay taxes on. If the stock is at $110 after the first year, you have a $10 gain, but you don’t get a tax bill on that $10: The gain is unrealized, and it’s not taxable until you realize it by selling the stock. You don’t pay taxes until you convert the stock into cash. People sometimes find this annoying, when they think about rich tech founders, and there are sometimes proposals to tax unrealized capital gains. On the other hand, people find this very convenient, when they think about their own portfolios. It would be annoying to have to pay taxes on your stock gains every year, and the fact that you get to wait until you sell to pay taxes just sort of feels right to most people. The way mutual fund investing works in the US is not quite like that. You buy a mutual fund for $100, and then it invests your money in some stuff, and it occasionally changes the stuff, and when it does change the stuff you pay taxes. The mutual fund is just a pool of investors’ money, and it passes the taxes on to the investors. When the mutual fund sells stocks — to change its portfolio, or to meet redemption requests from other investors — it incurs taxable gains, which create a tax bill for its investors. So, you buy the mutual fund for $100, and you pay like $3 of taxes per year, and then you sell it for $300 in 10 years and have only $100 of taxable gains or whatever. You pay some of the taxes as you go along, and some at the end when you sell. It is all perfectly reasonable as a matter of tax policy, but people find it annoying in their own portfolios. “Why should I pay taxes on this mutual fund,” they think, “when I haven’t even sold any?” The way exchange-traded funds work is that they are mutual funds that don’t pay taxes. The reason for this is … we’ve talked about it, and it’s fun, but for now it is fine to assume that the reason is “magic.” There is just a magic incantation, which ETFs have discovered but which doesn’t work for mutual funds, and which allows ETFs to work like stocks. You buy an ETF for $100, you hold it for 10 years, you sell it for $300 and have $200 of capital gains. No gains along the way. Very nice. Again: People find this very convenient in their own portfolios, which explains a lot of the popularity of ETFs. But I suppose they might find it annoying in the case of rich people. Bloomberg’s Zachary Mider and Surya Mattu report:
What was once a minor leak in the US tax system is now a torrent that’s costing the government around $48 billion a year.
That’s the amount lost to the Treasury Department from a loophole used by exchange-traded funds to defer or avoid capital gains tax, according to new Bloomberg estimates. The savings go almost exclusively to the highest-earning Americans, the estimates show, and could be set to nearly double following a recent policy shift. …
The ETF tax break is currently saving each of the highest-earning 1% of American households about $13,000 a year, on average, while those in the middle of the wealth distribution get about $23, the Bloomberg estimates show. If the whole mutual fund industry were to adopt an ETF share class, the richest could see their taxes fall by another $11,000. Those in the middle could save an additional $20.
Counterpoint, the US Securities and Exchange Commission has permitted mutual funds to adopt ETF share classes to move investors into the better tax treatment, and:
SEC Chairman [Paul] Atkins pitched the policy change as a reward to “everyday investors.” In an op-ed in the Washington Post, he noted that most American households own mutual funds. “These are our family members, friends and neighbors, and they are more likely to drive a minivan or a pickup truck than a luxury sports car.”
Presumably if you flipped the taxation of individual stocks, so that everyone had to pay capital gains taxes each year on all of their unrealized gains, (1) that would cost the top 1% a lot of money (they own a lot of stocks), (2) it wouldn’t cost the median household a lot of money (they don’t) and (3) you’d find it annoying to the extent that it cost you money.
The increasingly popular claim about prediction markets is that they offer businesses and institutional investors the ability to hedge real-world events: Prediction markets are not just for retail gambling, but an important risk management tool for the real economy. And then the awkward fact about prediction markets is that they are 80% sports gambling. And the synthesis of these two things is: “No, see, they offer businesses and institutional investors the ability to hedge real-world sports events.” This strikes me as dumb but here we are. We talked about it last week, because a bar called The Jeffrey offered free bar tabs if the Knicks won, and hedged on Kalshi. Approximately 80% of my readers then emailed to be like “do you know about Mattress Mack?” (Yes.) I wrote about some other possibilities, including this one:
A basketball team might make more money if it makes the playoffs; maybe its owner should bet against the team to hedge the risk of not making the playoffs. (Controversial!)
This struck me as controversial because, at least in US sports, there is (for now!) a norm against betting against your own team: If you’re betting against yourself, don’t you have an incentive to throw the game? Also, just, like, you’re supposed to have incentives to win the game. “We have hedged, so we are economically indifferent to making the playoffs or not”: No, bad, you should want to make the playoffs. There’s a reason that corporate executives are paid in stock and discouraged from hedging it; the incentives are supposed to be aligned. The whole point of, like, human endeavor is that if you accomplish good things then you get good things. Constructing an efficient hedge that makes you indifferent between accomplishing good things or not defeats the purpose of everything. Certainly of sports! But Liz Hoffman reports:
The owners of a top-tier Spanish soccer team had to move fast. Coming into the final game of the season, the club was on the brink of getting kicked down to a lower league, which would have meant millions in lost ticket and broadcast revenue.
So they turned to Kalshi, placing a multimillion-dollar bet against itself, in case the game didn’t break its way. In the end, the club squeaked by, losing its final game of the season by a narrow enough margin, 1-0, to keep its place in the top tier of La Liga.
But its prediction-market bet shows how these platforms are beginning to outgrow their YOLO beginnings and could one day underpin big parts of traditional finance.
On the other side of the Spanish team’s trade was Susquehanna, according to people familiar with the matter, which says it was the first quant trading firm to establish a dedicated prediction-markets trading desk. Susquehanna didn’t respond to requests for comment, but people familiar with the trade said the firm took home more than $1 million.
I assume that, in underwriting this trade, part of Susquehanna’s process was making sure that the team was hedging less than 100% of its economic risk. Like, if the team stood to lose $20 million in revenue by being relegated, and it made a bet that would pay out $10 million, fine, the incentives still point in the right direction. But if it made a bet that would pay out $40 million, it might decide to field a team of 12-year-olds for that last game. As far as I can tell Susquehanna might be the only firm making these sorts of markets in size on Kalshi, but give that like two weeks. I am really looking forward to reading about a soccer team that stands to lose $20 million if it is relegated, places eight separate $10 million bets against itself with different trading firms, and gleefully gets itself relegated for the payout. Maybe sports will be more fun with a little added market manipulation.
Banks Lay Groundwork for Mass Workforce Cuts as AI Takes Hold. Software buyout deals collapse to lowest level since pandemic after AI rout. AI Is Upending One of Finance’s Cushiest Jobs. Colleges Are Building A.I. Degrees, Hoping Students Will Come. Intesa gatecrashes rival’s bid for Monte dei Paschi with €30.6bn offer. Worries Grow That Hedge Fund Crowding May Amplify Risk in Crisis. Goldman Sachs takes fight to marketmakers in anonymised trading push. Blackstone looks to sell $2bn of stakes in private investment funds. Chilling in Money-Market Funds is the Hot Retail Strategy Now. Franklin’s Western Asset Management agrees $100mn settlement with SEC. The Cattle Empire That Turned Out to Be a Giant Ponzi Scheme. Venture capital for short fiction. 8 Indicted in $4.5 Million Scheme Involving Cheese, Meat and Cigarettes. If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Or you can subscribe to Money Stuff and other great Bloomberg newsletters here. Thanks!
|
No comments:
Post a Comment