Kristin Myers: Now, the people who trade for a living have spent decades and billions of dollars on one thing: getting information a fraction of a second before somebody else. That means private fiber, microwave towers, cables laid across the Atlantic just to save milliseconds. And that's how traditional markets work. Now, the newer ones, the ones that run 24 hours a day on blockchain, mostly run on the ordinary internet. Well, my guest's company is building that private road for them. But first, let's take a look at what they are building. The road from an idea to the opening bell starts right now.
On Wall Street, milliseconds can mean millions. But as financial markets move on-chain, trading 24 hours a day, seven days a week, the public internet may not be fast enough for what comes next. Austin Federa is betting there's a better way. He's the co-founder of DoubleZero, a dedicated high-performance network built for distributed systems, designed to move data faster and more efficiently between participants around the world.
And the stakes are getting bigger. Professional trading firms are moving deeper into on-chain markets. Asia is becoming a critical trading hub. Everything from tokenized assets to prediction markets could increasingly trade around the clock. So now here comes the real test: if the next generation of financial markets demands faster, always-on infrastructure, can DoubleZero become the network those markets run on? From Startup to Stock Exchange, Austin Federa is next.
And joining us now is Austin Federa, co-founder at DoubleZero. Austin, thank you so much for joining us.
Austin Federa: Thanks for having me.
Kristin Myers: So let's talk about the problem. We just heard about how a lot of traders are trying to save fractions of a second. So what made you think that DoubleZero really needed to exist? Talk us through the issue that you decided to solve.
Austin Federa: Yeah. So in traditional finance, there's billions of dollars each year spent on market access and market data access, and time is really the determining factor there. And you saw all of the major exchanges go through a period where they decided to say, we're going to put everyone's server in the same data center. We're going to try and equalize it. But most of that speed nowadays is arbitrage between different venues.
In blockchain, in this whole category of new finance, we still have very inefficient market data and very immature market data environments. It's very hard to actually get a complete picture of everything that's happening on Hyperliquid, let alone get it delivered to you as fast as you really need to make trades if you're a professional trader.
And so what we decided to build was a sort of next-generation fiber network. And this is a fully programmable environment, which is very different than what you get in a traditional network environment. Jump and Citadel and DRW run their own private networks, but it's just for their usage. You don't see this multiparty system. And so what we're trying to do is very similar to what the cloud did for compute, but for connectivity: make it programmable, accessible, open to all, and really democratize access to high-frequency market data in new finance.
Kristin Myers: So I want to talk to you then about multicast, because your site says that the technology behind this is essentially multicast. So that's the same way that, you know, the New York Stock Exchange, CME, other exchanges have essentially distributed market data for quite a long time. So is this something new, or would you say that this is more rebuilding the plumbing that already exists?
Austin Federa: I love that we're [unclear: "30s"] and we're getting to multicast. So multicast is an industry-standard technology in traditional finance that's been used for 20-plus years. The difficulty with multicast is you can't run it over the public internet. And what it does is it basically replicates data in a much more efficient and much more fair way than the way that you currently get data from a centralized exchange or a blockchain.
And what that really means is, if microseconds and fractions of a second are really the difference between making the trade and not, there's a whole compliance enforcement regime inside of these data centers to make sure that everybody is getting the data at the same time. That doesn't exist at all in new finance yet. And so not only can we distribute data more quickly, but we can make sure that people are receiving that data in a much more fair system than they're currently receiving it today.
Kristin Myers: Okay. How much more quickly? I have a feeling that you're going to tell me a very infinitesimally small number. But how much faster, and what are we, you know, measuring that speed against?
Austin Federa: Yeah. So, you know, blockchains have gotten much faster over the years. They're still orders of magnitude slower than NYSE or Nasdaq, orders of magnitude. So when we're talking about our speeds, we're talking about, compared to Hyperliquid speeds, sometimes 100 milliseconds or more as a performance advantage. Steady state, probably about 70 milliseconds.
That may not seem like much time, right? That is an eternity to a high-frequency trading firm or a market-making firm. If you're at a user interface clicking a button, you probably don't need 70 milliseconds of advantage, right? But we're moving very quickly to a world where AI agents will be trading on behalf of users. And that gap in analysis and execution between the best market-making firms and the best high-frequency trading firms, and what the average user is going to be able to do with a server co-located in Tokyo, is dropping every day.
Kristin Myers: So, as you were saying, the average person does not know the difference of seven milliseconds, right, compared to something else. I'm sure plenty of athletes do, if they're watching. They definitely understand that difference. But who is actually feeling that difference? Walk us through, right, that end user that's saying, I actually did need those milliseconds to be faster.
Austin Federa: Completely. So a difference of 70 milliseconds, which is what we provide on Hyperliquid, that's the difference between really making a trade and not. That is about half a block on any of these networks. And so the users who are clicking buttons on a user interface, if you're opening an app on your phone and trading, this is probably not something that's going to matter much for you.
But we see a lot of folks now who are semiprofessional traders. They may be running a $20,000, $50,000 book, and they've got some friends in a Discord. And I know it sounds crazy, but these folks on Kalshi are beating some of the big Wall Street firms at predicting CPI.
Kristin Myers: Wow. Okay. I was going to ask if we talked numbers, right, in terms of speed. Do you have any idea, monetarily, what the difference might be in some of these trades, or the average? In terms of winning that trade, what that might mean monetarily for an investor or for a trader?
Austin Federa: Completely. So the best analogy here is to look at the returns of HFT firms in general. When you see wild swings from year to year, sometimes there's a new alpha strategy they've opened up. A lot of the times, they found a way to get data or execution just a little bit faster. This is a market, generally, where most of the money is made by the fastest people. And that's what we see in traditional finance.
We see this in crypto as well. The fastest trading firms and market-making firms are making more money. And it also exists in an era where you don't have to comply with Reg NMS and NBBO and all these other things. So the pricing differentials can be quite different here. And if you see what happens today on something like Hyperliquid at market open and market close in US hours, there are huge amounts of arbitrage opportunities.
And so if you're able to move data even 5 or 10 milliseconds faster, from the pricing that's happening on NYSE and Nasdaq at open, or from CME to Tokyo, and trade on Hyperliquid, there's a lot of opportunity to be made there. And that is firms that are building custom programs to do this. But, I mean, I have built an AI-powered trading bot myself, just sitting on an AWS server in Tokyo. It's trivially easy to do. And that may be easy for me to say as someone building a company in this space, but I'm not an engineer by training. AI has really changed what's possible for all of us.
Kristin Myers: So I'm curious to know, because as you're mentioning, right, these big swings that we're seeing year to year not being about the strategy, it being about the speed, it seems as if it's an arms race that's being created in terms of getting to that pipeline, for someone who can afford a faster pipeline. Do you see it that same sort of way, that you're building a new arms race? Or, you know, giving access... I mean, do you find what you're building to be democratizing in a way?
Austin Federa: So we do, because the gap is narrowing, right? Right now, if you are trading on a user front end, or you're trading directly with the programs or the exchanges, there's a huge delta in what you can expect. One of the things we've seen with the advent of AI is that there are very smart people who have very good ideas on trades, but they don't have the technology to execute. And for years, that talent has been locked up at some of the biggest trading firms.
And now the democratizing isn't saying that anybody who has an idea on Twitter is going to go out and be able to make a lot of money trading. But you'll be able to test that idea and get the same class of data that you would if you were sitting in front of a terminal at a fancy Wall Street firm. And that's really a huge advantage for leveling the playing field. The big boys will always have an advantage, but it's about closing that gap.
Kristin Myers: Got it. So as you're building this company out, and as you're, you know, trying to grow, trying to scale, are you finding that the market is recognizing that it has this problem, or is there a level of convincing that you are having to do as you go out?
Austin Federa: Definitely. So we launched the company about two years ago, and the difference between then and now is very palpable. So two years ago, people were confused. Why would I need this speed? If I'm a crypto venue, why would I sell my market data? Like, crypto gives away all of its market data, and [unclear: "tried flights in"] almost $50 billion a year of business. And now it's different.
Now we have people who are building high-speed DEXs and applications coming to us saying, we would like to distribute our market data over DoubleZero Edge. We always had the HFT firms coming to us and saying, what's your latency? Can we get a demo of this feed? Can we try its performance? But we see now more and more just sort of average traders on Hyperliquid saying, hey, I think I've got an idea. I think I can code some stuff up with AI. How do I read the order book more quickly so I can take actions on it more quickly?
Kristin Myers: So say a little bit more about that, about some of your clients. You said you've been building this now for two years. When did you really feel like you got that first real yes? And is there a difference in the clients that you're seeing now versus the clients that you saw two years ago? And is there any sort of indication on the clients that you might get two years from now, compared to today?
Austin Federa: Completely. So we launched the market data product over the spring, and some of our earliest adopters were exactly who you'd expect them to be: large trading firms, or their crypto divisions specifically, or their new finance divisions that are looking to get faster market data access. And this is really a quirk of the way new finance is built, that there is not an ICE network you can just tap into and subscribe to data from the exchanges, like you can in traditional finance.
What we've seen recently is that that composition has shifted. We've added a lot more people that are at smaller trading firms, or a family office that is now saying, well, we used to just do mid-frequency, but maybe we have some opportunity in not exactly high-frequency trading, but arbitrage opportunities that maybe they wouldn't have taken advantage of before, because now they have AI helping power their execution layer, right? And that's been a really big change.
And now we're getting into just, you know, two guys that are trading a bunch of stuff on Hyperliquid, and they really feel like they've figured out something in the oil market, but they need better data to execute those trades. Those are sort of the new class of people that's coming. Looking forward, I wouldn't even say two years. I would say 6 to 9 months. Most of our users, I expect, will be AI agents, like agentic AI.
Kristin Myers: Absolutely.
Now, why would anyone build a private internet? Well, that's because the public one was not designed for this. When you send data across the internet, it takes whatever path is available. It might go one way this second and then a different way the next. And most of the time, that's perfectly fine, but in trading, it isn't. And it's not just speed. It's predictability. A route that's usually fast but occasionally slow is worse than one that is always the same. So consistency is key.
So traditional finance solved it with money. Firms built microwave tower networks between Chicago and New Jersey, because microwaves through air beat light through glass. They paid for cabinets inches closer to the matching engine. And that infrastructure exists for stocks and futures. Now, for markets that run on blockchains, mostly it doesn't.
We're back now with Austin Federa, co-founder at DoubleZero. So Austin, I want to pick up on what we were talking about just before the break. You were saying that the new clients that you would definitely be seeing in the next couple of months, within a year at least, I think you said 6 to 9 months, were going to be agentic AI. So AI agents: are you seeing those coming in from trading firms, from family offices? Who is utilizing that AI, and are they just using it to make the trades? Or are you seeing that AI pulling on that speed for other purposes as well?
Austin Federa: Yeah. AI in new finance, and traditional finance as well, has been a massive category. So one of the pieces that's really important to note is there's a huge speed differential between the matching engine at NYSE or Nasdaq and the matching engine at Coinbase or Binance or Hyperliquid. And so in NYSE and Nasdaq, we are still orders of magnitude away from AI being able to execute in a timely fashion.
But new finance is different. This is part of why this is being tried here. These systems are still comparatively quite slow relative to traditional finance, and what that means is the amount of time you have to wait for an AI to give you a response is a little less important. Now, when we say that this is going to be AI agent-powered trading, and that'll be most of the users of the DoubleZero network, there are a few different reasons for this.
This may be something as simple as what happens today, where you go to Claude or Grok or OpenAI and you describe a trade strategy you want to make. You give it your credentials or your wallet. You can connect it into DoubleZero data, and it builds you a trading program. And that program isn't always like asking a question and waiting on a response and doing that whole thing that we do. But that program has been built by AI. The next evolution of this is that AI, or inference, is fast enough that you can actually get a response directly from the AI within the time the market moves.
And so if you look at, like, Jane Street, one of the biggest purchasers of the new generation of AI chips, and most of these large trading firms have a huge AI division. A lot of that is for backtesting and analysis. But the holy grail will always be when the AI can make trading decisions itself. That will happen significantly more quickly in new finance than traditional finance, because the timing is so much more relaxed in blockchain and prediction markets than it is in traditional finance. So I would say look to see what happens with crypto trading and new finance over the next 6 to 12 months. That's going to show you very likely where we're going to see more of traditional finance going as well.
Kristin Myers: I was going to ask you on that timing. Do you think that we could see that within the next year?
Austin Federa: Absolutely.
Kristin Myers: You're describing those agents making the actual trades, because, to your point, that's the one thing that everyone seems to be going for next.
Austin Federa: Yeah. So right now, AI is a huge help in analysis and in coding. So you can describe a trade strategy: I would like to play arbitrage between these two venues. I have this thesis about where the market's going to be moving. I would like a long-short strategy. And, you know, if you hook it up to your Robinhood credentials, it can go build you that strategy already today. The difficulty there is you don't have access to fast market data, right? And so not only do you have this delay of your system, but you also have this delay of pulling data from the public APIs and interfaces. And that's where DoubleZero really comes in.
But there's another piece of this too, which is the agents can communicate with each other and communicate with the trading systems over the DoubleZero network, as opposed to going over the public internet. And so just like you can go to the cloud today and build a whole custom infrastructure deployment, and, you know, many companies do this, but also many AI agents do this. If you ask them to build you a website or something like that, they'll just deploy it on AWS and build the whole thing in the background for you. We can do the same thing with DoubleZero, because it's a fully programmable network. And this is already starting to happen today. I really think, when we look at the pace of advancement in AI, six months may actually be a conservative estimate.
Kristin Myers: I'm absolutely fascinated by that. But getting back a little bit to DoubleZero, because there are a couple of venues, and you've mentioned several of them: Hyperliquid, Kalshi, Binance. Are they buying that? What are they actually buying from you? Are they buying, to your point, that network? Talk to us a little bit more about that.
Austin Federa: Yeah. So the trading venues and blockchains themselves don't actually pay for access to DoubleZero. It's the consumers and traders that say, I want to know exactly what's going on in the centralized exchange, or get a fast feed of the Hyperliquid data. But part of the way that DoubleZero is built, different than all these traditional companies, is that all of the publishers, everyone in Hyperliquid that's sending us information, or on Solana, or with Kalshi, they actually get a cut of the revenue share. It's a revenue share model consortium, just like the exchanges used to be before they IPO'd and switched from nonprofits into publicly traded companies, or the way that consortiums like [unclear: "Go West"] are owned by the contributors.
This is really a network where the participants own the network, and that means that there's a much more win-win relationship there than you see in the sports betting world with sports market data, or something along those lines. And we think that's a really important idea, to rebuild some of this "networks owned by traders, built for traders" mentality.
Kristin Myers: So I'm curious to know. Binance, I hear Binance, and then Kalshi, and I think two entirely different beasts, right? Kalshi, very famously, essentially a gambling prediction market network, right? So talk to us a little bit about those differences. What do they have in common that makes them both come to you?
Austin Federa: Yeah. So one of the most popular instruments to trade on Kalshi is crypto futures and perps, as well as crypto up/down markets, these 15-minute "will Bitcoin be above this or below this within this 15-minute window." And, you know, you can look at them and you can say, oh, this is a prediction market. This is a totally new concept. And that might be true for elections. But these things are very similar to traditional futures and options markets. And so there's actually a lot of opportunity between multiple venues, the same way that you may say, oh, I've got futures on Coinbase and I've got perps on Binance, and if there's a spread between them, I have an opportunity between them.
And so what we also see are more and more traditional firms, as well as more and more individuals, getting into the prediction market space, because that's a place where they feel they have alpha that's not always based on speed. It's based on a complete picture of the market. And, you know, we see everything from ice cream shops hedging their weather exposure to individuals, right? If I'm going on a trip, it may actually be significantly cheaper for me to buy effectively bad weather insurance from somewhere like Kalshi, to say, will it rain in Italy when I'm going to be there? I can't buy travel insurance for the weather not being nice, but I might be able to do that through a prediction market.
And so we're seeing a whole bunch of different applicability here than we traditionally see. And in the prediction market space, these are even more opaque markets that have less good information structures around them. And so that's where we can come in and offer full level two Kalshi order book data to people over multicast on DoubleZero, and that's much faster than you can get it from WebSockets or with the FIX protocol. And it's also more instruments than you could possibly watch, because while there might be 50 assets you want to trade in crypto, there are maybe 3,000 or 5,000 different markets you would potentially want to trade in on a prediction market.
Kristin Myers: I'm going to try to get through these last two questions in about 60 seconds, so we're going to move fast. What happens when the market... excuse me, not the market. What happens when the network goes down?
Austin Federa: We have network outages all the time. Pieces of fiber go down all the time. There's scheduled maintenance. There's unscheduled maintenance. This is something that anyone who runs a fiber network deals with every day, day in and day out. The whole point of resiliency and redundancy is the system keeps operating, routes around problems, and keeps delivering data. The latency profile might change, but you will continue to get that data.
Kristin Myers: So, last question for you, Austin. I ask everyone this. We've got the bell over there. Have you thought about taking DoubleZero public and becoming publicly traded? What do you think about, if you do think about that day? And I also have to ask, when do you think that day might come?
Austin Federa: I think it's a really interesting proposition that more blockchain companies, or blockchain-powered organizations, are potentially getting access to public markets at the exact same time that public markets are now getting access to tokenized equity distribution through blockchain. And I think these can be very compatible and sympathetic ways of expanding market access, both in bringing more capital into crypto companies and bringing more accessibility into securities and equities that are traded in traditional listing venues as well.
Kristin Myers: All right. Well, we're going to leave it there. I'm absolutely fascinated by this. We're going to have to continue this conversation another day. Thank you so much, Austin Federa, co-founder at DoubleZero. Thank you so much.
Austin Federa: Thank you for having me today.
