I used to think a lot about “tokenized real estate.” During the crypto boom, people would sometimes pop up and say “until now, it was impossible for ordinary people to own a fractional share of a commercial building, but crypto makes that possible. We can issue tokens representing ownership of the building, on the blockchain.” I found this very confusing. In fact, for decades, it has been possible to issue shares of a commercial building. It’s traditionally called a “REIT,” a real estate investment trust; most REITs have lots of buildings, but a single-building REIT is possible. There are plenty of REITs listed on US stock exchanges. What did crypto bring to the table? I thought there were three main answers:
Technology: Maybe tokens on a blockchain are faster, more efficient, more modern, more 24/7, etc., than shares listed on the stock exchange.
Legality: Shares of a REIT — shares of anything, really — are securities under US law, and selling them to the general public requires registration with the US Securities and Exchange Commission, financial statements, disclosure, etc. Many people think this is a good thing — it’s good for investors to know the legally required information about the building they are buying shares in — but crypto people generally thought it was a bad thing, and preferred to avoid SEC registration. Perhaps issuing “tokens” rather than “shares” would avoid the registration requirements.
Gambling: It always struck me that, while there was a lot of enthusiasm from crypto founders to tokenize real estate, there was never any obvious enthusiasm from anyone for buying shares of buildings. Did ordinary people want to own some shares in some random commercial buildings? But arguably crypto solved this problem too, simply because in, like, 2021, there was a lot of speculative enthusiasm for owning all sorts of random crypto tokens, which would maybe carry over to owning tokenized real estate. “Tokens always go up, so these real estate tokens will go up, so let’s buy some.”[1]
Were these good arguments? Point 1 — “blockchain is good financial technology” — was plausible; it probably is a bit easier to spin up a new real estate tokenization than a new stock exchange listing. Point 2 struck me as completely wrong — “Just because you mumble the word ‘blockchain’ doesn’t make otherwise illegal things legal,” I once wrote — but opinions differed, and I am no longer sure that the mumbling is ineffective. Point 3 was the interesting one. The theory there was something like “if we have identified a new gambling behavior, we should co-opt it to finance real-world activity.” Like if people are going to blow their money speculating on memecoins, why not channel that speculation into building useful commercial real estate? (Anachronistically: Why not channel it into building AI data centers?) People want to gamble, you give them a gambling token, they give you money, you use the money to build useful commercial real estate, the gambling token goes up or down in value based on (1) the value of the real estate plus (2) independent speculative reasons, everyone is happy-ish.
I wrote a few times about tokenized real estate because it was a clear illustration of the issues, but in theory you could tokenize anything. Unique works of art, private-company stock, an athlete’s performance, whatever. All of these tokenizations raise similar issues: Are they securities? Do you have to register with the SEC to sell shares in them to the public? Can you capture some of the crypto speculative fever and use it to finance art or companies or athletes or whatever?
In many ways the successor to the crypto boom, these days, is the boom in prediction markets. And there is an analogous move in prediction markets, which nobody calls “predictionization” but which I will. Like! “Until now, it was impossible for ordinary people to get fractional financial exposure to a commercial building, but prediction markets make that possible. We can list an event contract on ‘Net Operating Income at 123 Fake Street Will Exceed $1,000,000 in 2027,’ on a prediction market, and people can buy that contract if they want exposure to that building, or sell it if they don’t.”[2] What do prediction markets bring to the table?
Technology: Maybe prediction markets are faster and more flexible at listing new contracts than stock markets are at listing new REITs, they have better user experiences, etc.
Legality: Is a bet on net operating income at 123 Fake Street a “security,” in the way a share of 123 Fake Street would be? I think there’s a pretty good case that the answer is “yes,” and we have talked about proposed “KPI binaries” (prediction-market bets on key performance indicators like the net income or total revenue of a public company), which arguably do count as securities derivatives and have to be registered with the SEC. But this is not at all clear, and there’s a decent argument that these contracts are not securities.
Gambling: In 2026, arguably, there is a fair amount of enthusiasm for gambling on prediction markets, what with them being large and growing sports betting platforms that advertise nationally. Maybe channel some of that enthusiasm into real estate!
Again I’m using real estate as an example because it clearly illustrates the issues, but again you could predictionize anything. Most notably, we have discussed the fact that a contract like “OpenAI to go public at a $1.5 trillion valuation” is a way for ordinary investors to get economic exposure to OpenAI before it goes public: Buying that contract is the rough economic equivalent of buying a share of OpenAI,[3] but you can do it without OpenAI filing anything with the SEC or disclosing its financial statements. (For now, you can only do it on Polymarket, which in theory is off-limits to US investors; Kalshi doesn’t list a similar contract, presumably because it is too obviously an unregistered securities investment.)
Kalshi wants to become a major challenger to traditional financial markets by encouraging big funds and companies to use prediction markets to protect against losses. Behind the scenes, it’s developing new contracts to win those big investors over.
That includes wagers aimed at niche corners of finance like the market for legal claims, where investors buy up claims at a discount betting they’ll be worth more once settlements are resolved. Last month, at a big investor’s request, Kalshi added new prediction contracts on the timing and size of payouts in the $2.8 billion settlement between college athletes and the National Collegiate Athletic Association.
The idea came from investment firm Grand Teton Systems, which has bought about $400 million worth of claims from athletes. Louis d’Origny, co-founder of GTS, said he reached out to Kalshi because he wanted a way for claim holders to hedge the risk that the payments might come in smaller or later than expected.
Under a court-approved settlement from last year, the NCAA is due to distribute back pay to college athletes who said they were illegally barred from sharing in revenue it made from broadcasting their games or using their name, image and likeness. But female athletes have challenged the settlement, saying they aren’t getting their fair share compared to male athletes, and that has put payouts on hold. ...
D’Origny said he suggested new contracts tied to some of the yet-to-be-resolved details. About 72 hours later, two new contracts went live on Kalshi—one allowing customers to bet on whether women will get a bigger share of the back pay and a second on when payments will begin. “Whats exciting is that this is a genuine hedging opportunity for real world risk,” d’Origny said.
Kalshi is veryinterested in “hedging,”[4] for somewhat obvious reasons.[5] But I’m not sure the important implication here is hedging. I think it might be distribution. I think the real long-term product here is something like:
GTS puts $400 million into litigation finance claims.
It would like to distribute some of those claims to retail investors to reduce its risk, recycle capital, make money from moving not storage, etc.
If it went and did an offering of $400 million of litigation finance claims to retail investors, that would be a complicated legal and marketing process.[6]
But if it got Kalshi to list a bet on those claims paying off, that’s easy! All of the process is on Kalshi’s side, the distribution is already there, the legal framework is easy (the US Commodity Futures Trading Commission lets Kalshi list almost whatever it wants) and everybody likes a gamble.
So ultimately GTS is able to syndicate part of its litigation investment to retail investors on Kalshi.
Now, I should say, this is a bit loose, and there are a few problems with this story in this particular case. First, the Kalshi contracts are not quite “how much will these claims pay out” or whatever; they are on “Will women get a bigger share of NCAA settlement back pay?” and “When will NCAA settlement back-pay payments begin?” They are not direct offsets to GTS’s cash flows.
Second, when a retail investor buys a litigation investment on Kalshi, GTS doesn’t get the money: GTS would pay cash for a “No” contract, rather than receiving cash. This is a solvable problem, but it requires more financial infrastructure. (We discussed that infrastructure in June.) Basically if GTS buys litigation finance claims and then buys offsetting hedges from Kalshi, it ends up with a package with a guaranteed cash flow, in any circumstance, of at least $X[7]: Either the claims pay out more than $X, or they pay out less than $X and the Kalshi bets make up the difference. That $X should then be financeable quite cheaply. You buy litigation claims, you sell the litigation-claim risk to retail, and you sell the remaining risk-free cash flows to, like, a bank for cash. Or to a private credit firm. Or your prediction market prime broker — or the market’s own clearinghouse — offers you portfolio margin on your combined position, lending you the cash against the guaranteed cash flows.
Third, the other problem with this is that, in this case, it is teeny. As of this morning, those two contracts had less than $6,000 of total volume. Roddan:
The contracts Kalshi created for d’Origny so far have drawn meager volume, and GTS itself hasn’t traded them yet because it hasn’t seen a favorable position to take. “We would hedge more exposure if we saw interesting odds,” d’Origny said.
Yeah, nobody wanted to trade tokenized real estate either.
Asymmetric bets
Here’s a pitch:
The Trump administration does a bunch of random wild stuff that has the effect, and quite possibly the purpose, of increasing financial market volatility.
If you know that stuff in advance, you can buy call options on the thing Trump is about to promote or the oil he’s about to make more scarce or whatever, and make a quick killing
I know that stuff in advance, because I have a guy on the inside.
So you should invest your money with me. The fees are high.
To be clear, this is a hypothetical pitch; I don’t know anything in advance. But if someone came to you with that pitch, would you find it compelling? There are some obvious risk factors:
Pretty illegal, though this is arguably more of a problem for the person making the pitch than for you as an investor.
There are probably 20 people out there making this pitch right now for every one person who actually gets inside tips from the Trump administration.
The fees are high.
This pitch was definitely better a week ago, back when the Trump administration was pretty loosey-goosey and not subject to any checks or oversight. They’re going to have less fun going forward.
Feels like you’re going to get a lot of volatility from this strategy. Even if you do have a guy in the administration, you’re going to get some bum steers.
Here’s a slightly refined pitch:
Random stuff, know it in advance, etc.
But I’m not a lunatic: If you invest your money with me, I will park it in Treasury bills. The Treasury bills are totally safe[8] and will pay me 4% a year. I will spend the 4% per year buying options on the wild stuff my guy tells me about. If those options pay out, I will quadruple your money and we’ll all get rich on the volatility. If not, I’ll lose the 4%, but you’ll still get your money back.
Is that pitch better? No! You could just invest 96% of your money in Treasury bills, earn a safe guaranteed return, and give me 4% of your money each year to yeet into wild Trump volatility trades. I am providing two entirely separate services, one of which (buying Treasuries) should basically be free, while the other (insider yeeting) is obviously bespoke and expensive. Why should you pay me to hold Treasury bills for you? Just size your trade appropriately.
Still, the form of this trade has an enormous psychological appeal, and we talk about it all the time. I once called it “a well-known bit of derivatives magic” and “a great, simple party trick that derivatives structurers can use to impress their friends.” You give me some money, I put enough of the money into risk-free Treasury securities to pay you back 100 cents on the dollar (without interest), and I use the rest (the interest) to buy wild risky stuff. I guarantee you your money back (without interest), so you sleep well at night. But I also buy you some lottery tickets, so you wake up in the morning excited to see what happened.
The most classic form is the literal “lottery bond,” where everyone invests money (often with the government), most people get their money back without interest, and some people get their money back with so much interest. It’s a barbell strategy: Most of your money is very safe and pays no return, and some of your money is absolutely nuts. For some reason people seem to like it when someone else does the packaging. Perhaps nobody trusts themselves to size the nutty trade appropriately.
Anyway, Bill Ackman is thinking again about his idea for a “black swan” or “asymmetric bets” fund, which we talked about in April. On Pershing Square Inc.’s earnings call on Tuesday, Ackman described the structure of the fund:
Our version of an asymmetric hedging-type entity would look like a U.S. treasury money market fund. We invest the capital in short-term U.S. treasuries, and the interest income, which today is something in the, let’s say, 4-ish-percent, would be our budget for the asymmetric investments. We design, in effect, a fund where you can’t lose money because your principal is protected by treasuries. So if you spend 400 basis points over the course of the year and you earn no return on that investment, people’s capital is protected, so we like that kind of structure as opposed to a so-called tail risk fund [where] we raise a smaller amount of capital, and over time, it burns up and goes to zero unless the world ends. This, we think, is sort of a much better model.
In terms of the — what we’re looking for, our big bets in this respect, and they tend to be actually relatively small amounts of capital, but they’re big notional bets, have been in cases where we’ve had a view of a black swan type event, think financial crisis, COVID crisis, or massive inflation spike in the 2021 timeframe that would cause the Fed to raise rates. We basically identified these as potential risks and then looked for an instrument that offered us an asymmetric payoff that justified some of the uncertainty associated with any of these kinds of bets.
But he also described a trade that got away (emphasis added):
The closest we came to putting on one of these bets in the existing portfolio was, I would say, late last year, early this year, when we said, look, it’s likely that we’re going to have some kind of war in Iran; that’s going to lead to a spike in oil prices. Is there a way to hedge that event in a way that made sense? And the problem was that the rest of the world was aware that there was a likely war in Iran that would lead to higher oil prices, and also because of the uncertainty, volatility was high. So we looked at a number of potential structures, and if you got it precisely right, you might have made 3x your money if you bought a one-month oil option and you had a friend in the administration who leaked to you they were going to war. I mean, that would be the best way to make money there. So we did nothing. We have nothing on today.
It’s possible that in five years we’ll look back on, you know, Tuesday and think “huh, it was weird that big institutional hedge fund managers were like ‘well obviously the best way to make money would be to buy short-dated call options when your friend in the administration gives you advance notice of a war.’” But he’s not wrong! Or he wasn’t, on Tuesday.
Pacing the frontier
There is a cynical antitrust theory of pacing the frontier. The idea is[9]:
Frontier artificial intelligence labs are big businesses in a very lucrative and competitive industry.
They have to spend a lot of money on researchers and computing power to build exciting new products that their customers will love and pay a lot of money for.
If one lab stopped doing that, its competitors would eat its lunch: Other labs would quickly introduce better products that customers would find more delightful, so the customers would all migrate to those labs’ products and the slower lab would lose revenue.
Everybody in every business would prefer to have the same high revenue but with lower costs and a more chill lifestyle.
The obvious solution is for all of the frontier AI labs to get together and agree to slow down: Spend less on researchers and computing power to build exciting new products, slow down the introduction of products, and keep charging customers the same prices for the existing products. If nobody introduces exciting new products, no customers will have any incentive to switch, and they’ll keep paying for the existing products.
So the labs would keep the same high revenue, but with much lower costs (for researchers and compute) and less stress.
If they just did that — if the big labs all got in a room and agreed “we’re all going to enter into a binding agreement to slow down our development to save money and not have to compete so hard” — then they would quickly be marched off to jail. That’s a classic antitrust conspiracy in restraint of trade!
But if the government passed a new law saying “all AI labs have to spend less money on compute and slow down introducing new products,” then they could all slow down, secure in the knowledge that their revenue is safe and their competitors would also slow down.
This theory has nothing to do with AI per se; this theory ignores everything that people find distinctive about frontier AI. This is a theory that applies equally to, like, widget factories, or social-media apps or business-to-business software-as-a-service. If you think that AI will cure cancer and usher in an age of miracles, then slowing down the introduction of delightful new products (like curing cancer) is an immense human tragedy. But, of course, if you think that AI will take over the world and kill everyone, then slowing down the introduction of exciting new products (like the world-ending superintelligent demon) is quite good.
As far as I can tell lots of people at the leading AI labs sincerely believe both things, and are willing to slow the cancer cures somewhat as the price to pay for making sure that the superintelligent demon is under control, which is an extremely reasonable tradeoff if you believe both things.
If you sincerely believe both things, you will want a coordinated slowdown in AI research purely for demon-control reasons. If you believe neither thing and are just in the widget business, you might want a coordinated slowdown purely for traditional monopolistic-profit reasons. If you are a cynic, the move is to get the government to believe in the demon, so that it will force everyone to slow down and maintain high profit margins at lower costs. “Guys guys guys, we don’t want you to unleash the demon, so we are going to force you to simultaneously slow down your spending,” the government says, and you grin.
If there were a coordinated slowdown in frontier AI development, what would that do to the profits and stock prices of the big AI labs? This is a bit hard to measure, in that the purest-play AI labs (OpenAI, Anthropic) are not public yet, but three possible answers are:
They would go down, because building frontier AI is lucrative, but the AI labs are willing to put safety over profits.
They would go up, because the cynical antitrust theory is true and slowing down AI development makes frontier AI labs more lucrative.
They would go up, because slowing down AI development happens to make AI labs more lucrative, but that’s not why they want to slow down; they’re really worried about the demon.
Anyway, in September, Anthropic Chief Executive Officer Dario Amodei put out an essay (“We Must Pace the Frontier”) calling for frontier AI labs to slow the development of AI for demon reasons, and suggesting that this could be done “via regulation that targets all US frontier AI companies.” Failing that, though, “AI companies can and should voluntarily work together to set standards,” though “with government mediation or waivers of antitrust restrictions.” (We talked about this, and the antitrust view, at the time.) Other frontier AI executives mostly agreed, but the Trump administration did not, which left this a bit in limbo.
But then this week … I don’t want to say “the US voted for massive restrictions on AI,” because that is not really how US politics work, but there was a pretty big electoral backlash to data center construction, and arguably to AI more broadly. And so, like, chipmaker stocks have fallen, because any version of this story — “cynically pump up AI lab profits,” “slow the demon,” whatever — is bad for Nvidia Corp.’s revenue. But elsewhere in tech, ehh, I dunno. Bloomberg’s Jeran Wittenstein and Ryan Vlastelica reported yesterday:
The shares of US technology giants led the market higher on Wednesday as investors bet that a Democrat-controlled Congress will slow the massive spending on artificial intelligence that’s eating up free cash flow.
Amazon.com Inc. was up as much as 8.3%, Microsoft Corp. gained 7%, Alphabet Inc. jumped 5.9% and Meta Platforms Inc. advanced 3.6%. Together with Apple Inc., which has avoided heavy AI spending and leaped more than 6%, the five tech behemoths are on pace to add nearly $1 trillion in market value on Wednesday. …
“The market is acting like this represents the end of the AI buildout,” said Robert Pavlik, senior portfolio manager at Dakota Wealth Management. “If that means the hyperscalers aren’t spending ridiculous amounts of money to build out compute, that means they could become cash-flow positive again.”
They’re down today though. I would have said “hyperscalers are spending ridiculous amounts of money to build out compute because they have made calculated bets that that will increase their cash flows in the long run,” but markets do not have to agree with them. The effect on pure frontier AI labs is less obvious. Anthropic is down a bit on Hyperliquid, but that OpenAI valuation contract on Polymarket has barely budged.
A&O Shearman will invest $1.1bn in its artificial intelligence strategy over the next five years, as law firms jostle to come out on top in an AI arms race.
The “magic circle” law firm told the FT that partners had committed to the investment this autumn, the majority of which will come from each practice spending at least 3 per cent of billable time on building AI tools for their area of law, as well as funds for the technology.
Practice heads will be able to decide how the 3 per cent target is delivered, either assigning people to work on it full time or spreading it across associates and partners. The firm will monitor each practice’s hours to ensure the commitment is met.
Like, if your law firm brings in $7 billion of revenue a year,[10] then pretty much 100% of that revenue comes from selling your lawyers’ time, and there is a quite straightforward and market-validated conversion rate between your lawyers’ time and dollars. And if you commit to spending 3% of your time for the next five years thinking about AI, well, you can just plug that into the conversion rate and conclude that it’s a $1.1 billion investment. It’s not a $1.1 billion investment in the sense that you are taking $1.1 billion of dollars and using them to purchase capital goods, or even to hire people to think about AI. But your time is money in some fairly literal sense, so spending enough time thinking about AI counts as a $1.1 billion investment.
Except! Except that the big question about AI in the legal industry is what it will do to the billable-hour revenue model. In four years, when A&O Shearman is starting on the last tranche of this investment (thinking about AI for another year), will it still be billing clients by the hour? Will the AI systems that it spent the previous four years building not be doing a lot of the thinking about the clients’ legal problems? (Will the AI systems be building themselves, for that matter?) The first thing A&O Shearman will have to do, in the 3% of its time it devotes to thinking about AI, is think about how to value its time.
Things happen
Goldman Taps 109 New Partners in Biggest Class Since 2010. Wendy’s Rejects Takeover Bid From Trian to Take It Private. Inside CVS’s Decision to Break Up a Healthcare Juggernaut. The Making of Dario Amodei. OpenAI to Change How It Handles Data Center Efforts After Outcry. China Blocks $5.4 Billion SoftBank Robot Deal, Flexing Control Over AI Tech. Election Raises Hurdle for Data Center IPOs After Peers Tumble. Developers Eye Golf Courses for Data Center Projects. Demand for Forward-Deployed Engineers Is Leading to $1.5 Million Paychecks. The Prediction-Market Bettors Who Turned Election Night Into a Real-Time Casino. UBS slashes spending on consultants as Credit Suisse project winds down. Gene Editing Is Giving Crops a Way to Tell Farmers When They’re Feeling Stressed. Oral history of Gronk.
Follow Us
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.
lot
