| In traditional finance, there are stock exchanges and there are brokerages. Stocks are bought and sold on the stock exchange, but you can’t just walk in and buy some stock. Only certain selected professionals can trade stocks on the stock exchange. If you want to trade stocks, you have to open an account with a brokerage; your broker will buy the stock for you on the stock exchange. [1] You can’t just open an account on the stock exchange and buy stock. The stock exchange doesn’t have accounts for retail investors. This is more or less a matter of historical contingency, though there are some benefits to this division of labor, and stock exchanges and regulators are sort of attached to it now. One arguable advantage of this division is that retail brokerages can be kind of embarrassing, without thereby rendering the whole stock market kind of embarrassing. Like: Robinhood used to show little animations of confetti when its customers bought stock, and regulators found that embarrassing, but it was embarrassing for Robinhood. If you were a stock trader at Goldman Sachs or Jane Street or BlackRock or whatever, moving tens of millions of dollars of stock at a clip, and your uncle was like “heheheh what about the confetti,” you could reasonably reply “that has nothing to do with me, I do not use Robinhood, that’s a retail brokerage, I am a sophisticated professional and there is almost no confetti at the New York Stock Exchange.” Nobody who started a retail financial market from scratch today, or even 10 years ago, would set it up this way, though. Nobody would be like “I am going to set up a central venue for trading of Pokémons, and offer a limited number of memberships to professional Pokémon trading firms, who can sign up customers and trade on their behalf on my venue.” What? No. You set up a website, and you let people open accounts on the website, and they can trade with each other. Perhaps professional trading firms will come to your website, perhaps you will enter into special arrangements with professional market makers to provide liquidity, but in general you are going to have an account-opening function and let people open accounts. In financial history this is more or less an innovation of crypto — crypto exchanges were the big financial institutions that most notably combined the roles of exchanges and retail-facing brokerages — but realistically it is an innovation of websites. The New York Stock Exchange started as a building where people met; the building could only fit so many people, and they all had to be in New York. A financial platform starting in the 2020s is naturally an online platform, and those constraints do not apply, so there is no real need for a separation of the exchange and brokerage functions. On the other hand, if your retail customer acquisition functions are embarrassing, that in a sense makes the whole market embarrassing. Here’s a Wall Street Journal story about how Polymarket pays influencers to pretend to make winning bets on Polymarket, to get other people to make bets on Polymarket: George Makihara ... is one of dozens of mostly college-age creators Polymarket paid to film themselves making fake trades and sometimes scoring fake wins, according to an analysis of more than 1,100 videos by the Journal, along with instructional materials and interviews with creators who have worked with the company. ... In its push to draw users to its unregulated platform, Polymarket has flooded social media with videos like Makihara’s, which appear genuine at first glance. In reality, Polymarket built near-perfect copies of its website, then instructed creators to make simulated trades on those dummy sites and hide that they were being paid by Polymarket. .. Polymarket said in a statement that it was “committed to maintaining accurate, fair, and transparent markets. We are part of a rapidly growing industry and are constantly evaluating ways to improve how we’re engaging and earning the trust of our audience.” The lead anecdote is about Makihara making a fake bet on Donald Trump saying the word “McDonald’s” in public in January, and pretending to win $100,000; in the real world, “more than 50 accounts made the McDonald’s bet in January,” and “all of them lost” because Trump didn’t say it. Oops! Also some of the fake bets were made on “poiymarket.com, which is indistinguishable from polymarket.com when the ‘i’ is capitalized,” terrific. There are definitely people running pump-and-dumps and boiler rooms in the stock market! But those people aren’t also running the stock market. We talked a few weeks ago about debt financing for artificial intelligence. People want to build huge amounts of AI infrastructure (data centers, chip fabs, nuclear plants, etc.), which might require borrowing trillions of dollars. The lenders want to see some steady cash flows. A data-center builder who strikes a leasing deal with Anthropic might have a hard time borrowing billions of dollars to fund development, “because Anthropic is still a startup and lacks the earnings to support such a large debt deal.” And so a custom has developed in which some big investment-grade-rated public company — Broadcom, Nvidia, Meta — backstops the borrowing. With that backing, the bonds get an investment-grade rating and investors are happy to buy them. There might, however, be a shortcut that bypasses some of this complexity. The leading AI labs are at this point trillion-dollar-ish companies, even if they don’t have a lot of earnings. (Anthropic expects net income of $559 million on revenue of $10.9 billion this quarter, its first profitable quarter.) “This data center is leased to a money-losing startup, but it’s really cool”: No, bad, not what bondholders want. “This data center is leased to a trillion-dollar company”: Yes, better. Even better, though, is “this data center is leased to a trillion-dollar public company.” Doesn’t “a trillion-dollar public company” sound creditworthy? Or $2 trillion. Bloomberg’s Brian Smith and Michael Gambale report: SpaceX is selling investment-grade bonds for the first time in what’s expected to be the start of a massive borrowing spree to fund the company’s AI ambitions following its record $75 billion IPO. … Elon Musk’s rocket, satellite and AI conglomerate is seeking to raise at least $20 billion from the offering, Bloomberg reported last week. Proceeds will refinance the bridge loan of roughly the same size, a facility that makes up the bulk of SpaceX’s $29.1 billion of long-term debt. SpaceX received ratings in the BBB tier from all three major bond graders last week, paving the way for cheaper borrowing. Moody’s Ratings and Fitch Ratings graded SpaceX’s debt at Baa1 and BBB+ respectively, or three steps above junk. S&P Global Ratings assigned a BBB rating, one notch lower. SpaceX lost $4.3 billion on revenue of $4.7 billion last quarter. “We expect to incur significant capital expenditures over a period of years before our AI products and services … become profitable, which may never occur,” SpaceX warns investors. In one sense, SpaceX is a giant AI startup that lacks the earnings to support a large debt deal. In another sense, SpaceX is a giant public company and ehh it’s fine, it’s fine. Moody’s said: SpaceX's Baa1 issuer rating reflects the company's exceptional franchise strength as the world's leading orbital launch provider and operator of the largest low earth orbit (LEO) satellite broadband network, Starlink. The company benefits from robust and expanding recurring revenue from Starlink, which has become the primary cash flow generator and underpins improving scale, margin expansion, and diversification away from more cyclical launch revenues. ... The AI segment addresses a large total addressable market across AI infrastructure, consumer and enterprise applications, and digital advertising, and represents significant long-term revenue and earnings upside as the company scales its compute infrastructure, develops its frontier models, and expands enterprise and consumer monetization channels. The company maintains conservative financial policies, alongside strong liquidity and financial flexibility, supported by substantial cash balances and continued access to both equity and debt capital markets.
Some of that — about the rockets and satellites — is about SpaceX’s actual current ability to generate cash. Some of it — “the AI segment addresses a large total addressable market … and represents significant long-term revenue and earnings upside” — is just “this is a big AI lab.” Some of it — “access to both equity and debt capital markets” — is just “this is a giant public company and people like its stock.” It’s possible the first component — the money-generating rockets — is essential to SpaceX’s investment-grade ratings, but I’m not sure. “This is a public company with plausible plans to make trillions of dollars selling enterprise AI, and people are lining up to buy tens of billions of dollars of its stock at whatever price it demands”: Doesn’t that sound like a good credit? Also doesn’t that describe Anthropic and OpenAI? I mean, not now, but when they go public? I guess the point here is that if you’re a big AI startup, you should go public not only to raise tens of billions of dollars of equity, but also to raise tens of billions of dollars of debt. Being a trillion-dollar startup is cool, but it might not move ratings agencies or bond investors; being a trillion-dollar public company will. Once there is an established public market price for your equity, and that price is $2 trillion, it is easier to persuade people that you’re a good credit risk. In that vein, here’s a 2025 working paper by Andres Almazan, Nathan Swem, Sheridan Titman and Gregory Weitzner titled “Access to Capital and the IPO Decision: An Analysis of US Private Firms,” with basically that takeaway: Going public allows firms to raise debt capital on more favorable terms. Indeed, we find that interest rates drop on firms’ bank debt after they go public. …
The reduction in asymmetric information after the IPO allows firms to borrow from a broader pool of lenders. First, we document an increase in the number of banks that IPO firms borrow from after going public. Second, we find that IPO firms increase their use of syndicated loans and bonds dramatically after going public. These results suggest that going public not only provides firms with additional equity but also facilitates access to syndicated loan and public bond markets, the former of which is consistent with anecdotal evidence that IPO activity is an important determinant of aggregate bank lending.
Elsewhere in AI lab credit, the Information reports on OpenAI: At first glance, the company portrayed by the AI startup’s financial statements, reviewed by The Information, resembles a lean, low-debt software business. Its balance sheet, as at March 31, had zero debt and less than $750 million of lease liabilities. Its cash-flow statement, moreover, showed OpenAI—one of the most hardware-centric tech businesses—spending just $46 million on capital expenditures in the quarter, less than even Salesforce, which sells business software. The reality is messier. The fine print of the financial statements show OpenAI has $665 billion of purchase commitments stretching out over the coming years for chips, energy and data centers. That’s because the company mostly rents computing capacity in other companies’ data centers, meaning it spends billions of dollars in leasing payments that flow through its income statement. A huge portion of modern financial markets is ultimately built on top of OpenAI’s creditworthiness. The sooner it goes public, the sooner it will, uh, have an investment-grade credit rating. You know the basic story. You buy some stuff at a store, using your credit card. This generates data: Your phone (and many of the apps on your phone) knows that you were at the store, and your credit card company (and perhaps your personal finance app) knows how much you spent and what you bought. Various intermediaries buy this data, aggregate it and sell it to hedge funds. The hedge funds know, more or less in real time, how many people go into each store each day and how much they spend, which helps them predict retailers’ earnings. “Sophisticated money managers have long turned to intel from credit-card receipts or satellite images of retailers' parking lots into an investing edge,” Bradley Saacks wrote last year at Business Insider, “but this data is increasingly becoming table stakes.” If you want edge now, you have to look for ever more obscure sources of alternative data. But that is all a stock-market story, and now we have prediction markets, which is to say that now sports gambling is integrated into financial markets. If you’re a sports trader (that is: an advantage sports gambler) at a hedge fund, what sorts of alternative data should you be chasing? I am sure that we will find out. The next frontier in alternative data will be sports data, and it’s going to be extremely annoying. For now, here’s Alex Shultz at Hard Reset (via Bruce Schneier) on wearable technology provisions in the Women’s National Basketball Association’s collective bargaining agreement: Wearables present serious privacy issues for “Average Joe” consumers, who are entrusting tech companies to safely store and protect their biometric data. Imagine the stakes for a professional athlete, whose entire livelihood could be affected by a single biometric data point. To give one of many realistic hypotheticals: a basketball player has a terrible game, and the coach wonders if they showed up to the gym hungover. The coach has access to the player’s wearable data, and checks to see when they went to sleep, as well as what their heart rate looked like during the night. ... It will not surprise you to learn that there’s an emergent gambling angle here: sports leagues would love to commercialize players’ biometric data, and sharp bettors would love access to data about, say, a hungover player. … The WNBA’s CBA clarifies that “commercial use of the data collected from an approved Wearable ... will require prior approval of the Players Association.” There’s an important clause at the end, though: the WNBPA’s approval “shall not be unreasonably withheld.” Traditionally credit card data is anonymized before it is sold to hedge funds, but I’m not sure that makes sense for sports data. You don’t want to know how many people in your metro area are drunk; you want to know if your point guard is drunk. Elsewhere in sports hedging, here’s a guy who’s gambling some free watches on the World Cup, and hedging on Kalshi. One of the coolest weirdest worst financial instruments in recent memory is Strategy Inc.’s “Stretch” preferred stock. Stretch is a floating-rate preferred stock, but its dividend rate floats not with some benchmark interest rate but rather with Strategy’s own market-clearing interest rate. Stretch pays a monthly dividend, and it trades on the stock exchange, and each month Strategy resets the dividend to try to make Stretch trade at $100 per share. If the dividend rate is 10% and Stretch trades at $80, for a 12.5% yield, then next month Strategy should reset the dividend rate to 12.5% to make it trade at par. When Strategy launched this thing last year, I compared it to auction-rate securities, which were the pre-2008 implementation of this idea. What is the idea? The idea is something like “a long-term fixed-income instrument with no duration.” [2] Stretch remains outstanding forever, but it is nonetheless essentially one-month financing: From Strategy’s or an investor’s perspective, it is economically equivalent to selling one-month debt and rolling it over every month. Of course, if you have to roll your debt over every month, you take the risk that one month you won’t be able to: Debt markets will get bad, or investors will get nervous about your credit, and they won’t want to buy your new debt. Strategy also takes that risk: Even though it doesn’t actually sell new debt every month (the Stretch is perpetual), its interest expense can go up rapidly if markets, or its credit, get worse. Stretch, though, is a softer and safer version of one-month financing. It’s perpetual, it’s preferred stock, and the dividend reset is in the discretion of the board. If Strategy decides “never mind, we’re going to let it trade at $80,” or even “never mind, we’re not going to pay dividends to conserve cash,” it can do that. The dividends on a preferred stock are not mandatory the way interest on debt is. But, of course, if you stop paying dividends on preferred stock, getting new financing — selling more preferred stock, for instance — gets much harder. You can’t exactly default on preferred stock, but if you stop paying it people will ask questions. Anyway, Strategy sells this thing to buy Bitcoin. At some level, the story here is “Strategy sells an assortment of weird preferred stocks to buy Bitcoin,” fine. But I think that the more useful analysis is: “Strategy has recreated banking to buy Bitcoin.” I kind of think that is Strategy’s analysis, and I wrote last month: The story of banking is that it is a magic trick for transmuting risky assets into safe liabilities. A bank lends money to companies and homeowners, which is risky (they might not pay back the money). The bank gets the money by issuing deposits, which are safe. A $1 deposit at a bank should always be worth $1. … You might think: “If people put their money somewhere safe, like a bank deposit, and then the bank invested the money in Bitcoin, that would make everyone happy. The people with money would feel safe, but also Bitcoin would go up. The magic transmutation of banking increases society’s capacity to take risk on Bitcoin. There will be more productive investment in Bitcoin, because of this magic.” … Essentially Strategy is issuing dollar-denominated one-month financing (at 11.5%!) designed to always pay back par, and using that financing to buy a big chunk of all the Bitcoins that are for sale. So that’s neat! Strategy is not actually issuing demand deposits to buy Bitcoin, because Stretch has perpetual maturity and Strategy can turn off the dividend. But it has approximately the economic shape of issuing demand deposits to buy Bitcoin, in the sense that, if the market gets nervous about Strategy’s balance sheet — for instance, because the value of its assets (Bitcoin) goes down — then Strategy’s cost of funding will immediately go up, just as it would for a regional bank. Anyway, at the Financial Times last week, Craig Coben had a good article about Stretch: The problem is that keeping Stretch at par has proved far more expensive than expected. Since the listing in July 2025, Strategy has raised the dividend five times, from 9 per cent to 11.5 per cent. Even that has not been enough. Stretch closed yesterday at $89, implying a current yield of roughly 13 per cent. If Saylor wants to drag the security back to par, the dividend will probably have to rise again, perhaps to something closer to 14 per cent. The difficulty, of course, is that Strategy generates little meaningful operating cash flow from its legacy software business. It can service those dividends only through fresh capital raising or by selling bitcoin. Coben analogizes Stretch to a “death spiral convertible,” an instrument that we have discussed, but I think that the better analogy is to a bank run. Strategy’s depositors (Stretch holders) have gotten nervous about the credit, so the interest rate has gone up. Unlike an actual bank, Strategy can’t be forced to give back their money, but an increasing interest rate is almost the same thing, and in fact Coben argues that Strategy should give back their money to save on interest expense: The cleanest solution may be a buyback. It requires a certain amount of chutzpah to position Stretch as a money market fund-like instrument and then repurchase it only months later, particularly against the backdrop of a stock market setting new records. Yet Stretch is beginning to operate like a tapeworm inside the Strategy belly. The longer it remains there, the more nutrients it consumes. It may be better to expel it sooner rather than later.
“Let’s issue bank deposits to buy Bitcoin” sounds like a great idea when Bitcoin is going up, [3] but it has some obvious problems when Bitcoin goes down. We have talked a couple of times around here about the re-equitization of the stock market. That is: For a few decades now, US public companies have bought back more stock than they have sold, because they were generating more money than they needed and returned the extra to shareholders. But now we are seeing a shift, where companies are projected to sell more stock than they buy. I have explained that shift in a pretty traditional way: Companies need more money than they generate, so they sell stock to raise it. We are near the start of a giant buildout of artificial intelligence infrastructure; companies expect that AI buildout to eventually generate huge profits, but for now it is expensive. Companies need to invest more in AI than they can generate organically from their cash flows, so they sell stock to raise the money. But there is an alternative, also very traditional trading story. That story is: “Companies sell stock when they think their stock is overpriced.” At the Wall Street Journal, James Mackintosh tells that story: When companies as a group turn into sellers, it’s a reasonable sign that stocks are very overpriced. Indiscriminate demand from investors incentivizes companies to step up to supply them, both with secondary offerings, like Alphabet’s recent record-breaking issue, or IPOs.
That story has sort of opposite implications. The we-need-lots-of-money-to-build-super-profitable-AI story suggests that companies expect the present value of all of this AI investment to be very high, so they are willing to sell stock to build it. The lol-our-stock-is-too-high-so-let’s-sell story suggests something else. The ‘Mass Affluent’ Are Losing Their Allure for Wealth Managers Navigating AI. Kevin Warsh’s push to axe Fed guidance may lift US borrowing costs, investors warn. Coke Takes on IRS With $20 Billion at Stake. Morgan Stanley Pitches Clients on a New Market for Data Center Loans. Morgan Stanley Plans $1.3 Billion Tower for Dallas Expansion. Alan Greenspan, Who Led Fed During Boom Before 2008 Bust, Dies at 100. A Hoarder’s Redemption: Bed Bath & Beyond Will Take Your Decades-Old Coupons. Denunciato recisiones. 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! | 
