Money Stuff: Bond Trading Gets Cheaper

One way to think about corporate finance is: There is a company that needs money. It goes to an investment bank to raise money. The bank is like “you can sell $1 billion of paper to investors, and then you will have money.” The company does this. It gets $1 billion of money, the investors get $1 billion of “paper” — that is, financial instruments, stocks or bonds or syndicated loans or hybrids or some other weirder thing — and the bank gets a fee. The company is happy because it has money; the investors are happy because they have put their money to work in a useful way for some promised return; the bank is happy because it gets a fee.

Another way to think about corporate finance is: There is a casino for trading financial instruments, paper, stocks and bonds and whatever else you’ve got. [1]  The casino needs paper to trade: Some paper is always going away (maturing, getting cashed out in mergers, etc.), plus the casino always wants to expand, so there is constant need for more paper. The investment banks run the casino; they take a cut of every bet. The more bets there are, the more money they make. They want more paper. Sometimes a company comes to a bank because it needs money; sometimes a bank comes to a company because it wants paper. “You can get great financing for your AI data center right now, please, give us some paper,” the bank says, and the company is like “well we do need money for the AI data center so sure.” So the company issues paper to the bank, which sells it to investors; the company gets money and the investors get paper and the bank gets a fee; everything happens just as in the previous paragraph, but none of this is the point. The point is not the corporate financing transaction — company sells paper to investors for money — but rather what comes next, the trading. What happens next is that the paper is constantly traded back and forth in the casino, and the people in the casino are happy, and the banks get a cut of that revenue.

One important thing to realize is that both of these models are true, at least much of the time. (In some rough theoretical sense things like private credit might not be like this, because private credit is traditionally a buy-and-hold sort of investment that is not traded constantly in the casino. But this is only approximately true, and we have talked about various forms of private credit liquidity, at the loan and fund-stake and retail-product level, because the logic of the casino is powerful.) The first model is the simple traditional one: Companies really do need money to build factories and AI data centers and whatnot, and the financial markets provide that money. The social purpose of raising money for companies is pretty obvious.

But the second model, the casino, is also true. The casino serves several purposes:

1. It’s fun, a place for smart people to match wits against each other, with the best wits winning. It is a general-purpose venue for rewarding correct understanding of the world, which (1) seems like it might be socially useful and (2) is certainly nice for people with correct understandings of the world, who can get quite rich in the casino.
2. It makes it easier for companies to raise money. The first model, in the first paragraph, could work without a casino: Companies could raise money by selling stocks and bonds to investors, and the investors could hold the stocks forever and the bonds until they mature, and successful companies could reward their investors with interest and dividends. But it works better with trading. Trading markets allow for price discovery, which makes the pricing of paper more efficient, so people can more confidently give money to companies because they know that the price is probably right. And they allow for liquidity, so people can more confidently give money to companies, knowing that if they need the money back they’ll be able to sell their paper. Because the investors are happier, the cost of financing goes down, and more companies can build more stuff because the casino exists.

The difference in the two models is in what the point is. In the first model, the point of the casino is to help raise money for companies. In the second, the point of corporate finance is to create trading volume and volatility for the casino. Which model you will emphasize will depend on where you sit. If you are a high-frequency stock trading firm, for instance, you live entirely in the casino and never (directly) help companies raise money. The supply of stocks for you to trade is entirely exogenous: The stocks just show up on your computer, and you make money trading them. If you are a debt capital markets investment banker, you live almost entirely in the money-raising world; your job is to raise money for companies, though to do the job you need to know a little about how the casino works.

Also the other difference in the models is in how the banks make money. If the paradigmatic way that a bank makes money is fees from companies, that’s the first model; if the paradigmatic way that a bank makes money is trading revenue, that’s the second. If banks’ corporate finance activities are where the action is, and they provide some low-margin trading services to help customers get in and out of investments, that’s the first model. If banks’ trading activities are where the action is, and they help companies issue paper as a loss leader to supply the casino with bets, then that’s the second. 

In practice neither corporate finance nor trading is really a loss leader for investment banks, and they trade off primacy with market conditions. (Broadly speaking volatility is good for trading revenue and bad for capital-raising revenue, while calm is good for capital-raising and bad for trading, but you can find ways to make money either way in any conditions.) You might have some sort of very-long-run efficiency view like “in the long run, markets will get more efficient and it will be hard for banks to make a ton of money in the casino, so in the long run the corporate finance function of markets will dominate,” but that does not seem to be especially true. (The famous paper is by Thomas Philippon.) 

Anyway here’s an IFR story about electronic credit trading:

Banks’ corporate bond trading revenues have slumped to their lowest level in more than a decade, according to benchmarking firm BCG Expand, in the wake of a rapid electronification of these markets that is eroding the profitability of credit trading desks.

The top 10 banks generated US$4.3bn of revenues in North America “flow” credit trading in 2025, BCG Expand said, less than half their 2021 total. That comes amid a structural shift in the once human-centric world of corporate credit towards an equities-style trading paradigm in which machines process a larger share of transactions.

This electronic revolution has sent trading volumes soaring to record highs in the US$11.5trn US corporate bond market as investors have found it easier to shuffle positions. But it has also compressed trading margins to their skinniest levels in history, making it harder than ever for banks to turn a profit despite this surge in activity.

Tough for those banks, but this seems like an obviously good story for market efficiency: Paying less to trade bonds is good. Also, is it bad for the banks? It sort of depends on what their job is:

The pure market-making focus of firms like Citadel Securities represents a different proposition to traditional banks, many of which believe they still need bond trading to support their debt underwriting activities. ...

“The [bid-offer] compression is very real, but it’s not a reason to panic. We can withstand it,” said the head of credit at a major bank, who predicted that spreads will keep tightening to the point where today's markets feel like a “golden age” for bid-offer.

“The reality is we need to have a secondary trading business. The top houses have the technology and the personnel to keep sharpening it and making it more efficient. But we’ve also got the capital markets piece, we have the derivatives piece, so we have many ways to access the client wallet.”

“We need to have a secondary trading business” to do corporate finance, is one way for banks to think about things, but sometimes some bankers think the opposite.

One theory of crypto is that there will only ever be 21 million Bitcoins. Bitcoin is scarce in a way that traditional financial assets aren’t: Bitcoin’s supply is set permanently and transparently by code, so no one can ever create any more, while banks can create dollars whenever they please. A dollar is just an entry on a list at a bank, and a bank can give you more dollars just by adding numbers to your list. It’s not quite that simple — when it creates dollars, the bank will create an offsetting asset (a loan) and liability (a deposit) for itself, so it doesn’t get magically richer, and its capital requirements will limit its ability to create too many dollars — but broadly speaking that’s true. This is called “fractional reserve banking,” and part of the reason that Bitcoin was invented is that a lot of people don’t like it. Feels fishy, that banks can create dollars. Better to just have 21 million Bitcoins forever.

But then Bitcoin became hugely popular and integrated into the traditional financial system, with the result that now there are a lot more ways to create, you know, fractional reserve crypto. In particular, crypto futures exchanges can effectively create Bitcoin supply: If you want to bet that the price of Bitcoin will go up, and I want to bet that it will go down, an exchange will happily match our bets without either of us owning any actual Bitcoin. Economically and psychologically, you now own some Bitcoin — you are exposed to its price moves, etc. — but you don’t own any of the 21 million Bitcoins. We just created some new Bitcoins for you.

But there is also a dumber and more literal version: You have an account at a crypto exchange, the crypto exchange has a database listing how much Bitcoin its customers own, and if it increases the number next to your name then … do you own more Bitcoin? Not really: Strictly speaking, you can only own as many Bitcoin as the exchange has backing your account, and if the exchange is just making up some Bitcoin then eventually people will notice, someone will get in trouble and the accounts will get fixed. But if you think you own more Bitcoin, and the exchange thinks you own more Bitcoin, and your trading counterparties think you own more Bitcoin, then maybe, for a while, you do?

Anyway here’s a crypto exchange that accidentally increased the numbers next to some names:

The hundreds of prize payouts were mostly just a few bucks each, part of a promotional campaign by a South Korean cryptocurrency exchange. The total reward pot: 620,000 Korean won, or about $425.

Then came a colossal mistake. A staffer for Bithumb, South Korea’s No. 2 crypto exchange, didn’t distribute 620,000 Korean won.

Rather, the prizes, due to an input error, emerged in a different currency: 620,000 bitcoins, valued at more than $40 billion. 

That meant a winner who should have received a sum of 2,000 won—enough to buy a cheap cup of coffee—reaped, at least momentarily, more than $120 million in bitcoins.

Enough recipients sought to sell or withdraw bitcoin that the market sank 17%, before Bithumb halted transactions after roughly 30 minutes. Those affected included investors who had held bitcoin before the botched giveaway. The losses totaled about $685,000, Bithumb says.

The company has since said it has reversed the transactions or had recipients voluntarily return more than 99% of the misdistributed bitcoins. But Bithumb is still trying to convince users who during the brief window of trading managed to offload more than 100 bitcoins, valued at roughly $9 million, to give back the equivalent funds. 

Look, if a crypto exchange accidentally gave me $9 million, I would fight to the death to keep it, just on principle. (The principle is “hahahahaha crypto exchange.”) Also:

South Korean lawmakers have raised concern over the so-called “phantom coin” issue: How Bithumb was able to distribute 620,000 bitcoins when its own stockpile appeared to have numbered closer to 50,000. ...

Cryptocurrency exchanges, by local law, can’t allow trades exceeding the actual supply of coins held in their digital vault. But that doesn’t appear to be the case with the recent Bithumb activity, representing a “catastrophic failure of internal controls,” said Lee Jung-soo, who advises the South Korean government on digital-asset policy.

Yeah I mean we can await the results of the investigation, but here I will tell you how I think “Bithumb was able to distribute 620,000 Bitcoins when its own stockpile appeared to have numbered closer to 50,000,” which is that (1) it had a computer that lists how many Bitcoin its customers have and (2) it typed bigger numbers into the computer. Obviously it shouldn’t have done that — it was a “catastrophic failure of internal controls” — but it did do that. And to the extent “a Bitcoin” means “an entry on the electronic list maintained by Bithumb,” then some people really got some extra Bitcoin. It doesn’t mean that in quite the same way that “a dollar” means “an entry on the electronic list maintained by a bank,” but it’s closer than some people would like.

We talked once about a (harmless) fat finger error at Paxos, the stablecoin company, which minted and quickly burned $300 trillion of its PYUSD stablecoin. That $300 trillion of PYUSD was supposed to be backed by $300 trillion of dollar assets, because that’s how stablecoins work, but it wasn’t backed by $300 trillion of dollar assets, because (1) obviously Paxos doesn’t have $300 trillion and (2) someone put the wrong number into the computer. I wrote:

In a good financial thriller, the person who accidentally created that $300 trillion would have sent it all to herself, sold it as quickly as possible and cleared, you know, at least enough to buy a first-class ticket to a non-extradition country. (The market price for $300 trillion of fat-fingered PYUSD is nowhere close to $300 trillion, but it’s not zero either, and you don’t need to get particularly close to $300 trillion to have a very good day.) But in the real world they just deleted the extra $300 trillion and went about their day.

Bithumb was slower, and some people apparently did make off with the money.

You know what’s a good piece of paper? A contract that costs $1,000 up front and pays you $61.25 per year forever. The two big problems in saving for the future are:

1. You don’t know how much money you will need each year, and
2. You don’t know how much money you will get each year.

This contract solves the second problem completely: You know you will get exactly $61.25 each year. (If that’s not enough, buy more than one of this contract. If you buy 1,000 of them, for $1 million, then you’ll get exactly $61,250 each year.) Most other financial instruments don’t do that. If you buy stock in Nvidia, it might go up or down. If you buy a 30-year US Treasury bond for $1,000, it will pay you about $48 per year for 30 years and $1,000 in 30 years, but then what? You’ll have to reinvest that $1,000, and maybe when you do yields will be lower and you’ll only get $25 per year. This contract — the $61.25 perpetuity — avoids reinvestment risk. You just get the same exact amount, guaranteed, forever.

Of course this contract does nothing to solve the first problem. What if there’s massive inflation and, in 30 years, $61.25 doesn’t buy what it used to? You probably wouldn’t want to put all of your eggs in this basket; you’d want to own some stuff with payoffs that vary with economic growth or interest rates or inflation. But if you have some fixed expenses, putting some of your money into an investment that pays a fixed amount forever might be a good deal. It might make life a little simpler. And if you are a pension fund, with lots of long-term fixed liabilities, this might be especially appealing.

The other problem with this, though, is how do you know you’ll get $61.25 per year forever? Sure the contract says that, but forever is a long time and things can change. If I promise you $61.25 per year forever, you might doubt my ability to keep paying you forever. You’d need to be really sure of my ability to pay, before you’d give me $1,000 for this deal. 

And so historically the main issuers of contracts like this were governments, [2]  which were considered pretty creditworthy; British consols had about a 250-year run. But if you look around in 2026 and think “who can I trust to pay me $61.25 per year forever?” I guess you could come up with worse answers than Google:

Alphabet Inc. raised almost $32 billion in debt in less than 24 hours, showing the enormous funding needs of tech giants competing to build out their artificial intelligence capabilities — and the huge appetite from credit markets to fund them.

The Google parent sold sterling and Swiss franc-denominated offerings, both of which were the biggest-ever corporate bond sales in their respective markets. Those deals followed Monday’s $20 billion dollar debt sale. The sterling issue included an ultra-rare 100-year note — the first sale with such an extreme maturity by a technology firm since the dot-com era, according to data compiled by Bloomberg. ...

Alphabet’s 100-year note was the first sale with such an extreme maturity by a technology firm since Motorola sold this type of debt in 1997, according to data compiled by Bloomberg. The market for 100-year bonds is dominated by governments and institutions like universities. For corporates, potential acquisitions, outdated business models and technological obsolescence make such deals a rarity. …

Still, demand from UK pension funds and insurers has made sterling a go-to market for issuers seeking longer-dated funding, and investors turned out in force for the century bond.

The 100-year has a 6.125% coupon and got strong demand. Bloomberg tells me that it has a duration of about 16.8 years, about a year longer than a 30-year US Treasury bond, so if you want to bet that long-term interest rates will go down this is a way to do it. But if you just want to get £61.25 per year forever, or at least for 100 years, or at least for as long as Google is around, this is a way to do that too.

Here is an article with the headline “California to ban blackjack-style games from cardrooms by April,” which seems to capture the spirit of what is going on: Blackjack has long been illegal under California state law, some cardrooms have operated “blackjack-style games” that arguably get around the spirit of that law, Native American tribal casinos object to this arguably-illegal competition, and now the state is tightening its regulations to ban those games.

Here, though, is a much funnier article with the headline “California imposes sweeping changes to blackjack in cardrooms,” which describes what sorts of “blackjack-style games” will still be allowed:

Under the new rules, cardroom players can’t “bust.” Instead, the winner of a hand is determined by which player has the most points toward the target point count, when compared to the player-dealer.

Also, receiving an ace card along with a 10, jack, queen or king no longer means an automatic win.

There is a certain participation-trophy vibe, “blackjack but you can’t bust.” It would hurt people’s feelings to bust!

Two quick points here: 

1. If you are a commodity futures exchange (or “designated contract market,” DCM) registered with the US Commodity Futures Trading Commission, you probably mainly operate an electronic exchange, but technically you can have a physical trading floor too. Members can come to the trading floor to trade their commodity contracts, instead of doing it electronically. There is not too much demand for that these days, as a matter of financial-market efficiency, but there is a lot of nostalgia for it among former floor traders. It would be a bit simplistic to say “the trading floor was really fun and people just liked going there,” but one does get that sense.
2. The CFTC regulates prediction markets, or “event contract” markets, which are essentially federally regulated gambling sites. Federal commodity regulation preempts state law, so these prediction markets can probably (it is still being litigated) ignore state regulation, offer gambling in states where it is otherwise illegal, let 18-year-olds bet and otherwise get favorable regulatory treatment. Currently they offer mostly sports gambling, and there is some regulatory reason to think that an event contract like “I will get closer to 21 than the dealer without going over” would not be a legal commodity futures contract, but I don’t know, man; who would stop it?

The synthesis of these points is that Kalshi should open some “trading floors” in California to let people “trade” blackjack “contracts.” You know, at felt-topped tables, with free drinks. This is not legal or gambling advice and is mostly a joke but also I feel like it will happen in the next two years and then I’ll link back to this column and be like “see?” Everything is so dumb but there’s no reason to think it won’t get dumber.

Paramount Sweetens Warner Bros. Bid Terms to Woo Investors. Trump to Repeal Landmark Climate Finding in Huge Regulatory Rollback. Atlanta Is Challenging Big Corporate Landlords Without Waiting on Trump. Elon Musk’s Go-To Banker Is Back in Action for the SpaceX IPO. Don’t let Elon Musk monopolize space compute. US Labor Board Gives Up Oversight of SpaceX in Victory for Musk. Inside OpenAI’s Decision to Kill the AI Model That People Loved Too Much. How One Prediction-Market Trader Played the Super Bowl—and Lost $100,000. No, Jeffrey Epstein didn’t trigger the 2008 subprime mxreltdown.

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!