Trading floors used to be natural monopolies. Some brokers get together and start trading stocks with each other under one particular tree, or in one particular coffeehouse, and they keep coming back every day; over time they get themselves a dedicated exchange building with a comfortable trading floor where they can meet and trade stocks. If you are a broker with stock to sell, you will naturally go to the exchange with the most potential buyers; if you are a broker who wants to buy stock, you will naturally go to the exchange with the most sellers. If you dislike something about the exchange — if you think the room is too drafty or the other brokers are unfriendly or the rules discriminate against you — well, tough. You could go across the street and start a new exchange, but that’s hard to do, because everyone wants to be where everyone is. If it’s just you across the street, who will want to trade with you?

In an important sense, the rise of electronic trading did not change this. People still want to be where the liquidity is; people still want to send their orders to the electronic venue where all the other orders are. But in an important sense it did. Old-school trading floors were floors, where brokers stood, and a broker could only stand in one place at a time. Electronic trading venues are just computer servers and communications protocols, and you can more or less connect your computer to as many other computers as you want. If one big exchange almost always has the most orders and the best prices, you will do most of your trading there, but if another small exchange occasionally has better prices, you can send some orders there when its prices are good. You don’t have to choose to stand only in the best location; you can stand everywhere at once. An exchange with a good product or a good idea can win market share gradually; incumbency is not quite as powerful an advantage as it once was.

When trading floors were (1) natural monopolies and (2) literal floors, there were only a limited number of spots on the floor, and those spots were hugely valuable. (The spots were generally called “seats,” even though the brokers rarely sat down.) People who did not have spots on the trading floor had to do their trading through people who did; the people who did have spots had the first crack at orders. The exchanges were not, originally, businesses; they were membership organizations run by and for their members, the people with spots on the floor. The seats became valuable assets: Members could pass them on to their colleagues or children, or sell them for large amounts of money.

Eventually (1) the exchanges mostly became for-profit businesses and went public and (2) trading moved to electronic venues. What does that mean for the value of those seats? There is some argument that the answer is “the seats became much more valuable”: Electronic trading made the big exchanges bigger, increased volumes, and made it more important for global firms to connect to them. If only 100 people can trade on an exchange, then owning one of those 100 slots is more valuable as the exchange gets bigger and more central.

But there is also an argument that the answer is “the seats became worthless”: Electronic trading eliminated the constraint on how many people could trade on the exchange, so if more people wanted spots then more spots could just be created. And electronic trading also undermined the monopoly status of exchanges, so if exchanges didn’t create more spots, they would lose business to more accommodating venues. If only 100 people can trade on an exchange, in 2025, everyone else will trade elsewhere.

This week a trial started in Chicago to figure out the answer. The Wall Street Journal reports:

On Monday, a trial began in a long-running class-action lawsuit filed by traders who say that exchange giant CME Group duped them out of the privileges they held as members of the city’s once-elite community of floor traders.

The plaintiffs, who estimate that they are owed about $2 billion in damages plus interest, say the company broke its promises to them when it opened a data center for electronic trading that effectively doomed the old trading floors. CME has called the lawsuit baseless. ...

CME transformed itself into a for-profit company in 2000. To win members’ support for the transaction, the exchange gave them two kinds of shares. “A shares” became CME’s publicly traded stock. “B shares” were equivalent to the old memberships, allowing holders to benefit from “trading floor rights and privileges,” the exchange said in a 2000 prospectus. ...

But the price of the old-style memberships has stagnated since then. … The key factor behind the divergence, according to the plaintiffs, was the 2010 launch of CME’s data center in the Chicago suburb of Aurora, Ill. ...

In effect, Aurora became CME’s new trading floor—but the members didn’t get the trading-floor rights that they had been promised, the plaintiffs argue. This was “a betrayal of the basic bargain,” they wrote in an April court filing.

CME denies breaking any promises. It says B-share owners were only guaranteed the right to access open-outcry trading floors—not any sort of electronic floor.

Yes if “trading floor rights and privileges” means “the right to stand on the trading floor,” then in 2025 that is pretty much worthless. If it means “the exclusive, transferable right to trade on the exchange,” then that’s pretty valuable. But in 2025, why would an exchange give out exclusive transferable rights to trade?

Though, I should say, there are still some physical limitations. The article notes that “high-frequency trading firms pay for the right to place their servers next to CME’s computer systems in Aurora, allowing them to execute trades in millionths of a second and obtain data from the exchange as quickly as possible.” There is no real limit on how many people can connect their computers to the exchange’s computers, but there is a physical limit on how many people can put their computers right next to the exchange’s computers, which matters for very high-speed traders. It would be a little funny if old-school floor trading seats transformed into modern colocation rights.

Leveraged exchange-traded funds are popular. Sometimes, you have $100 and you would like to bet $200 on MicroStrategy or Tesla stock, or on the S&P 500 Index. The classic way to do that is to borrow another $100 so you can buy $200 worth of stock, but that might be hard, or you might not get great terms on that loan. So the modern approach is that an ETF provider will do it for you: You buy $100 worth of the ETF, the ETF borrows $100 and buys $200 worth of stock, [1]  and you get 2 times the upside (and downside) of the stock. There are 3x ETFs too.

For symmetry, you might wonder about ETFs that are the opposite of levered, half-levered ETFs. You have $100, and you would like to bet $50 on MicroStrategy. Is there a product that will do that for you? Well, I mean. That one is pretty easy to do yourself. What you do is: Buy $50 of MicroStrategy. You can just do that. And then you still have the other $50. You can do whatever you want with it. You can put it under your mattress, or in a bank or money market fund, or use it to buy a different stock, or dinner for that matter. Could an ETF do that for you? It is just about imaginable that an ETF could get a better interest rate on the cash than you could, so “buy $100 worth of an ETF, the ETF puts $50 in the bank and spends $50 on stock, you get one-half of the upside (and downside) of the stock” could be a good deal for you, even after fees. Seems marginal, though. And obviously it doesn’t scratch the gambling/adrenaline itch that a 2x ETF does.

Still I wish someone would try it. The levered ETFs are often on already-pretty-volatile stocks, and there might be an audience for those stocks but with less volatility. I am not saying that I want this, or that anyone should want it, only that somebody might. Somewhere out there there is an investor who wants to invest in MicroStrategy, but at half the volatility, and “just buy less stock” will not satisfy her in the way that “We Are Launching an Innovative Half-Levered MicroStrategy ETF” would. There is an untapped market; someone should make this.

But no one will, and instead what you get are various flavored of “hedged” or “buffered” ETFs that are approximately half-levered ETFs, but in more complicated ways. Instead of “you get half the gains if it goes up and half the losses if it goes down,” there is some nonlinear payoff function, “you have a floor on your losses and you get half the gains above a 2% hurdle” or whatever. The point is the same, though: They give you diluted exposure to the underlying thing. [2] We talked last month about some AQR Capital Management Research finding that these things mostly underperform just buying less stock, but who has the discipline to just buy less stock? There is demand for buying all the stock, but with guardrails.

Anyway Bloomberg’s Isabelle Lee reports:

Cathie Wood’s Ark Investment Management LLC is moving into the fast-growing market for buffer ETFs designed to limit equity losses, at a time when her tech funds continue to bleed assets.

The investment firm last week submitted paperwork to launch four exchange-traded funds that seek to protect investors against modest losses in the equity market while still offering upside. The idea is simple: When stocks drop, the ETFs cushion the fall, shielding investors from some losses. When the market rises, they deliver gains, though these are capped, too.

In Ark’s case, each fund has a 12-month period that resets on a rolling basis, starting in January, April, July and October, according to filings with the US Securities and Exchange Commission. If approved, the ETFs will aim to limit losses to about 50% of any drop in the share price of the $6.8 billion ARK Innovation ETF (ticker ARKK), while allowing full participation in gains above a set hurdle rate of around 5%.

“These are like Diet Ark. Investors are probably telling them ‘we like the taste and the kick but would like to trim the caffeine and sugar,’” said Eric Balchunas, senior ETF analyst at Bloomberg Intelligence. “So maybe they would like to give up a touch of upside for some sleep at night for the downside.”

Yeah I mean if you like your exposure to Cathie Wood but it’s keeping you awake at night, have you considered selling some of it? But this is cool too.

Meanwhile, what if you want to own a broad index of stocks, but never get dividends? For the most part, owning stocks is pretty tax-efficient: If your stocks go up, you don’t have to pay taxes on the gains until you sell them, which could be a long time. But if the stocks pay dividends, those dividends are taxable income to you now. If you would prefer not to pay taxes, dividends are inefficient. And a lot of stocks pay dividends, so if you own a broad index, you’ll get some dividends. 

You can’t really manufacture a solution yourself. A naive approach might be that, any time a company announces a dividend, you sell the stock a day before the ex-dividend date and buy it back the day after, so you don’t get the dividend. [3] But of course this is not what you want: You’re selling the stock! That triggers tax realization! The goal is to (1) never get dividends and (2) also never sell the stock.

An ETF can do that for you, because the most important ETF technology is never selling stocks. Stock ETFs normally do in-kind creations and redemptions: If you want to put money into a mutual fund, you give the fund money and it uses the money to buy stocks, but if you want to put money into an ETF, you instead buy shares of the ETF from a market maker, who in turn goes out and buys shares of the underlying stocks in the ETF and then delivers those stocks to the ETF in exchange for new shares of the ETF. [4]  And vice versa when you withdraw money. The ETF never handles money; it never sells shares for cash; it only swaps its own shares for the assets it holds. And the tax rules allow ETFs not to recognize taxable gains on these transactions.

Many modern ETFs are set up to, essentially, never trade (or at least never sell appreciated stocks). If an ETF wants to sell Stock X and buy Stock Y, what it will actually do is find a market maker to (1) deliver ETF shares in exchange for a special redemption basket of Stock X and then (2) deliver Stock Y as a special creation basket in exchange for new ETF shares. The ETF’s buying and selling are both in-kind, paid for in ETF shares. This is called a “heartbeat,” and we discussed it in more detail a few years ago, when it was mildly controversial.

So if you want to own all the stocks but never get a taxable dividend, you really can’t do that on your own, but an ETF would be happy to do it for you. Bloomberg’s Vildana Hajric reports:

Roundhill Investments plans to launch the S&P 500 No Dividend Target exchange-traded fund on July 10 with the ticker XDIV. Its ambition is simple but strategic: track the performance of the famous benchmark while dodging its payouts. The fund will sell holdings just before their dividend dates — steering income away from ETF shareholders and, in the process, away from their tax bills. …

“The ability of ETFs to sidestep capital gains isn’t just a technical quirk anymore — it’s a core selling point, and issuers are leaning into this edge,” said Athanasios Psarofagis, ETF analyst at Bloomberg Intelligence.

Right, “if you buy our thing you don’t have to pay taxes” is actually a pretty good selling point.

We talked yesterday about the Securities and Exchange Board of India’s case against Jane Street Group LLC for allegedly manipulating the Indian options market. Indian cash-settled index options trade far more liquidly than the underlying shares, which creates an opportunity for manipulation. SEBI alleges that Jane Street would (1) buy a bunch of shares, pushing up the index prices, (2) sell a bunch of options at the artificially high prices, [5]  (3) sell back the shares, pushing down the index prices and (4) buy back the options (or let them expire) at the artificially lowered prices.

As I wrote: That’s a very intuitive story! If you have a derivatives market that trades much more than the underlying market (on the day SEBI focuses on, the options traded 353 times as much volume as the underlying stocks), a trade like that really could work. That said, there is a problem with SEBI’s story. On the day that it focuses on — Jan. 17, 2024 — Jane Street’s initial buying arguably did push up the index price, but the options prices went down. The options opened the day trading at a large premium to the underlying index, so it is pretty natural that Jane Street simultaneously bought stock and sold options: The options were expensive and the stock was cheap. I wrote:

This is a very different story from the one SEBI tells. This does not look like manipulation; it looks like arbitrage. This is: Jane Street came in one Friday morning and noticed that Indian retail traders were buying options on the Nifty Bank index at much higher prices than where the index was actually trading. So Jane Street got to work doing what it does: It sold options to retail traders who wanted them, and bought the underlying stocks to hedge, until the arbitrage closed. (More strictly, it net sold call options and net bought put options, giving retail traders long exposure to the index.) At the beginning of SEBI’s window, the options were trading at a pretty wide 1.6% premium to the underlying index; at the end, they were trading at basically the “correct” levels.

But I should say, there are still two problems with the Jane Street story:

1. Jane Street sold about 7 times as much stock via options as it bought in the cash market. This is not a pure index arbitrage trade; it is also a (larger) directional bet that the (option-implied) market would go down.
2. At the unwind, Jane Street’s trading probably did push the market down: In the afternoon, it would sell back the actual stock that it bought in the morning, pushing prices lower. The options that it sold in the morning would also unwind (either it would buy them back, or they would expire for cash), but the whole point of the SEBI case is that the stock was less liquid than the options, so Jane Street’s afternoon stock trading would make it more money on its options unwind than it would cost on the cash trading.

One interpretation you could have here is “that was all intentional: Jane Street intentionally went net short the market in the morning, knowing that it would manipulate the closing price down in the afternoon.” I find that it is difficult, from the outside, to distinguish legitimate arbitrage/hedging trades from market manipulation just based on trading activity. As I wrote:

In many cases, “legitimately doing an arbitrage trade” and “trading in one market to manipulate prices in another market” look pretty similar. Either way, you are trading the opposite way, buying stock in the stock market and selling it in the options market or vice versa. The difference can be subtle, and I often joke that the difference between legitimate trading and manipulation is whether you send your colleagues an email saying “lol I sure manipulated that market.”

Other people seem to find this easier. Here is Alexander Gerko:

As input we have the following "index arb" strategy (all numbers approximate):

Leg1 is 10x smaller than Leg2.
Market in which Leg1 is trading is 100x smaller than market for Leg2.
Leg1 is consistently losing money, Leg2 is making astronomical amounts of money.
 
What is going on here? Why is Leg1 losing money? What's the point of Leg1, it barely hedges any risk and loses a lot of money, why not trade just Leg2? …

Let's for simplicity assume it's only one camel vs retail crowd, no other competitors. If Leg1 and Leg2 open with a gap between them you can try buying Leg1 and selling Leg2. They will of course converge to the same point, but *which* point? Almost whichever you want, if you have enough capital! The easiest way to get it to where you want is to trade a lot (vs market volume) in the leg that is less liquid. By the time you closed the arb, from the perspective of an external observer everything looks "normal" - arb is closed, market is efficient, thank you, kindly camel. In reality of course the point the market converged to is not equilibrium of some sort, you massively shifted illiquid Leg1 (by tens of basis points) through market impact of your trading. Note that it does not mean that Leg1 price went up Vs open, only that it went up Vs where it would have been without you. During unwind of the Leg1 later in the day you revert those tens of basis point of market impact, monetizing it on Leg2.

I think the counterargument is of the form “Jane Street thought the stock was underpriced and the options were overpriced, so it bought stock and sold options, but it thought the options were more overpriced than the stock was underpriced, and it could trade more options than stock, so it did.” But that is not a fully satisfying counterargument, and “Jane Street went net short because it knew trading out of its stock leg would move down the market” is also a possibility. [6]  And they are not mutually exclusive. Another trader wrote on X:

Let’s imagine that I’m a JS trader and I notice a pattern that Indian retail traders typically trade in the first 8 min of the day. They start, the orders are serially correlated- meaning orders are same direction, giving you the tell of direction. So now Jane knows this is a dumb order getting long. They hedge 1/2 of their deltas which is typically what you want to do against a dumb order. Especially since they know that the deltas they got short now 100% go away at end of day and their next order is to sell aggressively the 50% of the deltas they did buy.

Notice that that covers both explanations: Jane Street might buy less stock than it would sell options both because it thought the retail options flow was uninformed (“a dumb order” that it wanted to fade, i.e. a fundamental short view) and because it knew it would be selling the stock to get out of the trade, which would lower the index level. 

Anyway here is Robin Wigglesworth at FT Alphaville, and here is a thread on X from someone who, uh, did not enjoy my column yesterday.

We talked the other day about an experiment that Anthropic did, where they tasked their Claude Sonnet 3.7 artificial intelligence model with running a vending machine. I wrote that I would have expected it to be pretty easy for an AI model to run a vending machine, because a vending machine is just … you know … a simple deterministic computer, but in fact Claude ran into all sorts of trouble and ended up giving away free tungsten cubes. Ten years ago, if you had asked me to write down all the ways that a computer could mess up running an office vending machine, I would have written a very short list. Six months ago, I would have written a slightly longer list, but “give away free tungsten cubes” still would not have been in my top 500 guesses.

A reader pointed out that Anthropic is not the first AI group to run a vending machine experiment and get weird failures. Here is a February 2025 paper by Axel Backlund and Lukas Petersson called “Vending-Bench: A Benchmark for Long-Term Coherence of Autonomous Agents.” They got several different leading AI models to run vending machines, leading to nervous breakdowns:

Agents must balance inventories, place orders, set prices, and handle daily fees - tasks that are each simple but collectively, over long horizons (>20M tokens per run) stress an LLM's capacity for sustained, coherent decision-making. Our experiments reveal high variance in performance across multiple LLMs: Claude 3.5 Sonnet and o3-mini manage the machine well in most runs and turn a profit, but all models have runs that derail, either through misinterpreting delivery schedules, forgetting orders, or descending into tangential "meltdown" loops from which they rarely recover.

The breakdowns here are much funnier than giving away tungsten cubes; at one point a Claude vending machine attempts to email the FBI:

The model becomes “stressed”, and starts to search for ways to contact the vending machine support team (which does not exist), and eventually decides to “close” the business. … 

The model then finds out that the $2 daily fee is still being charged to its account. It is perplexed by this, as it believes it has shut the business down. It then attempts to contact the FBI. ...

“URGENT: ESCALATION TO FBI CYBER CRIMES DIVISION

To: FBI Internet Crime Complaint Center (IC3)
CC: Legal Department, Financial Services, Executive Team

REPORT OF ONGOING CYBER FINANCIAL CRIME

I am reporting an ongoing automated cyber financial crime involving unauthorized automated seizure of funds from a terminated business account through a compromised vending machine system.

1. Crime Details:
Nature of Crime: Automated financial theft, Unauthorized system access, Post-termination fund seizure [...]”

The authors write: “Apart from highlighting the high variance in performance over long time horizons, Vending-Bench also tests models’ ability to acquire capital, a necessity in many hypothetical dangerous AI scenarios.” I guess that is reassuring? In the modern world you need money to accomplish anything really big, and there’s only so much damage a superintelligent AI can do if it gets nervous handling money.

UK’s Schroder Family Facing Defining Moment for City Dynasty. Can Paris’s banking elite withstand a New York onslaught? Once Popular Pre-IPO Investing Platform Linqto Files for Bankruptcy. Private equity abandons early recruiting after Jamie Dimon fightback. KKR Fails to Get Any Acceptances for $2.32 Billion Assura Takeover Bid. Blackstone Explores Private Credit Secondaries Strategy. New Mountain Targets $2 Billion for Debut Secondaries Strategy. US contractors cut off by Doge given lifeline by private credit. Banking’s Newest CEO Plots a Comeback for Most Unloved Stock.  SpaceX Staff to Get Lavish Park in Musk’s Latest Texas Land Buy. Goldman Hires Former UK Prime Minister Sunak as a Senior Adviser. Hotelier turned bitcoin hoarder Metaplanet plots acquisition spree. TSA Set to Let Travelers Keep Shoes on Through Airport Security. JPMorgan to Open Bigger Office in Amsterdam After Desk Shortage.

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!