The Money Stuff column titled Not Your Keys, Not Your DAT, dated Wednesday, October 28th, 2026. Programming note: We want to hear from you. Please take our quick reader survey. Section one: Bitcoin oopsie For a while last year, the US stock market would pay 2 dollars for 1 dollar of Bitcoin. It is hard to convey now how weird that time was. If you ran a tiny public company in mid-2025, the very best thing you could do was acquire 100 million dollars of Bitcoin to add 200 million dollars to your market capitalization. If you happened to have 100 million dollars worth of Bitcoin in mid-2025, the very best thing you could do was to plop it into a tiny public company in exchange for 200 million dollars worth of stock. These companies were called digital asset treasury companies, or DATs. Everyone kind of knew this couldn't last (it didn't), so there was a mad rush to get these trades done before the party ended. In September 2025, a company called Datavault AI Inc. got into this game. Sort of. Datavault was a small public company, a leader in AI-driven data visualization, valuation and monetization. In mid-September, it had about 116.7 million shares outstanding, trading at around 30 cents per share, for a market capitalization in the ballpark of 35 million dollars. On September 25th, it announced that it was selling 278.9 million shares, at a price of 53.78 cents per share, to Scilex Holding Co., another public company focused on acquiring, developing and commercializing non-opioid pain management products for the treatment of acute and chronic pain and neurodegenerative and cardiometabolic disease. (Start of Footnote one: Incidentally, this is not the first weird situation involving Scilex that we have discussed around here. End of Footnote one) Scilex was paying the purchase price in Bitcoin, valued at one hundred twelve thousand nine hundred fifteen dollars and 52 cents per Bitcoin, meaning that Datavault got about one thousand three hundred nine Bitcoin for its stock. (Start of Footnote two: Datavault was selling other stock in other transactions, and by the time the deal closed in November, Scilex owned about 265 million shares, out of about 552 million outstanding, for about 48 percent. End of Footnote two) The stock promptly shot up, hitting a high of 3 dollars and 11.8 cents in October, for a market capitalization of about 973 million dollars. Datavault did not, as far as I can tell, call itself a Bitcoin treasury company or DAT, and the announcement of the deal focused on its strategic goals, not the Bitcoin stuff. A sample: We believe the investment is particularly significant for biotech, strengthening the company's high-performance computing capabilities and supporting partnerships with the U.S. Department of Energy's Brookhaven National Laboratory and IBM WatsonX, said Nathaniel Bradley, CEO of Datavault AI Inc. By leveraging Web 3.0, blockchain-secured data trading, and AI-driven analytics, Datavault AI aims to deliver secure, scalable solutions addressing trust, data integrity, and monetization challenges. The partnership with Scilex will also support the continued development of previously announced platforms, including the International Elements Exchange, International NIL Exchange, and American Politics Exchange. Still it was a big slug of Bitcoin. In round numbers, putting 150 million dollars of Bitcoin into Datavault immediately added 900 million dollars to its stock market value. Wild times! The wild times didn't last, of course. The price of Bitcoin fell, DAT premiums collapsed, and by January Datavault had sold 491 Bitcoins back to Scilex for 40.5 million dollars, about a 27 percent discount to the price four months earlier. (It moved on to other things, like a dividend of Dream Bowl Meme Coin II tokens.) By the end of June, the date of its most recent financial statements, Datavault reported holding 837 Bitcoin, with a fair value of about 49 million dollars. It reported total assets of about 277 million dollars, with stockholders' equity of about 246 million dollars, so it is no longer especially a Bitcoin company. This Monday, Datavault's stock closed at 13.28 cents per share, for a market cap of about 126 million dollars. (It has issued more stock, and currently has about 950 million shares outstanding.) Never mind. But there is a hilarious coda to this story: Actually Datavault never got any Bitcoins at all. Yesterday it announced that its financial statements for the past year or so should no longer be relied upon, because (emphasis added): On September 26th, 2025 and November 25th, 2025, Scilex credited one thousand three hundred nine point zero four Bitcoin to the Company's digital wallet maintained at BICONOMY PTE. LTD. (Biconomy) through an off-chain transfer. Between September 26th, 2025, and January 30th, 2026, the Company sold 491.431 Bitcoin reflected as credited to its Biconomy account back to Scilex through an off-chain transfer in exchange for approximately 40.5 million dollars in cash (the Bitcoin Exchange Transactions). In June 2026, the Company requested a withdrawal of Bitcoin from the digital wallet maintained at Biconomy to an external account and has, as of the date of this Current Report on Form 8-K, been unable to complete the withdrawal. Subsequent to the issuance of the Affected Financial Statements and following a review of the transaction, the Company determined that the Bitcoin associated with the Scilex Equity Purchase did not meet the requirements for recognition as a digital asset of the Company. That is: In the original deal with Scilex, Datavault gave Scilex stock and Scilex gave Datavault one thousand three hundred nine Bitcoins at Biconomy, a large crypto exchange. Datavault did not own those Bitcoins on the Bitcoin blockchain; it owned them in an account on the ledger of Biconomy. Over time, it sold some of them back to Scilex for cash, again in transfers on Biconomy (not the blockchain). And then, this June, Datavault tried to take its remaining Bitcoins out of Biconomy and, uh, has been unable to complete the withdrawal. Its accountants concluded that it never had the Bitcoins at all: They did not meet the requirements for recognition as a digital asset of the Company. They were entries on a ledger at Biconomy, but when Datavault tried to turn them into actual on-chain Bitcoins, it couldn't. Oops! The stock closed down 50 percent yesterday. I don't know what to tell you. CoinDesk contacted Biconomy about the withdrawal request, but has not yet reported a response. It is not clear what happened. Maybe nothing; maybe the Bitcoins are all there, this is all a misunderstanding (since June?) and it's all fine. But it does kind of seem like, for a while last year, the US stock market would pay 2 dollars for zero dollars worth of Bitcoin. Section two: People are worried about non-GAAP accounting We talked a bit last week about Anthropic PBC's first-half net loss of about 80.7 billion dollars. Virtually all of that net loss was caused by changes in fair value of some stock-based instruments (convertible notes and preferred stock and purchase agreements) that Anthropic had previously sold, largely to Google and Amazon. Essentially Anthropic sold those investors some stock, the stock went up by about 80 billion dollars in the first half, those investors had about 80 billion dollars of profit, and so Anthropic had about 80 billion dollars of loss, which flows through its income statement. I tried to point out that that's weird. Most of the time, when a company issues stock and the stock goes up, it does not recognize a loss under US generally accepted accounting principles. Anthropic has a GAAP loss because these instruments are not quite stock; they have weird provisions that make them partially derivative liabilities under GAAP and thus require mark-to-market accounting. For most practical purposes, these instruments are stock, but as a technical accounting matter they are treated as mark-to-market liabilities. But if they were just regular stock, they would not have created any accounting loss, and when they eventually convert into regular stock they will stop being marked to market. It seems to me that this is a case in which US GAAP creates a misleading picture of the company's economic profitability. The story here is something like we basically broke even in our business, but because we issued some weird stock, we have an 80 billion dollar non-cash accounting loss reflecting the fact that our stock went up. But the stock going up is a good thing! And so in fact the general reaction to Anthropic's financials has been positive, and analysts talk about an adjusted net income of about 2.7 billion US dollars, adjusting out those mark-to-market losses. Because the mark-to-market losses do not tell you anything useful about Anthropic's actual profitability. There is, however, a widespread perception that US GAAP is objectively correct, and that if a company reports adjusted (non-GAAP) results, it is doing something tricky and misleading. At the Wall Street Journal, Jonathan Weil writes: There's a flip side to Anthropic's 80.7 billion dollar net loss for the first half of this year. On the other side of the ledger, Amazon.com and Alphabet booked large paper gains on some of the same financing deals. This tees up a vexing question as Anthropic prepares for an initial public offering as soon as November: Should gains and losses like those count in the alternative earnings metrics widely touted by Wall Street stock analysts? At Amazon and Alphabet, the gains count. But at Anthropic, the corresponding losses don't, or at least not in the company's adjusted profit figures. Those unrealized gains and losses are part of net income under generally accepted accounting principles. The analysts who follow Amazon and Alphabet included the gains in their consensus Street earnings metrics, even though the valuation marks are volatile and unrelated to the companies' core operating performance. Anthropic, which publicly filed its IPO prospectus Monday, is directing investors to a nonstandard metric called adjusted net income. By this measure, Anthropic had a 2.7 billion dollar profit in the first half, instead of the 80.7 billion dollar net loss. About 72 billion dollars of the excluded losses arose from marking various liabilities, including convertible notes, to market values. Excluding those losses arguably makes sense, on the grounds that they aren't indicative of ongoing operations. But if that is true, then the analysts who cover Amazon and Alphabet should be excluding the corresponding gains from the companies' Street earnings. But this is a coincidence. It happens to be the case that 80 billion dollars of the increase in Anthropic's stock value in 2026 counts as a mark-to-market loss for Anthropic, under US GAAP. But most of the increase in Anthropic's stock value, over the past few years, does not count as a mark-to-market loss for Anthropic, and it would be weird if it did. On the other side of the ledger, various Anthropic investors have 1.5 trillion dollars or whatever of gains from owning Anthropic stock, of which 80 billion dollars reflects GAAP losses for Anthropic and 1.42 trillion dollars doesn't. The 80 billion dollars of GAAP losses are a genuinely weird exception to the normal rule that, when a company issues stock and the stock goes up, that does not create an accounting loss for the company. On the other hand, when a company issues stock and the stock goes up, that does generally create an accounting gain for the owners of the stock. There is nothing weird about this asymmetry: If an investment bank makes money by trading stocks, everyone understands that that's income for the bank, but there's no offsetting loss for the issuers of that stock. That said, it is a choice in the accounting rules: It used to be, under US GAAP, that most companies did not report income when stocks they owned went up, but in 2016 that rule was changed, and now they do. Warren Buffett complained a lot about this at the time, arguing that it would severely distort Berkshire's net income figures and very often mislead commentators and investors. I disagree: that I bought stocks that went up, so I have income seems basically normal and correct, in a way that Anthropic issued stock that went up, so it has a loss does not. But if you're used to treating unrealized stock gains as not income, then treating them as income might seem misleading. Generally, though, it seems intuitive to me that: 1. When you buy stock and the stock goes up, you have economic and accounting income. (But not tax income!) (Start of Footnote three: See the next section. End of Footnote three) 2. When a company issues stock and the stock goes up, the company does not have an economic or accounting loss. 3. The books don't balance: Your gain is not offset by the company's loss. 4. Such is life! This is about accounting, but there is also something importantly true about it as an economic matter. If investors have made 1.5 trillion dollars by investing in Anthropic, that's not because Anthropic lost 1.5 trillion dollars. It's because Anthropic created 1.5 trillion dollars of value, which is shared with its investors. Stock investing, unlike some asset classes I could name, is positive-sum, and making money in stock investing does not require anyone else to lose money. (Start of Footnote four: See the last section. End of Footnote four) Section three: Buy borrow die Eventually no one will pay capital gains taxes in the US. The general rule is that you pay capital gains taxes when you sell stock (for more than you paid for it), so the way to avoid capital gains taxes is to never sell stock. There are two main reasons you might want to sell stock: 1. To buy other stocks (diversify, rebalance, etcetera), or 2. To pay for your lifestyle. So the trick is do those things, rebalance your portfolio and fund your lifestyle, without selling stock. The first problem is solved, and we have discussed it a few times recently. The trick is that, while you have to sell stock if you want to get rid of it and buy another stock, an exchange-traded fund doesn't have to do that. An ETF can get rid of stock it doesn't like in an in-kind redemption, acquire new stock in an in-kind creation, and not pay taxes on the transaction. This is a weird and possibly accidental quirk of US tax law, but at this point it is quite robust. But I'm not an ETF, you say, but that is a minor quibble. You can become an ETF: Someone creates an ETF for you, you contribute your stocks to the ETF, boom, you're an ETF and never need to pay taxes to adjust your holdings. This is called a 351 exchange ETF. Now. There are two problems with this. One is that it is expensive and complicated to create an ETF, and you probably can't do this with a ten thousand dollar portfolio to avoid 500 dollars of capital gains taxes. But over time that obstacle will go away: All of this stuff will get productized, automated, and cheaper. As with many tax trades, as with many trades, generally, this starts off as a bespoke expensive trade for very rich people, and ends up commoditized and available to everyone. The other problem is that the Internal Revenue Service recently announced a crackdown on too-cute versions of this trade; we talked about it last month. It is not yet clear to me how thorough this crackdown will be; the result might be in fact you cannot generally avoid capital gains taxes on rebalancing your portfolio, or it might be more like you need to pronounce the right incantations and wait three months and then it's fine. The second problem is also largely solved. The standard solution is called buy borrow die, and it goes like this: 1. You have 100 million dollars of appreciated stock. 2. You borrow money, from a bank or brokerage, against the portfolio. Say you borrow 50 million dollars at 6 percent interest. (Start of Footnote five: The Wall Street Journal says that the average loan to value is around 50 percent, while the charge for the borrower is around 1 percentage point over benchmark rates. Realistically you'd borrow at a floating rate; I'm using 6 percent, a bit more than 100 basis points over long-term SOFR swaps, for simplicity. End of Footnote five) 3. Now you have 50 million dollars to fund your lifestyle, and the borrowing is not taxable. 4. Of course you owe your broker 3 million dollars a year in interest. But your portfolio grows by more than 3 million dollars per year: If stocks return, say, 8 percent per year on average, then your portfolio grows by 8 million dollars a year, more than enough to cover your interest bill. (Just add the interest bill to the loan amount!) 5. Eventually you die, never having paid back the loan, and the net value of your stock (net of the loan) is higher than when you started. 6. Now your heirs inherit your stock with a stepped-up basis: If they sell it, they don't have to pay any capital gains taxes. 7. They sell enough stock to pay back the loan and keep the rest tax-free. Again, an extremely well-known approach. Again, there are occasional proposals to crack down on it, wealth-tax proposals are, deep down, kind of about buy borrow die, and occasionally politicians propose more direct approaches, but again it seems pretty robust. And, again, this is the sort of thing that starts as a bespoke trade for billionaires but is actually pretty easy to roll out to everyone. In fact, it is arguably easier to offer this to retail customers than to billionaires, because the retail customers are more diversified. If you have one billionaire client who owns 30 percent of the company she founded, lending her 50 percent of the value of her stake, 15 percent of the value of the company, is a concentrated risk and a complex credit decision. But if you have a million normal customers all of whom own some stocks, their overall portfolio is going to look like the stock index, and lending 50 percent of the value of that portfolio will be pretty straightforward. (Start of Footnote six: I mean, if you're a wealth manager in California, it's plausibly going to look like a portfolio of the five biggest AI companies. But otherwise. End of Footnote six) Here's a Wall Street Journal story about the commoditization of buy borrow die: Deloitte estimates that around 5.4 trillion dollars is extended in securities-based lending, mostly against stock and bond portfolios but also funds and crypto. The figure has risen sharply as more global wealth pours into markets and asset prices rise. Deloitte consultants said the market's growth rate averaged 12 percent over the past 10 years. It was more of a wealth management product in the past and not for the masses, said Eric Gutzwiller, who leads the credit business at Deloitte Switzerland. Now, not only the ultrahigh net worth, but also more and more the lower wealth segments all the way down to the affluent, and even retail, starts to explore that product. One perk for borrowers and for banks is the speed of this kind of lending. There is often no need for in-depth personal credit checks when the collateral is secured by the lender. Pricing is largely standard and can be calculated quickly based on the market value and liquidity of the asset. Xavier De Pauw, director of lending at Monument Bank in the U.K., said it would start lending against portfolios beginning early next year, so investors with a few hundred thousand dollars can adopt a billionaire mindset. We would like to make it available to pretty much everybody, De Pauw said. Right, these days it would be kind of weird to pay capital gains taxes. The technology exists not to. Incidentally, people sometimes worry about stock market leverage. A lot of stock market leverage has the basic flavor of people seeking higher risk: Retail margin trading, leveraged ETFs, hedge-fund prime brokerage, etcetera, all basically involve investors borrowing more money to increase their bets on stocks. But some stock market leverage is more like this: People borrowing half the value of their portfolios, not to increase their bets, but to increase their consumption without paying taxes. Section four: Sports bets Here is a Wall Street Journal op-ed by Tarek Mansour of Kalshi, arguing that sports bets on prediction markets are totally different from sports bets on sportsbooks. It covers the usual arguments. Kalshi, unlike sportsbooks, has no house that wins when retail customers lose. (Market makers might quibble.) Bars can hedge their beer sales by betting against the local team on Kalshi, but few hedge risk by putting a sportsbook bet on their profit-and-loss statements. (Mattress Mack has been hedging at sportsbooks for years.) Etcetera. But the argument that drives me particularly crazy is that we need federal regulation for sports bets: We recognize and support the calls for more regulation, including to address unique challenges in the sports and elections contexts. But those changes need to happen at the federal level, not through a patchwork of state rules that would fragment the marketplace. Imagine subjecting the New York Stock Exchange to 50 different sets of state laws. It would dry up the liquidity that comes with a national pool of traders and diminish the pool of expertise that determines prices. National markets require national regulation. Casinos don't. Yet when prediction markets started offering customers access to a regulated peer-to-peer exchange, state regulators who oversee casinos brought lawsuits supposedly to protect consumers. Against what exactly? They aren't demanding that sports speculation end. They seem to be fine with sportsbooks, which have fueled addiction nationwide for almost a decade. That's not really true, though. It's true of some states: Some states that are fine with state-regulated sportsbooks sued Kalshi for ignoring their state gaming regulations. (Start of Footnote seven: Those gaming regulations, notably, usually restrict sports gambling to customers who are at least 21 years old; Kalshi's federally regulated sports bets are open to 18-year-olds. States might worry that 18-year-olds are more prone to addiction than 21-year-olds, and that federal sports gambling regulation is not particularly focused on addiction risk. End of Footnote seven) But, for instance, Utah sued Kalshi for offering illegal gambling, and Utah is not fine with sportsbooks. Its attorney general said: Not only do our state laws ban gambling but Utah is one of only two states where gambling is also prohibited under the state constitution. For 130 years, Utah has been clear: gambling has no place here, and we will enforce our laws. Mansour's argument is that national markets require national regulation, and it would be silly for sports gambling to be regulated by the states. But sports gambling has been regulated by the states for almost the entirety of US history. Utah has banned sports gambling, in its constitution, for as long as it has been a state. Until this year, nobody thought that that was odd. Nobody thought we need federal sports gambling regulation to avoid fragmenting sports gambling liquidity. Nobody thought that Utah's constitution violated federal law. In 2018, the US Supreme Court ruled that states can regulate sports gambling; nobody at the time argued that the the Commodity Exchange Act prohibited that. Kalshi, and the US Commodity Futures Trading Commission, have discovered an entirely new area of constitutional law, in which century-old state laws about sports gambling are invalid because they conflict with the important federal interest in offering 18-year-olds liquid nationwide sports gambling markets. It's so weird! Section five: Things happen Wall Street's AI Debt Machine Is Starting to Choke on Supply. How Weil Gotshal, One of Wall Street's Most Storied Law Firms, Ended Up in Chaos. Nelson Peltz's Trian Offers to Buy Wendy's for 1.4 billion dollars. ETFs use perps to offer bets on private companies such as Anthropic and OpenAI. (Earlier.) JPMorgan's Skyscraper on Park Avenue Estimated to Cost Around 6 billion dollars. McKinsey, BCG Boost Starting Salaries for New Hires in US. US MBA Applications Plunge 27 percent as International Demand Drops. Home Buyers Who Enjoyed Cheap Adjustable Mortgages Are Now Getting a Nasty Shock. AI maths advances raise fears over security of global encryption. Where Have All the Car Commercials Gone? Polymarket's top trader is sports gambler Swiss Tony. Walking a robot dog paid around the same hourly rate as that for the flesh-and-blood variety. That was the Money Stuff column titled Not Your Keys, Not Your DAT, dated Wednesday, October 28th, 2026, as read by Robo Matt. All credit for Money Stuff goes to the biological Matt Levine and Bloomberg. Subscribe to the money stuff newsletter and podcast at bloomberg dot com.