Money Stuff: Quantum Bond Trading

One model that I have of financial markets is that they are a very general and efficient way for turning skill and intelligence and knowledge into money. People want money. Therefore financial markets create incentives for people to: 

• Develop skill, intelligence and knowledge, and
• Apply them to win at zero-sum-ish games of trading stocks and bonds and derivatives.

Is this socially beneficial? Well:

1. If people devote a lot of skill, intelligence and knowledge to financial markets, those markets will probably become better, in the sense of providing faster and tighter liquidity, more efficient price discovery, etc. That seems good. Normal people do not get especially excited about these social goods — ooh, the bond market is more liquid, ooh — but they seem better than the alternative. On the other hand these games often are zero-sum, and it’s plausible that a lot of the skill, intelligence and knowledge that professionals devote to financial markets are in the service of extracting money from other, normal people. (The ARK ETF IPO pop trade we discussed last week was quite clever but probably bad for ordinary investors.)
2. If people devote a lot of skill, intelligence and knowledge to financial markets, then those skilled, intelligent, knowledgeable people aren’t doing something else. The conventional something-else they could be doing is “curing cancer,” though the more relevant alternative is probably “astrophysics.” In any case, though, many smart people who are good at solving problems are solving problems like “how can we quote tighter bid/ask spreads on interest-rate futures” rather than problems like “how can we cure cancer,” because the money is better (and the feedback more immediate) with the interest-rate futures.
3. If a lot of young people think “I would like to have a lot of money,” many of them will look around and conclude that the most general and straightforward way to do that is by succeeding in financial markets. And then they will ask “how can I do that,” and the answer will be some form of “develop a lot of skill, intelligence and knowledge, preferably about machine-learning techniques though people skills are also pretty good.” And so they will go off and develop lots of skill, intelligence and knowledge. Surely that will have positive spillover effects. Not all of them will end up as hedge fund traders; some of them will enjoy astrophysics so much that they will grow up to be astrophysicists, or cancer researchers I suppose. Some of them will cure a little cancer during gardening leaves. I think a lot about the “high-frequency trading arms race,” in which a handful of proprietary trading firms spend billions of dollars to get data from Chicago to New Jersey microseconds faster than their competitors, so that they can buy shares of stock at stale prices when futures move. Seems socially useless, when you put it like that, but surely “smart people developing ways to send information around the world faster” can have other benefits.

Point 2 is arguably bad — “finance wastes our best and brightest minds” — but Point 3 is arguably offsettingly good — “finance provides incentives to create more best and brightest minds.” If the lure of hedge fund riches draws 100 extra people into physics, but 90 of them go to hedge funds and only 10 end up doing physics, that’s still a net gain for physics.

So I have one-quarter-jokingly suggested that some of the seeds of the modern artificial intelligence boom were planted by lots of people (including the founder of DeepSeek) learning machine-learning techniques because that’s a good way to get a job at a hedge fund. Now, of course, the most lucrative way to turn those skills into money is to make brief eye contact with Mark Zuckerberg, but financial markets are an always-available option that probably incentivized a lot of people to get into machine learning.

Or we have talked about the weather. Surely at some point in modern history some smart mathematically inclined young person has thought “what I really want to do with my life is study meteorology and learn to predict the weather, but how much money is there in that?” And then a hedge fund was like “tons actually, have you heard of natural gas trading?” And then she followed her (unexpectedly lucrative) passion and studied meteorology and can now advance the state of the art of predicting the weather, and those advances will first be useful in making natural gas futures prices more efficient but will eventually end up helping farmers or saving people from hurricanes or whatever. It just has to be good for humanity, collectively, in the medium term, if we get better at predicting the weather, even if the most immediately lucrative effect might be on natural gas futures pricing.

Meanwhile I am sure that quantum computing will do some very cool science-fictional stuff and solve some very deep problems, but what problems does it solve now? Obviously the first problem is providing tighter and faster bond quotes:

HSBC Holdings Plc said it’s achieved a world-first breakthrough in deploying quantum computing in financial markets, as a race intensifies among some of Wall Street’s biggest firms to embed the cutting-edge technology in their daily operations.

The London-headquartered bank said Thursday that it used International Business Machines Corp.’s most-advanced Heron quantum processor to attain a 34% improvement in predicting how likely a bond will trade at a given price. HSBC and the US technology giant applied quantum processing to an anonymized set of European bond trading data and found it could significantly enhance the efficiency of the market. …

HSBC’s trial involved examining how quantum computing could be used in over-the-counter trading markets where assets are bought and sold between two counterparties without any exchange or a broker in the middle. …

Josh Freeland, global head of algo credit trading at HSBC, said that at one point in the trial, a team of 16 physicists, machine learning and artificial intelligence experts were “working around the clock,” trying to replicate what the quantum computer had been able to do.

“If you could get something like this result everyday, that would be quite something,” said Freeland. “We spend all day looking for single-digit improvements, because when you repeat that thousands of times a day, it can really make a difference.”

In some ways this is the most boring imaginable demonstration of the power of quantum computing — people successfully traded bonds over the counter long before computers were invented, and now they are 34% more efficient — but (1) incremental progress in financial market efficiency doesn’t hurt and (2) cool technology! Is it socially beneficial for 16 physicists and AI experts to work around the clock to produce tighter bond quotes? Maybe! Will quantum computing breakthroughs free them up to spend more time doing astrophysics? Maybe!

The classical explanation for trend-following investment strategies is that investors react slowly to information. Some fundamental event occurs that significantly increases the present value of the future cash flows of some financial asset. The market initially underreacts to this event, so the asset’s price goes up only a little bit. Then more people catch on, and the price goes up a bit more. This keeps happening, until eventually people are jumping on the bandwagon and the asset’s price goes up more than is justified by the change in fundamental value.

And so a simple investing strategy is “buy stuff that has been going up, because it will probably keep going up.” This is called “momentum,” or “trend following,” and it has a pretty good track record in different asset classes. It tends to work over periods of multiple months: Stuff that has been going up for the last six months will probably go up for another month, etc.; going up for two days, or two years, is less predictive of future returns. 

If you combine that result (multiple-month trends) with the classical explanation (slow reactions to fundamental news), you get a crude intuitive model like “one important thing happens each year, and the market reacts slowly to it.” Obviously that’s not right. Lots of things happen all the time, and many of them are important to different degrees, and many of them offset each other. Still there is a human-scale plausibility to that model. Stuff does tend to persist; people who pursued one goal yesterday will probably pursue a similar goal today, and the results will compound. Ten years from now, if someone asks you “why did Nvidia go up in 2025” or “why did gold prices go up in 2025” or whatever, you will probably give a one-sentence answer referencing large-scale trends that only became apparent gradually in real time. And you’ll forget about a lot of the day-to-day details.

Again, this is an intuitive story about world events and how humans process them, but it’s not an absolute law of nature. Ten years from now, if someone asks you “why did Nvidia go down in 2025” and you answer “well there was a big AI boom that played out in stages over much of the year, but then of course the asteroid hit,” that’s two different things. [1] The slow reaction to some important information can be interrupted by the slow, or fast, reaction to some entirely unrelated and opposite information.

It doesn’t have to be an asteroid. People who pursued one goal yesterday will probably pursue a similar goal today, but who knows, man. What if every day the world resets to a brand-new state with no memory of the state the day before? What if global macroeconomic policy is made by flipping a coin each day? Well. Then no trend would be likely to persist for a month, and if it did that would be just an accident and not predictive of the next month’s trend. Trend-following wouldn’t work, not because of a failure of math but because of a … change in human nature? Change in the observed results of human nature? Change in the humans whose natures determine financial asset prices? “People just do random stuff each day” is not what is baked into traditional quantitative trading models, but that is because those models are built on historical data and regimes change.

This is all half-joking and exaggerated but here’s a Financial Times story about Man Group, which has had “three years of lacklustre performance” in “a ‘particularly unfavourable’ trading environment for the type of trend-following strategy that Man is famous for”:

Trend-following strategies, however, have faltered across the industry as markets have yo-yoed. When markets turn on the whim of the US president, it is difficult to bet on clear trends in asset prices — and to make money for investors. As AHL’s trend-following strategies have struggled, the group’s performance fees have also weakened.

Again, I’m exaggerating for effect, and it can’t literally be like “Truth Social has destroyed trend-following as an investment strategy,” but it is fun to think about. I wrote in April that “one model of capital markets is that volatility is bad and stable policy is good,” but:

Another model of capital markets is that people like gambling, so introducing some extra volatility makes markets more fun and exciting and gives people what they want. How much should you save for retirement? Should you borrow money to build a new factory? Boring! Boring! Don’t ask those questions! Ask more fun questions like “should I YOLO all my money on GameStop call options?” On this model, economic policymakers should lurch drastically from one policy to another, because that will make things more exciting for their audience. It will also get the policymakers more attention, and attention is the most valuable thing in the world.

And last month we discussed the hard times for trend-followers and I wrote:

When the world is basically stable, the stuff that went up last month will probably go up this month. When the US imposes huge tariffs one month and walks them back the next month, the stuff that went up last month will probably go down this month. The regime has changed; a new and stronger source of irrationality has been added to the market. It’s hard to make a living following trends when there are no trends.

But I also pointed out an oddity here: The long-term success of trend-following strategies is embarrassing; it seems like a sign of market inefficiency. The market, historically, has been slow to incorporate information into prices, so you could fairly reliably make money by just buying stuff that has been going up. Now it’s harder. I wrote:

In general I do not have a ton of sympathy for hedge fund managers who say things like “the market has become too irrational now, so it is hard for me to make money.” That’s not how it’s supposed to work! When the market is irrational, there are lots of opportunities for smart rational hedge fund managers; if it is hard for you to make money using your old tricks, that suggests that the market has become more rational. If I had told you in 2015 that trend following wouldn’t work in 2025, you might have said “ah, yes, as artificial intelligence becomes more sophisticated and hedge funds get more competitive, it should be harder to make money by just buying the stuff that went up last month.” But that is probably not quite what happened.

If aggregate human behavior is less predictable, then financial market prices will be less predictable, which can look like an increase in market efficiency but might instead be a decrease in, like, everything else efficiency. 

Anyway the FT reports:

In May, London-based hedge fund Man Group handed its quant teams a three-month long mission: figure out what was ailing its computer-driven hedge fund unit AHL.

Shares in the world’s largest listed hedge fund had fallen 36 per cent in the past year, key strategies were in the red and wealthy individuals were pulling their money.

The conclusion? “Nothing is broken,” said one person familiar with the review.

The algorithms are fine; it’s the market that is broken.

There is a deep weird tension in stablecoins, which I alluded to the other day. A stablecoin is simultaneously two slightly different things:

1. A stablecoin is a crypto token. It lives on the blockchain. It can be transferred permissionlessly between two parties. It is in a sense “digital cash,” the electronic equivalent of a $1 bill in your wallet. Nobody can stop you from transferring it to someone else, and once you do that there are no provisions for you to get it back. This is a foundational idea of crypto: The original Bitcoin white paper is titled “Bitcoin: A Peer-to-Peer Electronic Cash System,” and Bitcoin was designed to solve the problem that “completely non-reversible transactions are not really possible, since financial institutions cannot avoid mediating disputes.” Stablecoins apply the same basic principles, but unlike Bitcoin a stablecoin is worth a dollar, not a fluctuating value.
2. A stablecoin is an obligation of a company. [2] It lives on the balance sheet of an issuer like Circle (for USDC) or Tether (for USDT). You can transfer the stablecoin to me, and I can transfer it to someone else, and we can send it around indefinitely, but the reason it is worth $1 is because at any point someone could bring it to the issuer and say “give me back a dollar” and the issuer would do that. The transfers are not recorded on the balance sheet of the issuer (the way transfers of bank money are): If I send you a USDC, we don’t have to ask Circle to update its ledger; the transfer occurs on the blockchain. But of course Circle can see the blockchain: The blockchain is public and immutable, for one thing, but also, Circle needs to know who owns USDC if it is going to redeem them for dollars. If I go to Circle and say “here are 100 USDC, please give me $100,” I will demonstrate that I have the USDC by pointing to the blockchain. In a sense the blockchain serves as an external objective ledger of Circle’s liabilities, to which Circle has (sort of) read-only access.

Point 1 is central to crypto: irreversible, uncensorable, permissionless transfers. But Point 2 effectively means that the issuer can reverse transactions. If I send you 100 USDC, because you have done a scam and tricked me, then you irreversibly have those 100 USDC on the blockchain. But if I call up Circle and say “hey that was a scam, I want my 100 USDC back, can you help me,” Circle … could help me, no? Circle could announce “hey those 100 USDC” — which are identifiable to Circle and everyone else on an immutable public blockchain — “are not real USDC anymore, in the specific sense that if you bring us those USDC, we will not redeem them for dollars.” And so no one should accept those USDC as being worth dollars, because the link between those USDC and actual dollars has been severed. (Meanwhile Circle could make me whole by issuing me 100 more USDC, backed by the $100 of reserves that it held to back the USDC that I sent to you, which are now canceled.)

Same story if instead of a scam the transfer was a fat-finger mistake. Or if you are a terrorist, or a money launderer, or a drug trafficker, or a sanctioned person, and government authorities call Circle and ask it to reverse the transaction. Circle can’t actually reverse the transaction — the transaction occurs on the immutable blockchain, which Circle does not control — but because the transaction concerns liabilities of Circle, it could do things that have the economic effect of reversing the transaction.

My impression is that this does not particularly happen, for some combination of reasons including “it would undermine confidence in a still-nascent payments system that governments are for some reason encouraging,” “most stablecoin transactions actually occur at centralized exchanges and other custodial intermediaries (which do know-your-customer checks and try to prevent mistakes) rather than peer-to-peer on the permissionless blockchain,” and “nobody has gotten around to trying this yet.” But it does seem like the sort of thing that could happen. Also … maybe … should? Like part of the point of the modern US dollar financial system is to prevent or reverse bad transfers, where “bad transfers” is a loose regulatorily defined category including fraud, mistake and money laundering. If stablecoins are a growing parallel government-favored US dollar financial system, they should have similar mechanisms to serve that policy purpose.

Anyway the Financial Times reports:

Circle, the world’s second-biggest issuer of stablecoins, is examining ways to make it possible to reverse transactions involving its tokens, in a rare admission by a major crypto firm that it needs to take lessons from the traditional financial sector.

Circle president Heath Tarbert said a mechanism that allowed money to be refunded in cases of fraud or disputes would help the stablecoin industry’s push to become part of the financial mainstream.

“We are thinking through . . . whether or not there’s the possibility of reversibility of transactions, right, but at the same time, we want settlement finality,” Tarbert told the Financial Times.

Right, yes, the traditional financial system has spent many years trying to find the right balance between “settlement finality” and “obviously we have to reverse some bad transactions.” (Just ask Citigroup Inc.) Crypto’s founding ethos was to just not do that, to turn the dial all the way to “settlement finality” and not worry too much about mistakes or money laundering. Which is fine for a rebellious new financial system, but bad for a “push to become part of the financial mainstream.”

Elsewhere in “Bitcoin: A Peer-to-Peer Electronic Cash System,” here is a fun Kansas City Fed research briefing finding that, while no one ever really used crypto to buy a sandwich, the incidence of using crypto to buy a sandwich is now decreasing:

The share of U.S. consumers who report using cryptocurrency for payments—purchases,  money transfers, or both—has been very small and has declined slightly in recent years. The light blue line in Chart 1 shows that this share declined from nearly 3 percent in 2021 and 2022 to less than 2 percent in 2023 and 2024.

I suppose it is often the case that a technology is introduced with great hype to solve one particular problem, and then turns out to be more or less irrelevant to that problem but quite useful for some other problem. Here Bitcoin was introduced to allow for peer-to-peer permissionless payments, for which it has turned out to be a dud. But Bitcoin is very good at an unrelated thing, which is “going up in price.” (You can see why that would make it bad as a payments medium!) So it has caught on and become widely adopted, even though its original premise was kind of wrong.

Incidentally the point of that Kansas City Fed research note appears to be to express skepticism about stablecoins as a payments mechanism, in the face of growing optimism elsewhere. I don’t know if that follows. Crypto is not widely used for payments for a number of reasons, including (1) prices are volatile, (2) you can’t reverse payments and (3) you can’t send crypto from your bank’s app. Stablecoins fix the first problem and, with a certain amount of pushing to become part of the financial mainstream, could solve the other two as well.

We talked last week about a cool ETF trade. ARK Invest runs some actively managed exchange-traded funds, and those funds sometimes get allocations in hot initial public offerings. If you buy shares in an ARK ETF the day before a hot IPO, and sell it the day after, you can participate in the IPO pop: Those IPO shares will trade up, the ARK ETF’s value will increase, and you will share in that value. Of course the ARK ETF also has a bunch of other stocks, and maybe those will go down. Ideally you would do this trade on a hedged basis: Buy the ARK ETF, sell the underlying stocks, and participate only in the IPO pop. The mechanically simple way to do that is to borrow all the underlying stocks, deliver them to ARK to create new ETF shares, wait until the IPO pop, and then redeem your ETF shares by handing them back to ARK, getting back the underlying stocks and returning them to your stock lenders. 

Taking that trade one step further, I asked: What about not borrowing the underlying shares? Like, you go to ARK, you say “I’ll give you $1 billion of underlying stocks for $1 billion of ETF shares,” ARK says “sure we’ll settle that trade tomorrow,” tomorrow the IPO opens and pops and you call up ARK and say “I’ll give you back $1.05 billion of ETF shares for $1.05 billion of underlying stocks,” ARK says “sure we’ll settle that trade tomorrow,” you’re like “actually it precisely offsets our trade from yesterday so let’s just rip them both up,” ARK is like “uh sure I guess that works,” and all that has happened is that you got the IPO pop without any other cash or securities or anything else changing hands.

This is probably too cute, I said, though it is reminiscent of certain sorts of crypto trades where you can borrow money, do arbitrages, make a profit and return the money all in a single unit of blockchain time, with each transaction contingent on all of the others. If a trade works, you can do it without any financing (it retrospectively finances itself); if it doesn’t work, you can just not do it.

There is an old, popular, quite illegal form of this in traditional financial markets. It’s called “free-riding.” You write yourself a check for $1 million and deposit it in a brokerage account. Your broker is like “okay we got your check, what are you going to buy with the $1 million?” You are like “obviously zero-day call options on GameStop bro.” You buy the options. If they pay off, you have $5 million in your account; if they don’t, you have $0. Your check does not clear (because you have no money). Your broker calls you to ask for the $1 million. If the options paid off, you are like “oops minor cash-flow glitch, sorry, just take the $1 million out of my winnings.” (Now there is $4 million in your account.) If the options did not pay off, you do not answer the phone. (Now there is negative $1 million in your account.) The trade was self-financing if it worked and, you know, something else if it didn’t.

There are various ways for this to go wrong, with the main ones being:

1. Your broker will probably do some due diligence and might not give you access to the $1 million as soon as you deposit your fake check; and/or
2. Jail.

Here is a US Securities and Exchange Commission enforcement action [3]  against a guy named Aaron Freeman, who is accused of free-riding in various brokerage accounts. For a while he allegedly tried the traditional approach (writing himself checks), but it never worked, because the brokers never let him trade before the checks cleared (they didn’t). Then he allegedly moved on to the slightly more sophisticated approach of writing checks to his aunts, opening accounts in their names, and doing trades in their accounts. This occasionally worked, to the tune of $889,087.04 in purchases and $5,463.26 of losses (and no gains). It is illegal, but amusing and impressive, to free-ride and do good trades, or even high-risk high-reward trades that don’t work out. Consistently losing 1% on your trades sort of misses the point.

First Brands-Linked Financing Units File for Bankruptcy. Private credit CLOs. Corporate AMT. OpenAI and Databricks Strike $100 Million Deal to Sell AI Agents. The Investment Bankers Winning at the AI Deal Game. UBS emergency plan is not ‘executable’, says Swiss regulator. Germany’s Merz backs using frozen Russian assets for Ukraine. Brunello Cucinelli shares suspended as short seller makes new Russia claims. The Tech Fashion Darling Accused of Swindling Investors Out of $300 Million. There Are More Robots Working in China Than the Rest of the World Combined.

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!