In late 2023, early 2024, we had a hunch at SFC that compute would commodify, and that subsequently compute futures would become possible. We spent the next 6 months or so exploring the idea in depth, talking to specialized counsel about regulatory structures, having no-names conversations with the CFTC about their inclination to regulate compute as a commodity, speaking to traders about their interest to trade compute, and of course talking to anyone else who would listen to our arcane hobbyhorse.
We came away with less clarity than we had before the exercise. Commodities have existed in America for a long time, and as such many of the people we spoke to were, to put it bluntly, “lost in the sauce.” Deep experts in their field, of course, but after a career of interacting primarily with other experts in that same field we found the conversations around creating a net new commodity from the physical delivery side a bit like asking a school of fish to describe what it’s like to be wet. That is, until we spoke to Rich Jaycobs.
Rich has spent his career building financial markets, including some decidedly non-obvious ones, like futures on movie box-office receipts (one of the only 2 commodities contracts to be banned by Congress, the other is onions). He’s served as president and CEO of Cantor Exchange and Cantor Clearinghouse, and as CEO of ELX Futures, an electronic exchange trading U.S. Treasury and Eurodollar futures. Before Cantor, he was CEO of The Clearing Corporation, formerly the Chicago Board of Trade Clearing Corporation, and founder and CEO of onExchange, an early CFTC-approved exchange and clearinghouse. He was chief technology officer at Tudor Investment Corporation and a managing director at the New York Cotton Exchange. He’s personally had a hand in 7 approved futures-exchange applications and 6 approved clearinghouse applications, roughly 1/5 of all approved DCMs since the new process was introduced by the CFTC in 2000. Rich is the GOAT.
Aside from his impressive resume, Rich is also the most concise, salient mind I have ever met when it comes to commodities. He has a singular talent for breaking down his complex, highly regulated world into simple concepts and incentives.
Since Rich and I last spoke the idea that compute is a commodity has become almost passé. Indexes and exchanges have sprung up, partnered, and released contracts. The industry is moving fast, so fast I fear I’ve forgotten my basics. Below is a conversation where I attempt to get back to those basics to understand what problem futures contracts are trying to solve, and what actually makes a commodity.
Ethan: People are pursuing compute futures under the assumption that compute is a commodity and commodities should have futures markets. The debate tends to get stuck on fungibility, but what actually matters when you’re looking at a new market?
I take my wife to a baseball game. She understands the basics, and she can tell you a lot about baseball now. But baseball has a lot of nuance. You could also say it’s pretty easy: you just have to get a hit and score a run.
That’s how people talk about launching a market. We just need liquidity and an index. Get hits, score runs.
Young people are approaching the market, in my experience, with: I want to get data, I want to make an index, and I want people to trade my index. There’s enough volatility and curiosity about what’s happening in compute that everybody thinks there should be a financial instrument of some kind.
But I gave you that chart showing the steps, and you’re right: everybody has jumped to step six. The quality of the indices is still uncertain. How many brokers do you get information from? How do you average it?
Right now, I’m not seeing convergence between the physical side of the market and the financial-index side.
Early in our conversations Rich gave us a chart, which I’ve recreated here. The chart goes through the different structures a commodity market might take, more or less in chronological order.
What’s hot? AI is hot right now. Everybody wants to trade that in some way, make some kind of bet on it.
Ethan: What does that disconnect look like?
It started like a barbell. You have people who understand data centers, locations, and the differences between chips. And then you have people who want to bet on the financial index. Those aren’t necessarily the same people.
The financial guys have leapt right to it. That’s unusual, but there’s enough interest that they’re able to do it.
The stuff that connects the two, which makes the prices real and makes real commercial users participate, I’m not seeing so much right now.
Ethan: What do you think is driving that financial interest?
Gamblers. I’m going to say gamblers because they don’t have any direct connection to the GPUs.
What’s hot? AI is hot right now. Everybody wants to trade that in some way, make some kind of bet on it. I don’t think a lot of the people in that market are really paying attention to whether the index is any good or not.
Ethan: Have you seen this kind of speculative cycle before, or is the appetite to bet on everything something new?
The technology is new. The human behavior is not new.
I have this discussion with my wife. She’s so conservative it drives me nuts. Now she’s coming to me: “How come my sister owns all this stock? They’ve made all this money. We have to buy stock.”
“Sweetheart, we’ve had this conversation. You hate stocks.”
“I know, but they’re making all this money.”
Okay, we’re near the top. My wife wants to buy stocks.
I don’t know where the top is. But I think we’ll look back on this period of speculation and go: wow, that was insane.
Ethan: So when you’re looking at a more traditional market, what does that look like? How does a derivative normally bubble up?
Great question, it starts with you guys. You’ve got long-term contracts, and they start to peel off, or financial players step in there. I’ve got a 5-year deal to use this, and I want to turn that over into shorter, differently structured deals.
That buying and selling of compute, very much the model we talked about 2 years ago, creates: oh, now the price is going up and down a little bit every day, and now I want to lock that in. And now you develop the futures-market side of it.
Because your product has enough variability, somebody comes along and says: you know what? We need to standardize this right now. This thing is all fragmented because the data center in Stockholm is not going for the same price as the data center in India or Arizona or wherever the hell they are.
I’m going to create an index that weights it some way, and then everybody follows my index, and then we have a financial product. That’s the way it normally evolves.
Ethan: Where do private, over-the-counter deals fit into that?
I think in your space it’s broken into 3 models. You guys and the folks who have some physical connectivity. The second group that’s caught on is what you would normally see in the process: doing swaps on indices. I think that became the middle ground. And then you have your DCMs on the far end.
Rather than a standardized product, if you and I negotiate a contract, I can tweak it.
Well, I only want to do this chip. Or I only want to do this region. Or I only want to use these prices. Or I’m worried about manipulation, and I want to put a cap on it because I’m afraid it’s going to spike because somebody’s going to manipulate it.
We can negotiate all those things in a bespoke way. Or I want to have the average versus a particular point in time. We get to define all of those, just me and you, because we’re defining how our bet works. That’s easy to do.
Ethan: What becomes painful enough about those private arrangements that makes standardization and exchange-traded contracts worth the trouble?
Well, what usually happens is now I’ve got a portfolio of 30 of these things, just to pick a number. And so do you, and so do they, and they, and they, and they.
Now all of a sudden, our credit lines are getting all jammed up. I’ve got a credit exposure, and you know I’ve got a credit exposure, and he knows you’ve got a credit exposure, and we’re all looking at it: can we net these down?
We’re all exposed to each other. We know that. Can we come up with some math that will allow all of us to just net out whatever we’ve got here? We’ve got all these obligations to each other.
Ethan: And this is a combination of commercial participants and financial participants?
It’s almost always the financial participants that do this, because the commercial participants bought the insurance they want. They do maybe 4 transactions.
But if I did 4 and you did 4, and, oh, by the way, because I did a commercial transaction with you, I want to lay off 50% of it to this guy. So now I’ve got 2 transactions that are offsetting. But now he’s got that, and he wants to lay off half of that to that guy.
Oh, and then I want to lay off the other half of mine. Or I want to buy my half back, but he doesn’t want to sell it back to me. I want to buy it back because it’s profitable. He doesn’t want to sell it to me because it’s profitable. So now I’ve got to go do it over here.
My single deal with you all of a sudden winds up getting replicated through a financial-markets ecosystem. And that’s when the market grows, but it’s tied to your transaction.
But now I’ve got 10 people. This is why people say, “The derivatives markets are always 100 times the size of the underlying.” Yes, because we keep redealing the same trade between each other.
As the market moves my way, I take a little bit of profit, but I keep some of it. Then you retrade it, and you retrade it, and everybody retrades it. But if you actually netted the whole thing down, we’re back to your transaction.
Now, that’s not always true. Sometimes I’ll keep a little exposure. I’m happy to keep a little exposure. But your one transaction caused me to take it and parcel it out. Then they parcel it out, and then I buy it back, and we parcel it out again. Now all of a sudden, you get this volume. That’s the dynamic that causes the market to grow.
But now we get to this problem. We’ve done this packaging and repackaging of basically one trade in my example. On my books and records, it nets to zero, but I’ve got to worry: if he doesn’t pay me, I can’t pay him. And he knows that if he doesn’t pay me, I can’t pay him.
So we come together and say: let’s go visit the CME and see if we can’t take all this shit that we just did, give it to them. Tell them: standardize it, net it down.
And then it looks like: oh, here, my P&L was $350,000 on my $20 million of outstanding trades, and I have no balance-sheet obligation. That’s great.
Bad indices blow up when somebody has a billion-dollar trade on and the index can be manipulated for $10,000. That’s your number-one problem.
Ethan: We’re seeing a bunch of indices pop up in the compute space right now. What’s driving that?
Venture capital money is flowing excessively into the space. And people have to differentiate themselves to raise money.
You put those 2 ingredients together, and just like you said, you’re going to get 3, 4, 5, 10 indices born.
Because mine is better, and here’s why you want to invest in me, and I’m going to get an audience, and I’ve got DRW or 10 other firms interested in talking to me, blah, blah, blah, right?
Ethan: What’s the difference between a price that’s good enough to put in a newsletter and a price that’s good enough to settle a billion-dollar disagreement?
Bad indices blow up when somebody has a billion-dollar trade on and the index can be manipulated for $10,000. That’s your number-one problem.
That’s why index creation is its own specialty. How do you get a genuine price when there’s an extreme price move, while still having a filter that prevents manipulation from affecting the index? That’s an art.
I don’t know whether anybody really has that art down yet for your market.
Ethan: Why do we keep coming back to the physical side first?
Part of the reason is to see what substitutions people can make. Somebody tries to push prices up by buying a whole bunch of compute, but then they’re stuck with compute they overpaid for and can’t sell.
With financial settlement, that’s not the way it works. I jam the index up. Cash settles, I get paid. And if it drops by half the next hour, I don’t care. Because I’ve been paid.
Ethan: Does a market ever get past the point where people try to manipulate it?
There’s always somebody trying. The market has to be big enough to absorb the manipulation or call it out for what it was.
LIBOR was forever the standard. That was the standard for 30 years. The banks got so greedy, they manipulated it.
They wound up paying billions of dollars in fines to the U.S. government and others because they could not resist themselves from manipulating the damn thing.
Ethan: There’s this obsession in technology with first-mover advantage. People are trying to forecast the future, skip to the end, and start there. But when it comes to aggregating liquidity, what matters more: being first or being best? Will traders accept a worse contract or a worse index because that’s where the liquidity already is? Or will it shift to a better-constructed product?
Okay, so this is the ultimate—I’ll call it, with air quotes, “corruption.”
This very question is happening right now with Kalshi and Polymarket. The liquidity is being provided based on the “bribes” the market makers are receiving from them to provide liquidity to their markets. It’s all-out: forget the underlying, forget the quality of the product. Nobody gives a shit.
Here’s the art. If I’m a market maker and you’re negotiating with me, I’m going to say this:
“I’ve got a lot to do. I can make a lot of money doing a lot of things. I don’t need your shit.
“But I’ll tell you what. I’ve got 5 guys over there working on something else. If you’re willing to pay me 3 times what they get paid, cover all my losses should I lose any money in your marketplace, and give me a million bucks a year on top of that, we’ll prioritize.”
Ethan: Has that always been the case, or is this happening because there’s such a huge influx of venture capital and so many new markets coming up?
You nailed it. Look at Kalshi and Polymarket. I’m picking those two because the numbers are so big they just pop out. They’re raising money at $40 billion and $20 billion valuations, respectively. They’ve each got $1 billion in cash, and it’s a race.
You talk about first mover. It’s a race to be the winner, period. Forget first mover. They’re both effectively first movers in slightly different ways, and now they’re in a death battle to see one dominate the other.
And they’re also trying to draw some lines of peace between them because it’s getting too expensive.
A good chunk of that money is coming from investors. And this is the answer to your question in compute, too. You get this influx of money, so the easiest thing to do is buy your way there.
You can buy customers to trade your product through advertising. Kalshi and Polymarket advertising is flooding the market right now. That’s buying one type of user.
Then there’s the direct payment to market makers. That’s another type of user.
Net-net, it’s a cost of customer acquisition. It’s a cost of building your business.
In that model, they don’t care what the underlying is. The guy—or the group—that comes along with the fattest checkbook and says, “Trade my product and I’ll make it worth your while,” wins.
And that happens a lot in our space in the early days.
Ethan: But all these exchanges can’t keep spending investor money forever. What happens when the funding runs out?
Some of the startups, especially the later ones, are simply going to run out of money. The VCs aren’t going to continue funding them.
Sooner or later, we’ve reached the West Coast. There’s nothing out here but the Pacific Ocean, and we’re not talking about monetizing the Pacific Ocean.
Eventually, growth slows and people start asking what can be combined. Liquidity doesn’t do well fragmented.
There can be lots of licenses without there being lots of economically viable markets.
Ethan: You’ve seen these land grabs before. How obvious is it at the time which companies are going to survive?
When Cantor was interested in crypto, I went up the West Coast from San Francisco to Seattle and talked to at least 30 startups. Most had raised somewhere between $10 million and $50 million.
I went back to New York and wrote a paper saying: here are the roughly 30 names we’ve spoken to. You either have to deal with all of them, or you have to pick the horse you think will win.
It wasn’t obvious what was going to make any particular firm work. Brian Armstrong was impressive; what he had built was further along than most. But I know people who were there in the early days and left because they didn’t think the company was going to make it.
That’s where I think your space is right now. Not quite as many players, and not quite as well understood, but the same dynamic.
Ethan: A lot of tech founders have never really interacted with regulation.
And hate it too.
Ethan: And they hate it. From our conversations, the venue can make or break the product. What are you looking for in a partner, and what can go wrong?
You want to own your clients and control your stack. Software people understand that instinct very well.
The minute you don’t control or understand the regulatory process, you’re ceding something to somebody else. That’s a big deal.
As the entrepreneur who doesn’t want to have your own exchange, you’re getting married. All the complexity of deciding whether to marry somebody goes into this equation, except it’s the business version.
It goes well when the partners bring separate things to the table. One has infrastructure; the other has distribution. They aren’t trying to get into each other’s businesses.
It goes badly when one thinks it dominates the other, or when expectations aren’t met. I thought you were bringing customers, and you didn’t. You thought I was bringing technology, and it turns out my technology isn’t very good.
Then there are fees. We agree to charge the customer a penny. How do we split that penny? What adjustments get made before we split it?
A venture capitalist once put it to me this way: you’ve got a small company with the regulated side, and another small company with the technology side. “Why don’t we tie the 2 rocks together and see if they float?”
I think together we can solve the world’s problems, but neither of us is really equipped to solve our own problems, no less the world’s problems.
Ethan: What’s your advice to a company entering the space?
Control your own destiny. It’s complicated, but there are people you can hire who understand it. Start small and be ready to pivot. Partnerships are enormously distracting.
And be careful about exclusivity. The minute you agree to exclusivity, you’re betting that this partner is going to succeed for you.
“Till death do us part” is what exclusivity means. Unfortunately, corporations don’t die like people do. They linger and linger and linger and linger.
Ethan: Okay, you’ve got your contract, your index, and your venue. How do you actually aggregate liquidity?
Okay. So let’s go back to first principles, which is you “bribe” your way into it in the first position. Right. We talked about that.
Ethan: Yup.
Liquidity begets liquidity. That process we talked about, where people divide up and retrade a commercial transaction, eventually gets enough general interest that there’s somebody willing to buy and somebody willing to sell.
You also need a certain degree of passion and disagreement. One of the surest ways to kill liquidity is to have everybody agree.
Who’s the seller? Who’s the buyer?
Sometimes there’s something in a contract that’s hard to value. You and I do the math and come up with slightly different answers. That’s enough reason for us to trade with each other all day long. Our models vary based on some parameter.
The other thing about liquidity is passion. Just because I’m a Red Sox fan, I will lose $1,000 this month. It’s part of my entertainment budget. So I am a dummy for repeating behavior that I know loses me money.
Then there are people who don’t even know they’re in a market. I’m guessing you’ve bought gas in the last month.
Ethan: I don’t have a car.
Well, if you took a bus, they bought the gas. Or you took a plane. You used some form of transportation that used energy. Your money was a participant in the oil market.
You may use less of it, but you’re still a participant because you consume the product. Whether it’s my Red Sox or your transportation, the more of those players you have, the more natural liquidity you have. You can’t help but use oil-related products. Your machine’s got some plastic in it. You can’t help it.
Ethan: That seems like an interesting case for compute. As more work uses AI, more people are participating in the underlying market whether or not they think of themselves that way. How does that turn into demand for financial trading?
Exactly. The buyers and suppliers are making guesses, and those guesses are going to have timing mismatches.
If prices go higher, people want to build more data centers. People build because they expect demand to keep growing. But the timing doesn’t line up perfectly.
Those mismatches cause volatility. The volatility becomes apparent, and financial players jump in like sharks chasing minnows.
These aren’t little ripples on a pond. They’re oceanic changes. There’s an enormous amount of money being invested, and the supply-and-demand equations aren’t stable.
That’s what causes those of us who live in markets as a career to say: this would be cool to trade. I don’t know how I want to play yet, but I want a vehicle that lets me act when I’m convinced my research is better than Ethan’s research.
As long as there’s perceived upside, the world will get more exchanges whether it needs them or not.
Ethan: Futures exchanges have consolidated enormously. Does the world need new ones? What about companies that want to be the compute exchange and nothing else?
When I started, there were exchanges for different commodities: metals, oil, cotton, coffee, sugar, cocoa. There were grain exchanges in different cities.
Consolidation makes sense when an industry matures and the focus turns to cost. There’s infrastructure to maintain, and everybody has to connect to it.
But when people see a huge amount of potential growth, it’s a land grab. Everybody thinks they can get their piece.
As long as there’s perceived upside, the world will get more exchanges whether it needs them or not.
Ethan: Here’s where I get stuck. There’s interest in betting on compute, some commercial interest in hedging, and some private swaps happening. But the financial interest seems far ahead of the physical trading. A product people will bet on isn’t necessarily a product that gives a compute buyer or seller a useful hedge. What would tell you the two sides had actually connected?
The financial markets get liquid enough that the folks who sell compute start looking over there and saying, “I may be better off selling the financial product and taking whatever risk I still have on the physical side.”
There’s your convergence. But the opportunity has to be big enough for them to pay attention.
Right now, the financial markets haven’t overlapped enough with the physical markets for those operators to say, “I need to understand what’s happening over there, because I may be better off selling my compute through that index than selling it to the guy who’s actually going to use it.”
It may be worthwhile for me to let all those GPU cycles just go off as heat. No work product that the world cares about. Because I was able to monetize it over here. I don’t really care.
That happens in the electricity markets, but it hasn’t happened in compute because you haven’t linked them. It’s coming.
