| Let’s say I lend a software company $100 for five years at 8% interest. I give the company $100 in cash, and it gives me back a note promising to pay me back, which I enter into my accounting system as a $100 asset. The next day, Anthropic releases some new artificial intelligence tool, people get nervous about the software business, and software stocks and bonds drop. The yields on loans to other software companies, companies that are similar to my borrower, go up by 0.5% over the course of the day. I call up my borrower to check in, and they say “oh yeah, we saw that headline but we’re not too worried, everything’s fine.” How much is my asset — the company’s note — worth? There is a conventional answer, which is about $98. [1] The price of a debt instrument moves inversely with interest rates; as the market demands higher interest rates from borrowers like my software company, the value of those companies’ existing debt goes down. If I sold my note for $98, the buyer would get a yield of 8.5%, which is the new market rate. The “mark-to-market value” of my note is $98. But I might object: Well, no, I’m not going to sell my note for $98, or for any other price. I’m just going to wait my five years, collect 8% interest along the way, and get $100 at the end. I thought that proposition was worth $100 yesterday, and I think it’s worth $100 today. Nothing has changed, with the asset, or the company, or me. Oh sure sure sure stuff in the outside markets has changed, and other people are more nervous about software credit. But I don’t see how that affects me. I am not more nervous. They’ll probably change their minds tomorrow; it is nonsense to say that this loan was worth $100 yesterday and $98 today and for all I know $101 tomorrow. It’ll pay $100 in five years with interest, it’s a perfectly good performing loan, it’s worth $100 until I have some objective reason to think that it isn’t. [2] Probably a lot of you got very itchy reading that last paragraph, and I promise that I got very itchy writing it. But I want to argue that it’s not insane, and that which answer — $98 or $100 — is “correct” depends on the purposes for which you are using the answer. Here are some purposes: - If I run a hedge fund, the right answer is $98, and it is very important that I mark this note to market in my accounting system. For one thing, just as a matter of intellectual rigor and market discipline, my job is to make money every day, and I need to know exactly what every asset is worth on the market so that I can make good trades. (If I am trading notes like this every day, it seems silly for me to say “well I won’t sell this one so it’s worth $100.”) Also, a lot of outside people rely on my marks. I charge my investors a performance fee — 20% of their gains or whatever — based on market returns, so I need to price everything using market prices. And I probably have a lot of leverage — I borrow a lot of money from banks, secured by my investments — and the banks are sure going to mark those assets to market. The banks are giving me margin loans against the value of my assets, and if that value goes down — measured by market prices — I will get margin calls.
- If I run a mutual fund or exchange-traded fund, the right answer is $98, and mark-to-market accounting is again very important. Not, mostly, for the hedge fund’s reasons: I don’t charge performance fees or, probably, borrow much. But because the investors in my mutual fund can get daily liquidity: Every day, they can take money out or put money in. They will do this “at NAV”: If they take money out, I will pay them the net asset value of their shares; if they put money in, I will sell them shares at the net asset value. If the net asset value of my fund does not reflect market prices, this will collapse: If my investors can take out $100 when their shares are only worth $98, then they will take out too much money and leave remaining investors with too little. Also, if the investors take out too much money, they can effectively force me to sell the note (to pay them back), so its value really has to be $98, the price I could sell it for today.
- If I run a bank, I am a very special snowflake and generally am not supposed to mark loans to market in my financial statements. “I loaned the company $100, nothing has changed, so that’s worth $100” is the normal way that banks account for loans. The theory of a bank is that it can make loans and hold them to maturity, so it doesn’t need to mark them up and down to reflect fluctuating markets. There is some complication and nuance to this, a lot of people don’t like it, and it is not clear that it is really defensible anymore. We talked about it a lot in 2023, when there was a bit of a crisis in US regional banking, much of it caused by this sort of thing: Banks do not generally mark their loans to market in part because they assume they have stable funding and can hold the loans to maturity, and the 2023 regional banking crisis was a stark reminder that banks’ funding isn’t really that stable.
- If I am just a person investing my own money, none of this matters at all and I can tell myself anything I want. It is probably psychologically healthier for me not to check the market too often or worry too much about fluctuating market prices, so here $100 does kind of feel like the right answer.
Of course the trillion-dollar question right now is: What if I run a private credit fund? I lend a software company $100, the market starts to panic about software, I think the loan is fine: How do I value it in my financial statements and investor communications and my secret heart of hearts? It’s worth thinking about from first principles: - In some deep sense, what I am selling to investors, as a private asset manager, is a psychologically healthy relationship with their investments: I am selling them an asset that doesn’t trade every day, that doesn’t fluctuate with market prices, that doesn’t overreact to news, that appears uncorrelated with public markets. The fact that public markets are panicking about software credit has nothing to do with me, and deep down it is my job to shield my investors from those fluctuations. So, $100.
- A thing that I and others have emphasized about private credit is that it has a much more stable funding model than banks: Banks fund themselves with short-term deposits and so are vulnerable to runs that might force them to sell assets, while private credit loans are funded mostly with long-term locked-up investor capital that can’t run. So if banks don’t mark to market — “we don’t need to sell so the loan is worth what we paid for it” — then a fortiori neither should private credit. So, $100.
- But that’s only mostly true. In fact, private credit firms run retail-focused funds (“business development companies” or BDCs), some of which do allow investors to take money out — not as much as they want anytime they want, but a limited amount once per quarter. These transactions happen “at NAV,” and just like any other mutual fund, it is important that the net asset value be correct. If my net asset value doesn’t reflect market prices, investors are going to rush to redeem, because they want to get 100 cents on the dollar for assets that are worth 98. And if I pay them out at 100, then the remaining investors’ assets will only be worth 96, etc. So, $98.
- Also, private credit funds do use leverage. They don’t generally borrow as much or as short-term as banks or hedge funds, but they do some borrowing, some of it from banks, secured by their assets. The banks will take a keen interest in how much the assets are worth, they will want to keep their loan-to-value ratios low, and they might even have the right to send margin calls if the assets lose too much value. If I am getting margin loans against my assets, you’d better believe that those assets are getting marked to market — and probably by the bank, not me. So, $98.
- Also, like hedge funds, private credit funds tend to charge management fees (a percentage of assets) and performance fees (a percentage of profits), so correctly measuring asset prices is important. (This is less important in private funds, which often charge fees based on committed capital and realized distributions rather than market prices.)
One way to put it is that, when you look at the simple core function of private credit — raising long-term locked-up funds from institutional investors with low leverage to make loans that are held to maturity and don’t trade — you might think “ehh it’s fine not to mark to market very carefully; just assume most loans are worth what you paid for them.” And then if you look at the modern expansion of private credit — raising lots of semi-liquid money from individual investors, borrowing more money, trading the underlying loans — you might think “eek, actually it’s very important to mark to market rigorously and frequently.” And so you might tell a story in which private credit started out with a bank-like, hold-to-maturity, “it’s worth $100 until they stop making payments” mentality, and is now transitioning into a rigorous real-time mark-to-market mentality. And so Bloomberg’s Laura Benitez and Hannah Levitt report today: Apollo Global Management Inc. is ramping up its efforts to give investors more regular insight into the value of its opaque private credit holdings, just as a spate of redemption requests from such funds rattles the wider market. The firm, which manages $938 billion of assets, is preparing to start reporting the net asset values of its credit funds on a monthly basis, John Zito, the co-president of Apollo’s asset management arm, said in an interview. Ultimately, it aims for both daily NAVs and third-party valuations over time. Meanwhile, though, there is a lot of pressure on private credit firms’ net asset values. We have talked a lot recently about how private credit BDCs have been getting a lot of redemption requests from individual investors, and how they either have or have not met those requests. One way to think about it is that those redemption requests are a form of NAV arbitrage. If a BDC says that its loans are worth $100, but you think their market value is only $98, you should try to redeem out of that BDC: You’ll get cashed out at $100, which is more than you think it’s worth. We talked about this back in January, when I wrote that, “if you let investors take their money back, you will generally do it at NAV, … so NAVs really matter”: It is a classic bank run dynamic: If a private credit investment is worth $7, but you can cash out for $8, you should, and everyone will, and eventually there will be $0 left. And everyone knows this and will be twitchy.
The fact that private credit BDCs are getting historically high levels of redemptions might suggest that their investors don’t believe their marks. And BDCs limit redemptions to guard against this: If too many people are cashing out, that suggests that the fund is cashing them out for more than the actual value of their shares. [3] The fund sort of lives in the old hold-to-maturity world, but its investors live in a mark-to-market world. You know who else lives in a mark-to-market world? At the Financial Times, Jill Shah and Eric Platt report: JPMorgan Chase ... informed private credit lenders that it had marked down the value of certain loans in their portfolios, which serve as the collateral the funds use to borrow from the bank, according to people familiar with the matter. ... The loans that have been devalued are to software companies, which are seen as particularly vulnerable to the onset of AI. ... One person briefed on the bank’s decision said the valuation haircuts did not trigger margin calls at funds but were done to pre-emptively reduce the amount of credit available to the funds. … Publicly traded software stocks and debt have all plummeted this year. Private credit lenders, by contrast, tend to hold loans for the entire term and have not marked down their portfolios in lockstep. Private lenders have said enterprise software companies are still growing and expect their loans to continue performing, as investors backstop the borrowers. “Enterprise software companies are still growing and [we] expect their loans to continue performing” is not something you can say to a margin lender. Or at least, not to JPMorgan: JPMorgan is somewhat of an outlier in the private credit financing business as it reserves the right to revalue assets at any time. Most other banks require triggers such as missed interest payments.
At some level, marking private credit loans to market still feels wrong even to most banks. I wrote yesterday about how Kalshi and Polymarket are now offering binary options on where the S&P 500 stock index will close. There are, you know, various things to say about that, but one thing I said is that binary options are a particularly appealing product for retail investors because they are so simple. We discussed a 2025 paper,“Overvaluing simple bets: Evidence from the options market,” by Aaron Goodman and Indira Puri, finding that traders systematically pay more for binary options than they do for “strictly dominant bull spreads,” that is, conventional option combinations that pay more than the binaries in every situation. Why would you pay more for a product that is always worse? Well, because it is simple and easy to understand. But then I went on to quote the Bloomberg News story about the prediction-market S&P 500 contracts, which said that “the price on Kalshi for the S&P 500 Index ending the year between 8,000 and 8,200 points” was about 4 cents on the dollar ($2,190 for a $44,000 payoff), while “in the options market, one could pay a $2,190 premium to bet on a 8,000/8,200 call spread,” with a maximum profit of $20,000 per contract. And I said: “Sounds like the Kalshi call spread is cheaper?” Quite a few readers emailed to point out that, nope, these are not comparable trades: If the index finishes above the 8,200 cap, the Kalshi contract pays out zero, while the conventional call spread pays out the maximum. It is not quite strictly dominant (it pays less in the 8,000/8,200 band), but it is also not right to say that the Kalshi call spread is cheaper. Sorry. ChatGPT is not your lawyer | Nothing in this column is ever legal advice, but in this section, you should especially assume that every sentence ends with “(not legal advice!”). Anyway: We talked yesterday about whether Cluely did a little light securities fraud. Roy Lee, the co-founder and chief executive officer of the artificial intelligence startup, told a TechCrunch journalist some wrong numbers for Cluely’s annual recurring revenue, which the journalist then published. Later, Lee apparently asked ChatGPT “is this securities fraud,” and ChatGPT said no. Then he posted the exchange on X. My point yesterday was that several people sent me this exchange to be like “well, is it?” and that I basically agree with ChatGPT. But I also pointed out that “there are various not-best-practices things here,” and asking ChatGPT for legal advice is one of them. Not only in the sense that ChatGPT might get its legal advice wrong, but also in the sense that ChatGPT is not your lawyer and so your conversations with it are not privileged. If you call your lawyer and say “hey I think I did a little light securities fraud, what do you think,” and your lawyer says “yep definitely fraud,” your lawyer won’t go call the police to turn you in, and the police can’t make her: Your conversations with her are subject to attorney-client privilege. Whereas ChatGPT is not a lawyer, it’s a chatbot, and so it might turn over your conversations to the police. Here’s a Perkins Coie memo from last month: On February 17, 2026, the Southern District of New York, in United States v. Bradley Heppner, held that a criminal defendant's written exchanges with a “publicly available AI platform” are not protected by attorney-client privilege or work product doctrine and, thus, could be inspected by the government. The ruling appears to be the first federal decision squarely addressing privilege claims for communications with a generative AI platform of this type.
He was allegedly doing some securities fraud, got caught, asked ChatGPT for advice, and ultimately had to turn over the advice to the FBI: After receiving a grand jury subpoena and becoming aware that he was the target of a criminal investigation, the defendant—acting on his own initiative without instruction from counsel—used a free generative AI platform to prepare documents outlining his potential defense strategy and responses to potential charges. … The defendant later provided the AI-generated documents to his counsel, who reviewed them. In October 2025, a grand jury indicted the defendant on securities fraud and related charges arising from alleged misconduct as a public company executive. Upon arrest in November 2025, the FBI executed a search warrant at the defendant’s home and seized materials including the documents generated by the defendant’s use of the AI platform. Right, don’t do that. Though in Roy Lee’s case he posted ChatGPT’s advice on X, which also seems like it would undermine the privilege. I occasionally say around here, half-jokingly, that the stock market is a random number generator for gambling. At a micro level, a company’s stock price is obviously not a random number; it reflects the market’s expectation about the company’s future earnings, it will be higher if the company does good business things and lower if it does bad business things, and it is all analyzable in a logical deterministic way. But markets are highly competitive and pretty efficient and for most people the future path of a stock price might as well be random. Random number generators are of course important in gambling, which is why it is funny to half-joke that the stock market is a random number generator, but they are also important in computer science and cybersecurity. Some good (imperfect!) ways to generate random numbers are flipping coins, rolling dice, etc., but computers are pretty deterministic and don’t have hands, so they can’t do that. There is thus a demand for a source of random numbers that computers can use. Is it the stock market? A reader sent me this program from a “Financial Cryptography and Data Security 2026” conference last week, which includes a paper on “Ultra-high frequency random beacons from financial tick sequences” by Silvia Onofri, Andrey Shternshis, and Stefano Marmi. Here is their related paper: At an appropriate scale depending on the specific financial asset, multiple sequences pass standard randomness tests. Consequently, we propose a model-free approach for generating pseudo-random sequences from financial data, which can be leveraged for further cryptographic and other applications.
So there you go, the stock market is a useful random number generator. Regulators Plan to Relax Some Capital Proposals for Wall Street. Pushing private assets: are Europe’s investors at risk of a mis-selling scandal? Blackstone, Blue Owl Take Minority Stake in US Private Equity Firm Atlas. Morgan Stanley Cuts Jobs Across All of Its Business Lines. Deleted Tweet From Energy Secretary Sends Oil Markets on Another Wild Ride. Gulf disruption chokes sulphur flows supporting swaths of global industry. Oil Trader Pierre Andurand Made 6% Last Week in Chaotic Market. Hedge fund Caxton loses more than $600mn in Iran war fallout. Ackman’s Compensation and Other News We Learned From Pershing Square’s Filings. Bank of England fines Direct Line £10mn for overstating capital. The Tycoon Who Had a Secret Life as an Alleged Scam Kingpin. Lloyd’s of London says it will still insure ‘basically anyone’ in the Gulf. Merger blocked after lawyers Simpson Thacher miss deadline for appeal. 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! |
