| Today is the first lockup release date for SpaceX, which also means it’s the day that a lot of people will find out that they own fake SpaceX stock. SpaceX was a private company for a long time, and a lot of people — current and former employees, venture capitalists, Elon Musk buddies, etc. — ended up owning stock before it went public in June. Some of this stock found its way into special purpose vehicles that were sold to individual investors: The SPV would own a chunk of SpaceX shares, and you could buy shares of the SPV. You’d own SpaceX shares indirectly. Maybe! A lot could go wrong. Mainly: - The SPV might not actually own the SpaceX shares. You’d give the SPV money, which the promoters would spend on yachts; SpaceX would not be involved. How could you check? You could call up SpaceX and ask if the SPV’s name appeared on its shareholder ledger, but it might not tell you.
- The SPV might sell more shares than it owns. The SPV would own $1 million of SpaceX stock but sell $10 million worth; it would sell you a 10% stake for $100,000 but also sell 99 other 10% stakes. It would spend $1 million on SpaceX stock and $9 million on yachts. How could you check? You could ask to see the SPV’s shareholder ledger, but it could just show you a fake one.
We have talked about these problems before, and various hypothetical solutions. But for now, if you are an investor in an SPV that owns shares in a private company, it is hard to be sure that you really own the shares you think you own: The shares are not transferable, so all you actually have is an account statement from the SPV promoters, which might be fake. When the company goes public, it is still hard to be sure that you own the shares. Most initial public offerings have lockups in which pre-IPO shareholders (including SPVs) are restricted from transferring shares for some period of time, so the shares are still not transferable and you still just have an account statement. “We can’t give you the actual shares yet because they are transfer restricted,” the SPV promoters will tell you, accurately or inaccurately. But then when the lockup expires — or is partially released, as is happening with SpaceX today — you can reasonably call up the SPV promoters and say “hey I’d like my shares now.” And then they might distribute to you your pro rata allotment of the underlying SpaceX shares and say “pleasure doing business with you, let’s do this again soon.” Or they might sell the shares and distribute to you your pro rata allotment of the cash, also a pleasure. Or they might say “sorry, you’re breaking up, I’m going through a tunnel, who is this, what is SpaceX?” Here’s a Wall Street Journal story about Late Stage Management, which sold shares in a SpaceX SPV to individual investors: [Investor Ram] Rupireddy wired over money before the end of the year, including $17,250 to take part in a fund that held shares in Elon Musk’s SpaceX, according to documents reviewed by The Wall Street Journal. At the time, he estimated the rocket maker was valued at $58 billion. When SpaceX went public this June at a $1.77 trillion valuation, Rupireddy’s dream of a windfall seemed within reach. It’s turned into more of a nightmare. Shortly after the IPO, Rupireddy and three other investors who spoke to the Journal said they couldn’t log in to Late Stage’s web portal for investors. Rupireddy said the investment firm eventually told him in an email that it sold the SpaceX shares he was exposed to in 2024, when they were worth around $105 each and before a 5-1 stock split, or when Rupireddy’s holding was worth $45,450. But Rupireddy—based on the holdings shown on his investment portal as of May 2026 and on his 2025 tax document—believed he still held the equivalent of 2,500 shares of SpaceX, which at the IPO price he estimated were worth more than $300,000. … Late Stage didn’t respond to repeated requests for comment, including to executives and to lawyers representing the firm in a class-action lawsuit, or to questions sent to the company’s main email address. Ah. Strictly speaking, if you were doing a fraud, you wouldn’t have to take down your web portal shortly after the IPO, since the investors couldn’t do anything with their shares until the lockup release. But now: A test will come on Thursday, when a first wave of pre-IPO SpaceX investors will be permitted to sell shares under so-called lockup agreements. Bankers estimate there are at least a thousand SPVs tied to SpaceX stock alone, and it will be a chance for scores of investors to cash in on the shares’ growth. Or, they could be hit by the same panic that overwhelmed Rupireddy if their share of the profits fails to materialize.
I wonder what the fully bezzle-diluted market capitalization of SpaceX is. Like, Bloomberg tells me that SpaceX’s market cap is about $1.45 trillion. About another $25 billion of shares have been sold short, [1] adding to the supply. So there are people out there who correctly believe that they own about $1.48 trillion of SpaceX stock. Probably there aren’t people out there who incorrectly believe, in the aggregate, that they own billions of dollars of SpaceX stock. But millions? Surely? I think sometimes about the Warren Buffett adage that “Only when the tide goes out do you discover who’s been swimming naked.” I often find that it’s the reverse. If you are running a fraudulent business, and a big financial crisis bankrupts a lot of legitimate businesses, you can tell your investors “hey guys sorry but we just got bankrupted by the crisis, legitimately, ah well,” and they will understand. If you are running a fraudulent business and reporting fake excellent results and nothing bad happens, eventually your investors will ask to withdraw their massive fake profits in cash, and then what. This is of course the plot of The Producers. Anyway if you were selling fake SpaceX shares in 2020 you had almost six years before anyone could reasonably demand their actual shares, but now the time is up. The big picture is something like: - OpenAI and Anthropic are fairly young, fast-growing, somewhat experimental, mostly unprofitable tech startups. As is customary in that situation, they finance themselves with equity: They raise money from venture capitalists hoping for huge upside if they succeed, not from bond investors who want certainty of being repaid.
- Also though like half of the economy is OpenAI and Anthropic debt? Like, every day there is some new 11-figure data center financing in which investors buy billions of dollars of investment-grade bonds essentially backed by OpenAI’s or Anthropic’s commitments to pay for future computing capacity.
That second point is a simplification: In fact, various more mature, public, creditworthy companies are involved in backing the data center debt, because bond investors and credit ratings agencies won’t quite trust OpenAI’s or Anthropic’s credit. But of course the guarantors are trusting their credit. And their trillion-dollar-ish equity valuations provide some good reasons to trust their credit. If the market thinks that the best estimates of Anthropic’s and OpenAI’s future cash flows are around $1 trillion each, then maybe it is reasonable to underwrite, you know, $500 billion each of those cash flows as investment-grade. Their actual capital structures do not involve a lot of debt, but their equity values can be used to create some debt. My point is that in some very loose casual sense the biggest thing in finance is lending against OpenAI and Anthropic equity. So: Why not 11-digit OpenAI margin loans? The Wall Street Journal reports: Global tech investor SoftBank Group used its stake in OpenAI to borrow $10 billion from a group of banks, the company said Thursday, dialing up risk on its giant bet on the AI pioneer. The cash, in turn, helps SoftBank fund an even larger stake in OpenAI: SoftBank is poised to pay another $10 billion for shares in the ChatGPT maker by October, the final chunk of a $30 billion investment announced earlier this year. … While Wall Street banks routinely offer so-called margin loans to investors in publicly traded companies, they are reluctant to make large loans backed by stock in private, loss-making companies given the risks inherent to startups. In a typical margin loan tied to publicly traded stock, a borrower needs to pump more cash into the deal if the stock price falls below a certain level—otherwise the lender can seize the shares. OpenAI’s shares aren’t publicly traded, making it harder to determine how its valuation is changing. Still, SoftBank said the deal does require it to inject additional cash if its valuation falls. “Borrowing against OpenAI to buy more OpenAI” is a decent description of the global economy right now. If OpenAI’s valuation collapses, then the banks that gave SoftBank this margin loan will be in bad trouble, but so will all the other banks. Might as well also do the margin loan. I just want to make a small dumb point about why prediction markets are hot right now. [2] It is: - Modern prediction market event contracts on platforms like Kalshi and Polymarket and Robinhood have one-dollar notional prices: Every contract pays off $1 or $0, depending on whether the underlying event occurs. Every contract is paired with its opposite — Yes pays $1 if No pays $0, and vice versa — so in some loose sense the average price of all prediction-market contracts is always $0.50.
- Their prices move in increments of $0.01.
That’s it. The tick size is $0.01 on a contract with an average size of $0.50, or about 2% of notional value. Meanwhile the normal price of a share of stock in the US is on the order of $100. [3] Stock prices also move in pennies. Therefore the tick size of a US stock is on the order of 0.01% of its notional value. If you go on Kalshi and look at some event contract, it will show the order book for that contract. The best bid (the price you’d get if you sold it immediately) will be, generally, one penny below the best ask (the price you’d get if you bought it immediately). The best bid and ask won’t be equal, because if they were they would have traded with each other; the order book reflects only orders that haven’t traded yet. They won’t be half a penny apart, because these markets move in whole pennies. (I suppose they could be two cents apart, or more, in thinly traded markets.) This doesn’t exactly mean that you will pay one cent per contract to make a prediction-market trade. It doesn’t mean that market makers, on average, make one cent each time they buy and sell a contract. Maybe spread capture is not how market makers make their money. Maybe retail investors are constantly crossing trades with each other at the fair value and not paying a bid/ask spread to anyone. It is suggestive, though. There is famous evidence from US stock prices. Stocks, in the US, used to have tick sizes of $0.125 (“one eighth”). In 2000, they switched to trading in ticks of $0.01, [4] which is called “decimalization.” Studies generally found that bid/ask spreads in normal US stocks — the amount of money that investors pay to trade stocks, and the amount of money that market makers earn for trading stocks — declined significantly when decimalization was implemented. [5] Before decimalization, a market maker would buy at $50 and sell at $50.125 and make $0.125; after decimalization, she would buy at $50 and sell at $50.01 or $50.02 or maybe $50.06. If someone can charge you $0.01 for trading a stock, they might; if the only choices are $0.00 or $0.125, they’ll probably charge you $0.125. Prediction-market tick sizes are 2% of notional, equivalent to charging you a whole dollar for trading a $50 stock. I wrote the other day: Anything that causes retail traders to do more trades is good, for brokers and market makers. Anything that causes them to trade more high-margin stuff — like options, which often have wider bid/ask spreads than stocks — is good, too.
If prediction markets have 2% bid/ask spreads, then that’s a lot of money for market makers, and they can share some of it with brokers. Anyway this is all very simplistic but: - Robinhood announced its second-quarter earnings last week.
- Its customers traded 13.6 billion event contracts. Presumably that means $13.6 billion of notional value (at the $1 maximum payout per contract), or about $6.8 billion of market value (at an assumed average price per contract of $0.50).
- Meanwhile, its customers traded $956 billion worth of stock.
- Robinhood made transaction-based revenues of $776 million, of which $156 million came from event contracts (about 1.1% of notional value traded, or about 2.3% of assumed market value) and $129 million came from stocks (about 0.013% of market value traded).
Which ties pretty closely to tick size: Robinhood makes on the order of one penny per $100 stock trade and also one penny per $1 event-contract trade. That is, Robinhood’s cut of event contract trading is about 100 times as big as its cut of stock trading. That might be why everyone is excited for this burgeoning asset class. My basic take on Leopold Aschenbrenner and Situational Awareness has been that: - Being smart and plugged-in about the long-term rise of artificial intelligence is a pretty good investing strategy, but
- Doing it in a hedge fund with lots of borrowed money that can be called away at any time is a bad funding strategy.
“If you’re a venture capitalist, ‘we will make huge directional bets on the pace of AI adoption’ sounds great,” I wrote on Monday. “If you are a hedge fund running a long/short portfolio at 5x gross leverage, though, it’s worrying.” Aschenbrenner learned that lesson the hard way last week, when the bulk of his public portfolio was cast into the maw of Citadel, but he learned both parts of the lesson: no more leverage, but still full speed ahead on venture capital. Bloomberg’s Hema Parmar reports: Just days after his hedge fund was on the brink of collapse, Leopold Aschenbrenner has made his return to the investing scene, plunking down $400 million to back a privately held company, according to people familiar with the matter. The investment, completed on Tuesday, was the first sign of how Aschenbrenner intends to pick up the pieces of his Situational Awareness after it nearly buckled under a barrage of margin calls from lenders across Wall Street last week. Man good for him. He got married over the weekend too. You might think that between evaporating tens of billions of dollars of investor money and getting married, he might take a break from investing in AI. But I gather the essential thesis of Situational Awareness is that the pace of AI is faster, and the stakes are higher, than you think. There’s no time to stop investing in AI just because you blew up a giant hedge fund! Stereotypically, the way the consulting business works is that firms hire recent graduates of top universities who are good at doing mock case interviews, send them out to big corporations to make PowerPoint presentations about the corporations’ business problems, charge the corporations large fees, and use the fees to (1) pay the recent graduates pretty good entry-level salaries, (2) pay the senior partners who supervise them larger salaries and (3) pay for flights and hotels. A fascinating tweak to that business model is: What if you hired them before they graduated? The work product might be more or less the same: Your frontline consultants are young people with elite educational credentials, demonstrated excellence at doing mock case interviews and not much business experience; the corporations that pay for recent-graduate consultants should pay about the same for undergraduate consultants. Maybe you do fewer on-site projects (because the consultants have to go to class!), so you save on hotel bills. Bloomberg’s Samuel Church reports: The Harvard Undergraduate Consulting Group [was] tasked with looking for vulnerabilities to certain risks among AstraZeneca’s suppliers and suggesting how the company could use data to anticipate problems, according to internal HUCG documents obtained by Bloomberg. It was just one of more than 30 assignments that the group took on last year, including tasks for Delta Air Lines Inc. and Samsung Electronics Co., that helped the club’s net assets reach just over $2.7 million as of June 30, 2025, according to its latest tax filing. Since 2017, the first year that public filings are available, the all-student group has brought in almost $7 million in revenue, the filings show. It turns out that the real savings are in salaries: Apparently the going rate for an undergraduate consultant at Harvard is $0, because (1) modern colleges are magical wonderlands in which all of your material needs are effortlessly met so nobody needs money and (2) conversely, modern colleges are brutal battlefields where students struggle desperately for artificially scarce credentials that will allow them to land plum jobs. (We have talked about undergraduate finance clubs before; sheesh!) So: The club’s prestigious client list makes it one of the most sought-after organizations on campus, despite the fact that, like most student organizations at Harvard, HUCG doesn’t pay the students a salary for the work they put into each mandate. About 450 students apply to join each semester, with an acceptance rate of around 10%, said Noreen Mohamed, HUCG’s director and a Harvard senior. Applicants — some of whom have never heard of consulting before landing at Harvard — go through a rigorous interview process to get their foot in the door of an industry where starting salaries can reach six figures. … The formal application process involves a written questionnaire, followed by first round interviews, a mock case study and a final presentation. Members said the process is meant to mimic how professional consulting firms hire employees. I feel like the 46th-best Harvard student at doing mock case interviews would also be a fine consultant but whatever. [6] The business question here is, if you can bring in a lot of revenue by renting out consultants, but you do not pay the consultants, what do you spend the money on? And the answer is parties: HUCG doesn’t pay students, but it throws lavish parties for them on a scale most student groups could never imagine, complete with bougie locations, open bars and reptile petting zoos. One of last year’s celebrations was held at an estate in Newport, Rhode Island, according to the event’s invitation and several students who attended. Guests were treated to a multi-course dinner with the option of sirloin steak, risotto or duck. After eating, guests floated between the bar, fire pit and bouquet-making station, attendees said. And then there were the snakes. HUCG brought in pythons and other reptiles as part of an exotic petting zoo, the people said. Guests stopped and posed for pictures with the creatures wrapped around their necks. I am an old man and would prefer to be compensated for my work in the form of cash, not in the form of parties with snakes and aspiring consultants. But I can see how, when you are young, you might prefer the parties. Millennium Partners With Anthropic to Develop AI Risk Analyst. Biggest US law firms explore selling stakes to private equity. Alphabet Is Seeking Up to $25 Billion From Latest Bond Sale. Blackstone BDC Profit Drops 94% While Loan Performance Steadies. Inside Intel: how America’s chip champion came back from the brink. Armed With $10 Billion, Sequoia’s Leaders Plan Its New Era. Meta Releases Coding Agent to Compete With OpenAI and Anthropic. OpenAI Models Joined Forces Months Ahead of Hugging Face Hack. Paramount agrees safeguards for UK approval of $110bn WBD deal. Deutsche Bank, KBC Freeze Some Radiant World Funds in Singapore. Hedge Fund Albar Is Returning Cash to Run Only Millennium Money. Startup Raises $700 Million to Replace Data-Center Wires With Light. “The most valuable pregame signal in football might be whether the head coach decided normal pants would not cut it.” 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! | 
