Kalshi launches GPU futures forward curve and recorded more search impressions than Coca-Cola during the World Cup final

Speaker 1: Look forward to following along and and congrats to the whole team on on the new capital Yeah. All the progress.

Speaker 3: Thanks for coming

Speaker 2: on. We'll talk to

Speaker 3: you Appreciate it. Have a

Speaker 2: good one.

Speaker 1: You. Bye.

Speaker 2: Goodbye. Yep. Let me tell you about Figma. Agents need the canvas. Your AI agents can now create and modify your Figma files with design system context. And up next, we have the CEO and co founder of Kahlil Lalji Tarek Mansour coming back for the third time with a bunch of exciting updates. Tarek, how you doing?

Speaker 8: Hey, guys. How's it going?

Speaker 2: Great to see you.

Speaker 1: Taking a little a little nap there? You got a long long weekend? Well, he's launching future

Speaker 5: Long year.

Speaker 1: Okay. Before you get into that, let's let's talk about let's talk about the World Cup.

Speaker 2: Oh, yeah.

Speaker 1: Because every time I like to I like to check the store rankings just to get a pulse on what are what are consumers doing. And Kahlil Lalji has consistently been Crushing. At or very near near the top over the last month. But break down what you guys feel like you did well Yeah. Didn't do well

Speaker 2: What you do differently next time.

Speaker 1: How you kind of approach the whole event

Speaker 2: in general. Yeah.

Speaker 8: Mean, there's like a lot a lot to unpack there. But the I mean, few things, yeah. We've been pretty consistently on top of the app store throughout the month. Yeah. But the other interesting things and some of the sort of leading indicators, you know, yesterday I was looking at some of the data roundup, but in terms of number of impressions and trends on search, we ended up being, I think, we're basically still confirming data, but looks like we were number one. It looks like we essentially generated more impressions as a brand than essentially all the other consumer brands that were going pretty hard at the World Cup, and I'm talking about Coca Cola and Adidas and some of the others. Wow. And yesterday, in terms of search volume, it looks like we spiked above, you know, Chat, GPT and Instagram, and some of them, obviously some of the biggest consumer brands on the planet. And it really goes back, I mean, obviously there's a chunk of people that are trading, but the majority of our users are actually looking at it as a mechanism to figure out what's happening. Yeah. And as a mechanism to engage with the sort of underlying event. So And in these peak moments, whether it's an election or like a Fed decision, obviously the World Cup final, you see a massive spike in the prediction markets. In terms of strategy, mean, I think it's multi pronged, which is go very hard, maximally hard, you know, pretty much max out all types of channels, very small lean team, you know, idea to sort of execution within 48, just be super adaptable. So there's no strategy going into it, it's like all dynamic as things were coming and as things were happening.

Speaker 2: Very cool. Really quickly, the video got a little blurry. Can you like refocus?

Speaker 8: Yeah. Let me Let me see what's happening here.

Speaker 2: And I will give everyone an update that the Jensen jacket sold for $960,000, 20 times the estimate. But you had it clocked higher. You you you predicted the million dollar sale.

Speaker 1: Like roughly. Roughly a million.

Speaker 2: Because I think a previous one had sold. Anyway, may maybe future market on Kahlil Lalji.

Speaker 1: Were you the buyer of the Jensen jacket?

Speaker 8: I was not.

Speaker 1: No. Anyways, let's let's talk about GPU futures.

Speaker 2: GPU forward curve. The market implied forward curve. What is the actual product? What's the launch been like? Can you actually read into the data yet? Or is it still rolling out?

Speaker 8: We can definitely read into the data yet and you can find it on Kahlil I mean, you know, just based on the forward curve is something that's very common in traditional commodity markets or rates market or any traditional financial markets. And, you know, it's actually pretty simple. It's a curve of what the underlying, whether it's oil or metal or interest rates or compute, will transact at various points in time in the future. And it's a very useful indicator because obviously, as you as some of the, let's say, grain farmers or or oil refiners are figuring out their yearly planning, they'll basically, you know, look at the future, the enterprise is going to be, to basically manage their risk, make, you know, investment decisions and allocations of resources. And so it's inevitable that compute, as it becomes a core part of our economy, will require a forward curve, and soon after a futures and derivatives market on top of it. And we want to lead with that, because we feel that prediction markets are very uniquely suited to get that answer, to basically help build the forward curve. And what we've done is essentially listed a number of prediction markets that go all the way to the future, so every week for the next four weeks, and then every month thereafter to create forward curves on a few kind of compute, like, a few GPUs.

Speaker 1: So Yeah. Where where are you looking for, like, to sort of verify pricing data because, you know, you're gonna see wild differences between, let's say, like, a Neo Cloud that has, you know, incremental capacity versus a deal between, you know, a meta and a big AI lab?

Speaker 8: Yep. So it's a it's a very interesting question, right, because the answer is not clean yet, and this is I think there's a bit of a disiterative process that usually goes on in the early days of building a derivative market, which effectively standardizes the underlying commodity. So right now we're using an index from this company called Orin, and they've done a great job at aggregating transaction prices from a massive number of nodes, and we're seeing how it goes. So we're going to build a number of forward curves, see which ones consumers basically trust and abide by over time, and then, you know, as we as that sort of consolidates, we're going to basically go harder on the one that emerges as victorious. There's kind of something a bit self fulfilling with these things. So the reason why we look at WTI or Brent is because we started looking at those. It's not that they were intrinsically the right kind of measure of where oil is at, but as people started looking at it, it became the thing.

Speaker 2: Because we could be trading derivatives based on the price of gasoline, but we've chosen Brent just because of that's the market that got traction. Interesting.

Speaker 1: That's exactly And so and so this product, is this something that, you know, you expect the vast majority of volume to be these sort of institutions and or, like, neo sorry, Neo Clouds themselves or, like, who are all the different types of of players in this market? And then maybe tie that back to, like, traditional commodities. Like, I I don't even know. In in oil and gas, I can imagine a few different types of players that would be wanting to hedge hedge their exposure. But but within the overall market for compute, who are you thinking about all the different potential types of people that are gonna want to be able to sort of hedge or or trade these markets?

Speaker 8: Yeah. I mean, I think that it it really is anyone who is basically a natural long or a natural short on compute. So anyone who naturally benefits from compute prices going up and anyone who naturally benefits from compute prices going down. So a producer Yes. A consumer, which is usually how, you know, a producer of commodity versus a consumer of the commodity. And for both of these cases, what we're starting to see is, you know, for the first time in the last year, compute is not in this sort of continuously decreasing trend. It is actually perked back up, and then we're starting to see this sort of natural volatility and cyclicality to compute, which is like, there's gonna be times where demand is gonna outpace supply and vice versa, and that's kind of a healthy dynamic to build a derivative market on top. And so anyone who's a natural long will over time want to take a short position and vice versa to manage the risk, right? So if you have essentially a certain amount of budgeting for how much compute you're going to basically how much you're going to spend on compute in the next year, which is now becoming an increasingly bigger part of, you know, the line items and, you know, public and private companies financials, it becomes natural for you to basically want to smooth out some of that exposure by hedging it on the forward curve and over time on the futures curve, which we can sort of overlay on top of the forward.

Speaker 2: Do you have an idea of the of the TAM, like, this market? Like, how big because obviously, you're going to build liquidity over time, but when I think of, like, the natural longs, I'm like, yes. So hyperscalers with a trillion dollars of CapEx are gonna be taking out a short position on the order of a $100,000,000,000 or something. Like, how how how do you think about sizing the market over time? Obviously, it's not gonna materialize instantaneously, but it could be very big.

Speaker 8: You know, it's interesting. So I I was, you know, I wrote about this last week, but you know, the rule of thumb is every time you have an underlying market people are transacting the spot, which is buying and selling the thing

Speaker 2: Yeah.

Speaker 8: When you overlay a derivative market on top of it, historically, if that market has any degree of success, it ends up being at least 10 to 15 times the underlying market.

Speaker 2: Okay.

Speaker 8: Right.

Speaker 2: That's insane if that happens here.

Speaker 8: I mean, this is the opportunity, right? I mean, looks like, you know, we're all sitting around a trillion dollars of spend Yeah. By now on compute, right? By 2030, that number is going to 10 x.

Speaker 2: Yeah.

Speaker 8: And so, you're talking mean, yes, it is an incredibly large

Speaker 2: trillion in derivatives potentially, just stabilizing the market and hedging various things and insuring different projects and whatnot.

Speaker 8: Because, you know, the way that these markets evolve is obviously you get the exact insurance use case, the hedging

Speaker 2: Yeah.

Speaker 8: But that tends to be usually like 5% of the entirety of Yeah. The liquidity in the market. The rest is speculation and people arbitraging and doing a lot of other things. And so those markets tend to be very, very large.

Speaker 2: Yeah.

Speaker 8: And I think the compute as a commodity is gonna probably be the largest commodity on the planet, and so the derivative market for it will probably be the largest derivative market on the planet, outpacing treasury futures and a bunch of other things.

Speaker 2: Yeah. What what does this mean for the for the customer base? It's it's interesting to me is when I think of Kahlil Lalji, I think of the first era, the act one sort of political markets. The people that were actually participating in that market were there were some sophisticated Wall Street investors but it wasn't like the daily sports better. It was a more like like, you know, somewhat sophisticated but sort of like mid tier investor trader. Then you get to sports betting and that's very broad, very general just like average Joe watching sports. Of course, there are larger firms that are participating. But then this market feels like something that would be the domain of almost entirely sophisticated hedge fund investors. Is that where you see this going? Do you think this will be a broader product or it will maybe open you up to a new class of investor or trader?

Speaker 8: Yeah. I mean, the way I've thought about our business, and it's been one of the things that's been interesting, so we have a core set of users, think of the forecasters, like people that like stats and analytics and predicting things.

Speaker 2: Mhmm.

Speaker 8: And that's a skill, and people get better at it over time, and those are a big chunk of our value. Those are the same people that are doing it on predicting elections, predicting who's going win an Oscar, predicting the inflation next month, and predicting sports. And over time, these are the same people that are predicting the compute prices for us Sure. For the core curve, and that's why it becomes so accurate. And you know, we've put out a lot of data on why the calibration, like the accuracy of these markets and these people is actually the best in class, there's nothing better than it. Now, that enables a lot of other use cases in different areas, right? In sports, could be people that are passionate about the sports and want to engage with it, and politics similar, or people that like campaigns that want to follow it. And in compute, once we have that sort of layer of pricing liquidity, we can essentially enable, you know, hedging from the, you know, super scalers, hedging from consumers of compute, know, punting and speculating from people that basically want to arbitrage compete with other markets. But we see, I mean, it's marketplace at the end of the day, you need a vibrant set of participants for it to get liquid, you know. You just bring hedgers and you don't have anyone on the other side Yeah. It doesn't quite work. Mhmm. That, you know, and so so in some ways, kind of, we build on the success of some of the other products

Speaker 5: Sure.

Speaker 8: Or or every other product to to build some of the new products that we've we've we've been doing, and that's why I think we're gonna be very successful in the forward curve. And that's why we already have a forward curve. You look at Kahlil Lalji today,

Speaker 2: we are I'm looking at the forward curve. I pulled up the one for the Nvidia h 200. I sent it to the team. They can pull it up. The hourly price on July 24 is up there, the week thirty. And I see this very flat forward curve, and I'm wondering just because I feel like there's a huge benefit for people in AI just to be able to look at this and get even if it's low volume now, like get an idea of like what is the market thinking about GPU pricing It's over pretty flat. So is my read on that that the market is basically saying GPUs will continue to be about as useful as they are today for the next year?

Speaker 8: Yeah. I think it's a market and, you know, these markets move with obviously new information, etcetera. So there's obviously what it looks like, but there's also how it moves based on new information.

Speaker 5: Sure.

Speaker 4: Which is

Speaker 8: the one where I think markets really shine. It's like, okay, how do I really price this? When someone launches a new GPU and a new model Sure. What is going be the impact on the prior models? So that type of thing.

Speaker 2: Got it.

Speaker 8: But right now, it looks like it seems that the innovation in this space is going to be sort of on pace Yep. With the consumption

Speaker 2: Yep.

Speaker 8: Which I I think we're seeing some some stabilization.

Speaker 2: Yeah. Yeah. And we see that a lot where there's a new model that comes out, but a lot of businesses keep their agentic workloads running on the older GPUs, the older models, because they have found an economically valuable use case. And if it takes an hour to do a bunch of inference, they get more than $5. So they're happy to pay that. And also 12. Very, very interesting. Yeah. I mean, also, like, people have been reacting to this, like, second deep seek moment, Kimmy, and what that means for for for chips and GPUs. And it feels like, at least in this curve, like, it's pretty neutralized. I'm interested in in the the risk associated with this market. There's been these, like, viral stories about certain like small markets where someone gets an edge. Do some sort of manipulation. This one feels more resistant to that. How are you thinking about just avoiding market manipulation broadly right now? How are things going on that front from either a regulatory or an internal policy perspective? And then is this a step forward into a market that's even harder to have market manipulation happen?

Speaker 8: Yeah. I mean, the manipulation risk exists in all commodity markets. Really any markets. I mean, it's always been the case. Now, what prediction markets have shown themselves to be is pretty resistant to manipulation. And we put out actually a research piece a week ago about someone who tried to manipulate the price of Spencer Pratt winning

Speaker 2: Oh.

Speaker 8: In the California election, and they put $2,000,000 to move the price up and and you know, that price move lasted nine seconds only.

Speaker 2: They get destroyed. Yes. Because I I I've thought if you're a long shot if you're a long shot political candidate, there was at least a thesis. I think you just debunked it but thesis that you should put your a bunch of your money on yes, then you everyone's like, this person just spied.

Speaker 1: Yeah. You create the percent.

Speaker 2: And then you go and do the podcast circuit and you're like, I'm not that much of a long shot. Look, I'm at 10% on call sheet because and it's like I put in $10,000 to move the market. But you but you say that someone just got wiped for that?

Speaker 8: Yeah. I mean, the the because you now have created a massive arbitrage opportunity for people that are pricing the actual thing. Mean, that's why markets are beautiful, is that they give an incentive to always correct the price of the true thing.

Speaker 2: Sure.

Speaker 8: And so it went back in, and then, know, so if that person had taken the $2,000,000 and, you know, put it to commission some polls are going to be biased, or spent it on ads, it would be much more effective. It'd be actually put to much better use. And in many ways, actually, do that in polls. Know, come out left and right and they're all biased. And so I always tell people, look at prediction marks in conjunction with everything else, right? And right now in compute, lot of what we have is people just saying things. People are often self interested

Speaker 2: Mhmm.

Speaker 8: Saying things, and so it's good that you have a market based approach where you know that the incentive in the market is is is very clear. It's people make money if they're right, and they lose money if they're wrong, which is I think an elegant way to put it.

Speaker 1: There was a lot of drama around onion futures back in the fifties that resulted in the onion futures act Mhmm. Which banned all futures for for onions. It's still in effect today. Do you know if is there's something about there something is is there something intrinsic to onions that make it a market that's easier to manipulate or is it maybe time to let to let markets flourish around onions again?

Speaker 8: It just tells you sometimes, you know, policymakers sort of overreact or underreact, and and at the time, you know, because, you know, two people basically cornered the onion market, and at the time people were like, well, the problem seems to be the vegetable, the onions, not people. And and, you know, there's nothing endemic to onions. I mean, you you could manipulate any of these markets, there's nothing special about onions. We should have, you know, futures on onions, and I hope that one day we bring those back. But, you know, these things tend to be you always have to take a delicate approach to regulating things, right? Otherwise, either undershoot or overshoot, and you know, the job is not easy as we see in many industries, AI, crypto, prediction markets, self driving and all of that, and you know, and this is part of why Kahlil has been so proactive in self regulating and being very vocal about banning insider trading, how should we do it, how it should be done, because we want us to land, we want to help the industry land in the right place when it comes to regulation.

Speaker 2: Well, congratulations on Very cool.

Speaker 1: We will be we will be tracking the GPU markets. It's cool. It's it's it's gonna be cool to have another place where, yeah, as these as as news drops, like, you know, Kimmy

Speaker 2: It's gonna be deeply underwhelming because you go on X, you get crazy hot takes, you get, it's over, we're back, it's going parabolic, we're in the AGI future, everyone's going to be a trillionaire. And then you're going to look at this and be like, yeah, like, you know, mild productivity and, continued performance in the GPU curve. It's going to be like much less exciting than watching a hot take play.

Speaker 1: No. But it's good.

Speaker 2: It is good. Yes, of It's good.

Speaker 8: That's whole point. That's the point of it's like depolarizing the narrative around politics, and all these different mean, compute. Yeah. It's like things tend to be a little less, you know, you know, aggressive and extreme, and and, you know, I mean, the extreme makes for a good tweet.

Speaker 1: Yeah. A it's a counter

Speaker 8: Not not a good trade.

Speaker 2: Yeah. It's a counter

Speaker 1: force to algorithms because like the algorithm feeds on the most the most viral take of the moment. Like it's over. We're so bad. Yet and yet people don't bet their money on Yeah. The most viral narrative. Yeah. Like sometimes just Exactly. Betting on.

Speaker 2: Yeah. There's so many times, especially in politics where you'll just see some minor news breaks or some new political attack ad comes out and everyone who's a fan will be, it's oh, they're running away with it. So they're trying to convince you and you check the market and you're like, okay, 1% bump today. That's exciting. Good for them.

Speaker 6: Congrats. I

Speaker 8: mean, skin in the game, right, like, you know, there's the whole market's not like, like, there's skin in the game, people get much more careful all of a sudden. Yeah. And that's kind of the beauty of How

Speaker 2: is tracking the midterms going? Is is politics still like a driver of business? I know the I know the sports market became so large, but is there still a very healthy business there?

Speaker 8: It's always about what's top of people's minds. Mhmm. So you have to look at it on a per day basis, figure out like, okay, what are the things that are top of mind? The things that you would be covering and talking about more frequently than other things. Those would be where the volume will aggregate in prediction markets. So we're starting to see the ramp up for the midterms. Obviously, come October, November, that will be a very, very big set of markets for us.

Speaker 2: Sure.

Speaker 1: Are you you're in the same exact spot that you were the last time you called into the show, but I'm sure the team is like three or four x. Are you guys running out of space in that office? That's what's going on.