Jeff Yan 将整个金融行业带上 Hyperliquid 的计划 | Odd Lots

1. Closing Banter and Sign-Off

I hope you feel a little bit guilty about making everyone pay attention 24/7. We need a break. I think it should be an option. You can take a break!

I’m Joe Weisenthal.

And I’m Tracy Alloway.

Tracy, we’re recording this October 13th. In a couple of days, we’re going to be in Chicago and, as we do when we go to Chicago, talk to some legends of the trading industry.

Yeah, it’s great to go to Chicago. We’ve never done a floor tour of any of the actual trading places, and I know that we should do that.

2. The "Inverse Pyramid" of Financialization

Yeah, we should do that. But, of course, it’s always an exciting time to talk to the trader community because there are just so many interesting things happening in the world of trading, per se.

One of them is, of course, the rise of prediction markets. Essentially, I think we’re in this moment where so many different instruments can be expressed as other instruments in various ways. You could have stocks that are tokenized, and you can have tokens that are ETFs.

Yeah, like crypto tokens that are put in ETFs.

You can have prediction markets that actually functionally predict a stock, and then they become a stock. Essentially, you can replicate the economics of a stock. People are rejiggering the way you can express things with this incredible basket of new instruments and platforms to do it on. Financialization on steroids.

Can I go on a slight tangent for a second? I have this media theory that I call the inverse pyramid.

The way I think about traditional media, it was sort of like a food pyramid. You had original reporting as the layer at the bottom. You had old-school reporters making phone calls and getting scoops, and then maybe in the next rung you had some people writing editorials or opinion columns on them, with maybe a little bit of video eventually on top.

Nowadays, the pyramid is completely inverse, where the smallest layer is the original reporting, and there’s so much content built on top of that. I often think about finance today in those terms, in the sense that the underlying is kind of the smallest part of the market now, and there are so many new instruments built on top of it, whether it’s derivatives—

Yeah, futures and options.

—or now prediction markets and things like that. Derivatives have been a bigger market than the underlying for a while, but it feels like the inversion of that pyramid is just getting bigger and bigger.

That’s a great take, right? I didn’t know where you were going with that until I got there.

You think of something like a Skynet [?] or something like that. There was some company, and there is a stock in the trades, and then there’s an index, and then there are 3x leveraged ETFs, and there are tokenized versions of it, and so forth. Then there are prediction markets on where it will close in the next 10 minutes. I don’t know if that exists. If it doesn’t, it will exist. It probably will.

There is so much to do on this topic. Of course, we did that episode a few years ago with Austin Campbell talking about the potential for blockchain to unify and bring together a lot of different markets. I think it was one of the more compelling crypto conversations that we’ve had.

One of the things getting a lot of attention that combines all these things is this entity called Hyperliquid, which people who know it are obsessed with. It’s not a household name yet, but the people who do know it are very into it.

It’s not a household name, even though it’s been growing incredibly fast.

3. Hyperliquid Introduction and Mascot

The one thing I knew about it before starting prep for this episode was the mascot.

Yeah, which we have here, and it’s called Hypurr. I always approve of a company with a good mascot. So here we are.

Well, the company is a little bit known for being secretive. They don’t do a ton of press, so we’re actually going to have to interview the mascot today because we were—

No, no, that’s not true. Amazingly, we really do have the perfect guest on the podcast, someone we want to talk to for a long time. When I’ve told some people that we’re having him on, they’re thrilled.

Jeff Yan, the co-founder and CEO of Hyperliquid Labs, is here in the studio with us. So, Jeff, thank you so much for coming on Odd Lots.

Thanks for having me.

Why don’t you give us the basic view of what Hyperliquid is today? We’re going to go into this, but big picture, if someone says, “What’s Hyperliquid?” what do you answer them?

4. What is Hyperliquid? The Protocol Explained

Hyperliquid is a protocol. The tagline, or aspirational goal, is “housing all the finance.” I think the community is serious about that, so we can go into what that means. But the short of it is a neutral layer on which finance can ultimately happen.

It’s the same way that electronic trading was a step-function improvement over pit trading only a couple of decades ago. We view Hyperliquid, the protocol, as a step-function improvement over electronic but fragmented markets.

I know Joe said Hyperliquid is not necessarily a household name just yet, but it might be on its way to becoming one, because just a couple months ago Donald Trump was talking about Hyperliquid and talking about U.S. regulators working to bring Hyperliquid to the U.S.

At the moment, if you’re in the U.S., you are, quote-unquote, not allowed to use Hyperliquid. Although my understanding is that a lot of people get around it using a VPN. When Trump said that—and feel free to refute me on that; it’s just what I read on Reddit—were you aware that he was going to be talking about Hyperliquid?

I was not aware, so that was a huge surprise to me.

5. Trump Mention and US Regulatory Context

Hyperliquid Labs works on the protocol. I suppose it may be good to give a breakdown of the various pieces. There’s the protocol. It’s a neutral blockchain, a permissionless set of validators who operate a network collectively, which is a global view of a single, consistent state.

Hyperliquid Labs is a team that builds technology for this blockchain. We are a super-small team.

What I think the CFTC is referring to when they say they want to bring the protocol to the United States is that it is, in fact, against the terms of service of Hyperliquid.xyz and many other interfaces for U.S. users to use the derivatives trading on the protocol.

But the CFTC, or I think the current administration, wants to take a proactive approach toward on-chain rails for finance. It’s a kind of technological view that markets can be improved and that the U.S. should not play an extremely restrictive role in this, but rather should find ways where we can incorporate this technology into the existing financial system.

Let’s talk more about the ecology, or the ecosystem, of Hyperliquid and the players, because you mentioned the word protocol. At various times, Bitcoiners would say, “No, this is the new protocol, the Internet of Money,” and Ethereum people talked about how it’s a protocol. This word gets used a lot.

Why don’t you talk a little bit about it? It’s a blockchain, but it obviously has higher aspirations than a typical crypto project. I understand there are validators, deployers, and so forth. Give us the landscape of who does stuff on Hyperliquid and what their roles are.

6. Blockchain Protocol Mechanics and Validator Network

It’s a good question. This is a very deep question, and it’s a lot of stuff.

At a high level, a protocol is basically kind of what it sounds like. There’s an HTTP protocol, like a TCP protocol. It’s a messaging framework by which different validators, in this case, agree to communicate by these rules.

The interesting thing about a blockchain protocol, as compared to a purely informational-format protocol, is that the rules, under certain assumptions—in the case of Hyperliquid, that two-thirds of the stake of the network is what we call non-Byzantine, which is to say that they operate under the normal rules of the protocol—give you all these great properties.

These validators can operate independently. Basically, that’s all you need to assume. The network will come to a consistent view of the world.

This is really important for finance, right? It is the first layer. We’ve talked about the pyramid of things that need to happen for a global financial system. The very base layer of the information in the system is that everyone needs to agree.

If we don’t agree on what the ledger of balances is, for example, then all bets are off. At a minimum, we all need to agree that there’s this much in your account, there’s this much in Tracy’s account, and there’s this much in my account. We also need to agree on what the prices are.

There are some basic things that every participant needs to agree on. In a regulated exchange, that’s pretty simple: we trust this company, and they’re going to mark it.

But in a decentralized one, you need some mechanism to get independent validators to all agree.

Yes. I would draw the distinction that there is an orthogonal question of whether there is a regulated entity deploying the markets.

And in fact, I think there very much will be. It’s not an either-or. I think you can benefit from the resiliency of an on-chain network even if a single registered entity has permissions over the markets that they deploy.

The benefits of the technology are that, if you want to run a 24/7, 365-day exchange, it’s not a simple task. Any sort of mistake that you may make, whether it’s with the matching engine, the clearing, or any of the vendors in the stack, ultimately becomes a single point of failure for a very critical piece of infrastructure for the financial system.

If the rails on which it is built consist of a diverse, distributed set of validators, where one or even many can go down at any moment in time and the network still maintains its important properties—that everyone agrees on the state and that it continues progressing forward at the usual throughput—then all of a sudden, you have a much more stable foundation on which to build your regulated exchange.

Maybe just to hammer this point home, could you walk us through an example? Say I buy $1 million worth of a certain perpetual futures contract. What’s your most popular perpetual futures contract at the moment?

It actually varies. Between real-world assets and things like Bitcoin, I think right now it’s the perpetual on Bitcoin’s price.

7. How Trading Works: Perpetual Futures Example

Okay, well, let’s say I buy $1 million worth of perps on Bitcoin. Who is actually taking the other side? Is there collateral involved? Where is the collateral kept? Who determines the liquidation price when the trade ultimately unwinds? And who loses money if, for whatever reason, I can’t actually pay? I was going to say, if by some miracle Tracy makes money, who is on the losing side? Walk us through how that works.

Well, first of all, I just want to reiterate that as an American, you’re probably not in a fair position.

Okay, okay. Pretend that I live in Singapore.

Okay. If you’re a Singaporean living in Singapore or a European living in Europe—for anyone who lives in a free country—

How can I trade on my Austrian passport? This is interesting.

Oh, you know Austria.

Yeah.

The mechanisms, all of which you mentioned, are not so different from futures contracts on a traditional exchange. Perhaps the one that is interesting here is that there’s a counterparty in the order book.

The liquidation is set by the system, so it’s not subjective. It’s not like someone is actively setting liquidation prices. It’s a rules-based property.

As a property of the clearing system, the unique thing here is that the clearing system is on-chain. The threshold at which it happens and the mechanism by which the liquidation happens, which we can get into, are analogous to the traditional default waterfall in a U.S.-regulated futures exchange, for example—analogous, but with different names.

The execution of what is called HyperCore, which you can get into as well, is kind of like a submodule of the blockchain state itself. It autonomously carries out the mechanisms of clearing and liquidation.

Let’s talk about that. By the way, I think a classic sign that something is going to be a big deal, in my opinion, is when Bloomberg starts listing it. Years ago, I think it was 2012 or 2013, the first time that on the terminal you could look up the Bitcoin price, it was like, “Oh, that’s probably a sign that maybe something is going to happen here.”

Recently, there is now a HYPE page, and you can look at 24/7 things. They trade Brent there, and they trade WTI, the S&P 500, Circle, SanDisk, Micron, and a bunch of other stuff. Of course, it’s got a lot of engagement, I see.

Yeah, it’s super cool. No, it’s a big deal. Kudos to you all for integrating the technologies.

Okay, so now here’s a key thing. I’m looking at, let’s say, an NVIDIA perp. It trades on Hyperliquid, and I believe it is issued by an entity called TradeXYZ. So let’s bring in who TradeXYZ is and how they got into a position where they can issue an NVIDIA perp. Walk us through how that emerged on Hyperliquid.

8. Deployers, Oracles, and the Dutch Auction System

We use the word “deploy” because I think “issue” has many connotations. This is a technical process by which anyone can use the network in different ways.

On the Bitcoin network, you have the same kind of foundational idea: a globally consistent state and a distributed set of participants. But on Bitcoin, you can move Bitcoin around. That’s all you can do. On Hyperliquid, it’s a lot more complex, as a system that ultimately can power the financial system needs to be.

One of these functionalities is that players can come and stake some amount of the native token.

How many HYPE do they have to buy to become a staker? How many do they have to accumulate?

They have to stake 500,000 currently.

HYPE—not $500,000 worth of HYPE, but 500,000 HYPE coins. Okay, keep going.

The stake is basically a way for them to unlock these deployment functionalities that are implemented on the blockchain itself. They can effectively define an oracle, or more of an oracle system.

Suppose it is a mechanism by which a perpetual contract tracks the one number that is exogenous to the system, which is the index price. That’s kind of the beauty of a perpetual, right? It’s so powerful, so expressive, and yet it ultimately boils down to one number.

And this is really important, right? Because with a physically settled future, the future ends and you get the oil at the end. With a cash-settled future, there’s some entity and it’s like, okay, the future expires. With a perp, there needs to be an oracle. It never really ends, but there needs to be a price oracle that’s always going to be, I guess—

Why is it called an oracle as opposed to just the underlying price?

Oracle here has the meaning of the word “oracle” in a blockchain context. It’s something that is published by a single publisher or multiple publishers that tracks something that is off-chain. It is not knowable from purely on-chain characteristics.

This will change in the future because I think as spot markets move onto Hyperliquid, you can imagine a system where the oracle is actually no longer published, but rather derived from ongoing prices.

Okay, just to finish my question about how the HYPE or the NVIDIA perp got there: TradeXYZ is a deployer. They accumulate the HYPE so that they’re allowed to deploy. How do they get the right, or the ability, to actually issue the NVIDIA perp? How did that happen?

It’s a simple matter of sending a transaction to do the deployment. The thing here is that they’re inheriting the technology of the order book and the clearing.

Yeah.

The existence of USDC itself, in a way that’s well connected to the entire Circle infrastructure, is another example of something that a deployer inherits. They simply send the transaction designating the specifications of the contract.

They, as suppliers, push this. In TradeXYZ’s case, I believe there is a system of publishers of the oracle price.

The magic of it is that it is truly a distinct set of responsibilities between the protocol and the deployer, but together they form this incredibly compelling ultimate product for the end user. This is the idea of modularity that I think is unique to on-chain financial systems.

Where does the liquidity come from? This is a classic issue for new exchanges and new markets. How do you make sure that there’s enough money floating around to take the other side of a trade?

The liquidity is just a free market. It’s an order book. Of course, whoever places an order must have the collateral necessary to back the order. But other than that, the protocol is simply the mechanism by which buyers and sellers come together.

I’m trying to come up with an analogy here. It feels to me like, to some extent, the deployer—like TradeXYZ. But there are others. We keep mentioning them. There’s TradeParagon, and there are some others whose names I don’t know.

They’re a little bit like an ETF issuer, right? In a traditional stock market, they’re a little bit like a deployer because some companies have single-stock ETFs, et cetera. They’re responsible for this instrument maintaining a price relative to some underlying thing.

In the case of an ETF, it’s usually a basket of constituents, et cetera. But the thing I’m still trying to understand is: how can people issue an NVIDIA perp? How did they get this particular slot in the system, as opposed to someone else?

Anyone can do it. There are no protocol-level semantics about what the perpetual is. You can view the protocol as a kind of basket of tools, and a deployer can come and use it as they see fit.

The second part of your question is: what is the mechanism by which this kind of scarce resource might be allocated? That is governed by an on-chain mechanism. It’s a Dutch auction.

In short, it’s a way to allocate the scarce resource of deployment slots. This is scarce because each order book—each clearing system for a specific market—takes computational resources of the network.

As the network scales, I think these resources become less scarce. But ultimately, they will always be a limiting factor for anyone who wants to deploy on the system.

The Dutch auction is a way whereby, in every 31-hour interval—currently, 31 is simply chosen so that the time at which the auction resets changes time zones, as it is relatively prime to the number 24.

Approximately once a day, there’s an opportunity for any deployer to deploy something. The Dutch auction is nice because normal auctions have a few undesirable properties. The incentives are a bit odd, so you’re incentivized to penny the other bidders, withhold information, and spam it at the very end. It has all these interesting, I would say, suboptimal game-theory properties.

Dutch auctions are nice, especially for launching systems, because they go in reverse. The price starts at a value and decreases linearly. It’s spam-resistant because the first person who bids gets it. The price decreases in this case linearly, although I suppose it doesn’t have to be linear, and it resets to twice the last clearing price. So there’s a natural correction mechanism by which it approaches the fair market value of the scarce resource.

Do you get excited to see who wins those auctions? Do you have any idea of, “Oh, they plan to do this, or issue or deploy whatever, if they win this auction”? Or do you have an idea that this particular deployer has come up with interesting ideas before?

9. Permissionless Deployment and Market Access Philosophy

I don’t know. It is a permissionless system, so I don’t personally track it. I know that people have built tools where you can see some pretty cool stats. This is another nice thing about chain systems: the data is transparent, verifiable, and auditable. Anyone can build these really cool analytical tools about how these deployers tend to build and how they use the tool in specific ways.

But I’m honestly just too busy to look into this stuff in depth. I do think it is very cool, broadly, to see what is being done—taking markets that are traditionally confined in various ways and extending them, making them more broadly accessible. Whether that’s operating 24/7 or just giving retail investors, or everyone around the world, more equal access to the same financial opportunities, I’m a really big believer in that.

I think many people, whether they’re building on-chain or even working in traditional settings, are working in finance toward the same end goal. We’re really just taking different routes, but we’re all trying to make incremental progress. Ultimately, the thinking is that the financial system is there to better serve as many people on this planet as possible. Many layers removed, this permissionless deployment system is ultimately accomplishing that goal.

One example is IPOs, especially this year. If you look at them, they’re a really great example of how wealth is being created, and tremendous amounts of wealth are being created. But it’s being created in a way where it’s gatekept. I don’t think this is done intentionally. I think it’s just a byproduct of how the markets work right now.

If you have trillion-dollar, $2 trillion, or $3 trillion companies IPO, you can pretty confidently say that most of the wealth creation has happened at that point. So I think it is a failing, or a limitation, of the existing system if, up until that point, very few people have been able to trade these opportunities.

I guess one question that I always have with this type of stuff is: if you’re buying shares in Anthropic or whatever when they IPO, you’re getting a claim on the company’s dividend, right? Assuming they have a dividend, there’s going to be some free cash flow, right? The net present value of all future cash flow.

Whereas if I’m an unaccredited investor purchasing something that’s tracking Anthropic shares on a platform such as Hyperliquid, I’m not getting that cash flow.

But you’re not getting the cash flow if you’re trading futures on something as well, right?

So I’m not saying that it’s a strict improvement over the system, and I’m not saying that it actually makes sense for everyone to trade these things. It’s more that there are a nonzero number—there are, I think, an order of magnitude more people who could benefit from something.

Of course, there’s a parallel track of work, which is that we should all be working hard to educate as many people as possible so that financial opportunity is just that: an opportunity. Whether it fits someone’s specific risk profile is more a question of education. But if the opportunity itself is completely gated off, then that’s a nonstarter. There’s no way, and no amount of education, is going to get you out of that.

So I think you need both. Right now, there is just so much work to be done, but I think the first step is taking the system itself and making it fair and democratized.

We’ve been talking a lot about the market structure of Hyperliquid, and I would like to get back to that, but maybe people would be interested in a little bit about you and your background. You used to be at Hudson River Trading. You started Hyperliquid a few years ago, after the insane 2021 year that crypto had. What were you looking out into the world and seeing? Most people might have thought it was a contrarian bet, that crypto was just a bunch of meme coins and stuff, and that was over. Why don’t you talk to us a little bit about this choice? You didn’t raise any VC money. Talk to us about this choice to pivot your career and take this big shot at building something new. What did you see out in the financial system such that you thought there was this big missing piece?

10. Jeff Yan's Background and Motivation

Hudson River Trading? I worked there right out of college and had an amazing experience. I only have positive things to say about them. They’re obviously a wildly successful company, but when I was there, you didn’t feel that day to day. It felt a lot like a startup, like a tech startup. I think they worked hard to have that culture. I learned a lot about markets and met great people.

With building an on-chain system in 2023, the vision was a bit more laser-focused and narrow: it was to build something on-chain that people actually need. When we looked around and just looked at what people were doing in the world, and where on-chain technology could really improve things, we didn’t see it as, “Crypto was only about meme coins,” or whatever.

There were these exchanges offering perpetual futures in a centralized context, and they were very clearly offering something that users benefited from and wanted. There was amazing product-market fit. I think now you see this in traditional contexts, where many people have crossed that threshold of acceptance, which is really cool.

But even in 2023, there were many venues in crypto offering perpetual futures, and they were all doing it in a centralized manner. So it was very clear to us that there was an opportunity to bridge the gap between the user experience and product experience of a centralized perpetual venue, on the one hand, and the on-chain technology and properties such as custody of your own capital, on the other.

I think the counterparty risk of using an exchange is always going to be there, whether it’s in a crypto or non-crypto context. If you could have your cake and eat it with those two properties, the idea was that this would be a step-function improvement over what people had access to at that point.

11. Future Vision: Housing "All of Finance"

How serious are you when you say you want to house all of finance? What are you looking at next, other than perps? Could we go old school and have physically settled commodities futures or something? Is that possible for you?

I think we’re very serious about the idea. Of course, it is aspirational. Even 90% of all finance would be a huge success, and we’re a long way away.

Physical settlement is always going to have a very important place in society. I think this is a misconception about perps and, honestly, about our general vision. We’re not trying to displace anything.

In fact, I think the technology can benefit everyone who currently operates markets in a more traditional setting. Whether these markets are perpetual markets or settled-dated futures is irrelevant, because ultimately what you really need is an order-book clearing system and global distribution and concentration of liquidity. All these properties don’t really matter, regardless of what the underlying is.

I’m personally not very opinionated about what ought to exist. I think the system to house all of finance should be a collection of tools. We call these things primitives because I think it evokes more interesting questions about how these things can compose with each other. But effectively, they’re tools for people to do what they need to do.

It’s funny, there’s no Hyperliquid Wikipedia page. I don’t know—I looked because I was trying to see. Maybe there was one and it was deleted. One of the moderators said, “Typical crypto spam.” Very doubtful.

This is a notable topic, and the page was deleted. But if you guys have 90% of finance one day, it sounds like my page, which got deleted because I was a journalist of little notability. If this person gets to stay humble, I’m not going to call this person out by name, but this moderator is going to feel pretty dumb one day if you guys host 90% to 100% of finance and he’s like, “Very doubtful. This isn’t notable.” Anyway, we’ll be revisiting this page in the future.

So you decided to build it, and you coded it yourself. This was when coding was still hard, because now anyone can code. Can you talk a little bit about what that was like—building an exchange at your computer, how that all worked, and how long it took?

Well, we don’t view it as building an exchange.

That’s great for you, right? You’re building kind of a protocol and network.

Yeah. The important thing here was that we looked around and tried to find something existing at that moment in time that could effectively, with an incremental improvement, house all of finance. The conclusion, unfortunately, was that there were many technological gaps.

It’s a bit like if you were building consumer tech in the early 2000s. Maybe you realize you go down this rabbit hole of, “Oh, wait. We’ve got to buy computers. Maybe the internet doesn’t actually work that well. Maybe we should work on some networking things.” This is where the vision of an AWS kind of analogy came up.

I really think that the thought process, realizations, and ultimate generalization of the infrastructure probably have some parallels between these two stories. Hyperliquid, of course, has had a much earlier stage and smaller scale, but that was effectively the realization: “Okay, I guess this is a much bigger project to do well.”

It’s a very big project, and maybe it was a bit of blind optimism. I think the fact that it was hard was almost more inspiring to me. It made it seem more worth doing, and it felt more like, “Well, if we don’t at least give it a shot, who will?”

It involved digging way down into the consensus layer—just how these different nodes talk to each other—and questioning all the assumptions about how networks were built at that time to come up with a solution that ultimately can house all of finance.

Because if you look around at even blockchains today, they’re very admirable, and they all have their own selling points. But I don’t think any of them, other than Hyperliquid, has this precise vision of a global financial system that can actually just have one global state.

12. Technical Constraints and Validator Requirements

You mentioned computational resources as a constraint on listings, deployments, or deployment auctions earlier. What’s been the evolution of your data needs? I’m really curious how big you guys are now. Are you running 18 data centers that are all the size of three footballs, or whatever cliché people like to use? What’s your token spend like now, and how much of a constraint is this really?

That’s a good question. I think this is also a misconception. First of all, we actually don’t run these things. The foundation runs some validators, but ultimately the current set of validators is 30-something, and the top 27 of them by stake, at any moment in time, operate the network.

The specifications for running a node are actually very minimal, which I think is quite cool and counterintuitive. I remember a validator at some point tweeted something like, “Wow, my mind is blown. I just realized that on this machine”—a very humble, rudimentary machine that anyone can procure—“you can run the global financial system.” Of course, we’re not there yet, but that is the vision. It doesn’t matter how many things are built; the network should not be prohibitively expensive to operate.

That being said, you can’t run it on your cell phone, so there is some sort of trade-off here. I think a reasonable approach is that it should not be, like you said, data centers the size of football stadiums to run this thing. Right now, you need 128 gigabytes of RAM. That’s not the cheapest laptop you can buy anywhere, but you can buy a laptop with that specification. That’s kind of the sweet spot.

13. Governance, Team Anonymity, and Neutrality

You have this vision. Maybe you get to 100% and swallow all of finance, which is pretty cool and very ambitious. Hyperliquid sells itself—well, you don’t do a lot of media, which is why we’re thrilled to have you on. I believe you have a completely pseudonymous co-founder. There’s not much known about your team in general. I think people know more about the mascot than they do about you. People don’t know if there’s a specific address or whatever that they could call.

Do you envision that eventually changing? Do you think there’s going to be a demand for, “Look, we’re running a lot of the financial system through this platform. We sort of want to know the people who are running this”? Or do you envision it as a kind of supreme neutrality, where we don’t know who’s running the internet and we don’t know who’s running Bitcoin? Do you see yourself and your team truly fading out, so it’s just a protocol like TCP/IP?

I actually view the team as much bigger than Labs. Who runs the thing? It’s really a collection, a constellation of different teams. We mentioned them earlier, right? The Nasdaq-traded ETF, or digital asset treasury company—they’re part of the team. They’re the team going to investors and explaining what this technology is. Dave is very well known.

I think this is a miscategorization of who runs the thing because, like you said, who deploys the market? It’s the deployer. There are many deployers, and there will continue to be more.

Kraken announced earlier, for example, that you set the rules for HIP-3. You set the oracle for pricing, the leverage limits, settlement, and everything else. You are a governing entity.

This is no different from the Ethereum blockchain, where the developers govern 90% of the world’s finance. If the world is eventually going to have all finance be on this chain, people are going to expect to know more about the people behind it. Ethereum has gone sideways for several years. If the world is going to be on this thing, do you think people are going to expect to know more about the people running it?

I guess what I wanted to say there was that the Ethereum blockchain is a collection of developers taking community feedback, but ultimately it has to be written by developers. Unfortunately, maybe you’re right: Maybe at some point software will no longer require developers to oversee the conversion of thoughts and ideas—a very democratic process—into code. Maybe there will be no human intermediary in that process in the future. But for now, there is, and there’s this role that developers play.

I would not say that the core ideas behind the network are owned by the developers. I would say it’s a manifestation of what a system should be. It should be fair, it should be transparent, and it should support these tools.

Maybe a slightly different way of asking this question is: With HIP-3, a builder can deploy a market, set the oracle for pricing, set the leverage limits and settlement, and everything else. I trust the builder to do all of that, and it’s all laid out very clearly and transparently. What exactly am I trusting Hyperliquid to do in that scenario? What provision is Hyperliquid making for me, and what am I hoping you will do?

If the question is about Hyperliquid Labs, I think the trust assumption is in the output of the software—the node software that exists. There’s an assumption that it works and adheres to the specification. I think this is true of any blockchain. Ethereum has a specification and multiple client implementations. When you run a node, you pick an implementation and run it. I suppose you are trusting that implementation—or you can read it, but generally you’re trusting this very complex, millions-of-lines-of-code system—to adhere to the rules of the blockchain.

As you said, all of the quote-unquote running of markets is done by a combination of the deployer specifying the specifications of the contract and publishing this oracle price, which is tethered to some real-world thing. The technology itself is sort of a reflection of the deployer’s intention.

I don’t think there’s any inherent trust going the other way, from the protocol to the deployer, because users who use a deployer’s market to trade are trusting the deployer of the market. That’s where the trust is. If you’re trusting a contract to track the price of gold, then you are implicitly trusting that the deployer is publishing something that represents what they said it would.

I don't know if this is a great analogy, but with the rise of ETFs and passive investing, there are all these questions about how far that can go. Don't we need some people to trade the underlying thing?

I'm curious if there's some sort of analogy with oracles. Several years ago, there was an incident with a Solana-based thing called Mango Markets, where someone was accused of manipulating the underlying token. The size of the trading relative to the underlying token got so extreme that it was very easy to manipulate the underlying price that was transmitted through the oracle. Then you make a lot of money because the thing moves price.

What are the constraints on the volume of perps relative to the underlying trading? Does that create a constraint for Hyperliquid? There was also an incident earlier this year where there was a bad quote from SK Hynix, and the oracle price deviated very dramatically, causing some losses. Talk to us about the oracle constraint.

14. Oracle Constraints and Market Manipulation Risks

I'll take those in reverse order. The SK Hynix thing—I think TradeXYZ published a very detailed postmortem, so I'm just quoting here—but that falls into the category of an implementation issue. Their system kind of worked as intended, but their system perhaps was not intuitive. I think they indeed found it was not intuitive. There were refunds, and ultimately they sorted it out.

These things happen. This is not unique to Hyperliquid, TradeXYZ, or even blockchain. This is just a software property. Apple ships operating systems with bugs, and they patch them.

The first question is interesting. A priori, you might assume that the volume ratio needs to be capped. Maybe you would think, “If there's more volume in perps, then it doesn't work because the underlying trades less.” But I don't actually think that's true.

I think it can be healthy. If you look at, for example, futures tracking the S&P 500 versus the underlying components, or even ETFs, there's a big volume differential there. The important thing is that the underlying needs to have smart people—the right people—with the ability to provide liquidity into price moves that do not reflect the underlying value of the thing.

This is where the market design, the mechanism, is really important. You can imagine a world where there is perfect information about the underlying companies, and someone trades a bunch of the perps because there's a lot more volume going on there, or the futures for the S&P 500, and then the underlying is traded very thinly. The price moves 1%. The fact that it's traded thinly does not really matter. What matters is that someone needs to say, “That 1% move is not correct. Whatever size you want, I'm here to trade against you.”

If you look at the S&P 500, I'm very confident that there are market makers who would happily trade at far less of a dislocation. You get this effect where price discovery can happen on a derivative, and the underlying can still be a reflection of some real value in the world.

Can deployers ever get rejected for having an oracle because you think it wouldn't work because the underlying isn't liquid enough, or it's too hard to trace a price, or something like that?

The person does not even understand what that means. There's no semantic encoding in the protocol. What the protocol does know about is, for example, that there's a constraint at the protocol level where, between two oracle price updates, the relative difference between those two prices cannot be more than 1%.

This is a hard-enforced rule. In fact, when you try to push an oracle price that's 2% away, what happens is the protocol clamps it to be 1% away, so then you can push it again. It's relatively user-friendly.

You might think this constraint is important because, given the sort of leverage in the system, if you move it too quickly, perhaps the system cannot mutate its state machine in an orderly way. But if you break it down into a composition of something that's more like a continuous price movement, then it's fine.

These kinds of mathematical constraints—the protocol very much has a lot of intentional mechanism design around them. But whether it's, “This thing doesn't make sense,” I just don't think that question can be answered. We can't do subjective liquidity judgments.

In the same way, if you're running a traditional exchange or futures exchange and you have some vendor that provides the order-book software, that vendor surely cannot know whether the order book represents something appropriate. They say, “Here's the technology. It allows people to place buy orders and sell orders. You can specify a price and a size. This is how it works.”

15. 24/7 Trading and War-Time Price Discovery

We mentioned at the beginning that Hyperliquid is not a household name yet, but it is getting there. It did get some attention this spring after the war in Iran broke out, because here is a venue that had 24/7 WTI and Brent futures trading on it.

People are sickos, and they want to know—or maybe they're not so sick—they really want to know what's happening with the price of Brent at 1:00 a.m. on a Sunday morning. Where's the price coming from? There's no oracle transmitting it, right? There's no actual live spot-cash trading or anything.

Talk to us about what's happening at 1:00 a.m. on a Sunday such that there is a price of a WTI perp.

This is up to the deployer, kind of like the previous question. The semantics of what it is are up to the deployer to define. This is part of the specification of the contract, not in the protocol.

An interesting thing to point out here is that the fact that perpetuals today, by and large, have an oracle source based on an underlying price 24/7 is interesting, but it is not necessary.

You could actually have funding. The mechanism of funding is basically to make derivative trading, over time, converge to whatever the underlying price is. You can get that effect with periods of free-floating derivative prices and then periods of funding payments based on the discrepancy between the derivative and the underlying.

If you work out the math for funding, you can relax that constraint. I think it will converge more slowly if there are periods with less funding, but I think the mechanism itself still accomplishes the goal.

Do you have any institutional traders on the platform now?

I think many of the larger traders are institutions. It requires significant sophistication to, for example, market-make in any setting.

You're here in the U.S., and presumably you want to get a lot more finance. So presumably, at some point, if that's going to happen, there are a lot of legacy institutions that may be interested in figuring out how their business can work with Hyperliquid, beyond just the trading shops, which are probably active.

Can you talk about the volume today? What do you see? Who do you see in 5 years? Could you see the banks of the world, or the Roundhills or State Streets, becoming deployers and issuing stuff?

When you think about where Hyperliquid could become, could you imagine some of these household names slotting into some part of the ecosystem here?

16. Institutional Adoption and Regulatory Framework

I would push back on them being “legacy.” That's very much not the vision of Hyperliquid, at least not how we are seeing an ideal future.

“Legacy” implies that they will be disrupted or displaced. The vision here is that the institutions you mentioned have great relationships with a bunch of their users. The users want to use them. They trust the brand. They provide a valuable service. I don't think any of that should change.

But I think, to your latter point, it's in their interest to adopt new technology that improves their offering. There's an analogy here with the cloud and AWS.

I don't know if, back when AWS first came out and went around telling people, “This is great technology. Why don't you consider using the cloud?” the companies said, “Wait, we built our own servers. Are you trying to disrupt us?” I think there could have been that pushback.

Fast-forward to today, and basically you could see the very same institutions eventually becoming deployers. I don't even think that's remarkable. I think it's an important development, but I don't think it's remarkable that that could happen.

This is a slightly weird question, but if you were to be regulated by the CFTC or the CME, could you see a world where the deployers are the ones who get regulated versus the protocol?

I can very much see that world. I think that's an astute way to frame it. It's a good understanding of how, as we said, there are different roles in the setup.

And I think in the same way that internet technology—TCP/IP, for example—is a technology used to build financial stacks, it is not regulated. The regulation sits at the layer where there are controls. Indeed, the deployers are the ones that have these explicit controls. So it’s a very natural way for the protocol to continue playing its role as a networking and liquidity infrastructure layer.

17. Path to US Accessibility

As Tracy mentioned, President Trump called out Hyperliquid. But what actually would have to happen so that an American could trade on Hyperliquid in the normal way without having to use a VPN? What needs to happen from a regulatory standpoint for that to happen?

Yeah. Well, again, if you’re an American using a VPN, you’re violating the terms of service. I think—don’t do that. This is strictly not a laughing matter. You’re breaking the rules.

For it to happen, the CFTC in this case needs to—well, it’s a good question what they need to do. In some sense, the core principles of the CFTC are agnostic of the underlying technology. That’s why they’re called core principles, not technological rules or something.

On the other hand, I think there is a certain gap in understanding what is hot off the presses, in some sense. Regulators are conservative, obviously, and I think to everyone’s benefit, they want to move quickly and embrace innovation. But there’s a very heavy responsibility with that, because there’s an existing system that, by and large, works well, serves many people, and has protections built in place.

I think there’s naturally this gap in wanting to understand whether this technology is truly compatible with our core principles. Ultimately, I think the technology has a lot of benefits, and many people are beginning to see this. The players see this because they’re using the technology day in and day out. Regulators are more conservative and slower to see it, but they’re beginning to see that if anyone in the world can provably audit that the sum of user balances equals the amount of USDC sitting at this address, that’s actually a really good property.

Even in traditional settings, there are a lot of regulations in place, but there’s kind of a trust there, right? If someone asks, you have to tell them: “Here’s the bet. Show them the bank account. Show them the ledger of user balances.” But that is a sort of process. Whereas if this is a property rather than a process, all of a sudden, I think this moment is quite beneficial. So it’s not like it’s compliance or new tech.

I hope you feel a little bit guilty about making everyone pay attention 24/7. You need a break. I think it should be an option. You can take a pill. All these poor financial journalists are working on weekends now. I like being able to check quotes on the weekend.

Jeff Yan, thank you so much for coming on the podcast. There are many more questions and things to dive into, but that was a fascinating introduction. Please come back sometime.

Thank you for having me.

That was a lot of fun. I feel like we threw a lot at the listeners there. Some of these terms—I would recommend that people who are listening do a little more research on what these things are, particularly how perps work.

Yeah. If you’re in markets, if you care about futures, et cetera, I think it’s very much worth knowing a little bit about the market structure and perps, how they’re funded, and so on. Because regardless of what happens with Hyperliquid specifically, this primitive of perps has been growing for a long time, and I think there’s very good reason to believe they’re going to be with us for a long time as a popular trading instrument. It feels very inevitable, as does 24/7 trading at this point.

Do you have a take on why so many of these crypto things end up inventing their own language to go along with it? Like deployer or oracle?

Yeah, it’s a good question. Maybe I don’t know. I would imagine that the idea of a price oracle probably existed in computer science before the trading world. I don’t know. It’d be interesting to know.

But the structure of Hyperliquid to me clearly feels distinct in many ways. One thing that we didn’t get into is that, for example, Ethereum exists and there are different exchanges that exist on Ethereum. Uniswap exists on Ethereum. One thing that’s interesting and distinct about Hyperliquid from Ethereum is that the protocol itself sets the rules for the exchanges.

This idea that there’s a 31-hour interval in which a deployer can bid to list an item, and that this is baked into the HIP-3 version of the protocol, is very interesting and distinct. The fact that the lab takes active governance in how these exchanges are run is also interesting.

We know, for example, that on Solana or Ethereum, each probably has 10,000 dog coins. Those don’t exist on Hyperliquid because there’s an auction process. You can’t just spam the network because dog coins are hot and decide to list dog coins.

I think one other thing we didn’t get into, but I think it’s interesting, is: let’s say you and I assembled 500,000 HYPE tokens so that we could then be a deployer and bid into these auctions.

Yeah. There are already Nvidia perps. There are already SK Hynix perps. For us to have a successful thing, part of our strategy is figuring out what’s going to be the next big thing. Maybe Central Asian ETFs, like the Kazakhstan index, are going to be hot.

Yeah. So maybe our strategy is, “Let’s build those.” But there’s probably this whole other metagame. Maybe at some point we should do an episode with one of the deployers about how they think about what to bid on so that their instrument becomes the thing everyone wants to use.

That’s it. Lots of interesting questions.

Oh, totally. Lots of game theory.

Yeah, totally—deploying those new markets.

But, shall we leave it there, for now?

Let’s leave it there. This has been another episode of the Odd Lots podcast. I’m Tracy Alloway. You can follow me @tracyalloway. And I’m Joe Weisenthal. You can follow me @thestalwart. Follow our producers, Carmen Rodriguez @carmenarmen, Dashiell Bennett @dashbot, Cale Brooks @calebrooks and Kevin Lozano @kevlloydlozano. And you can chat about all of these topics 24/7, including markets and crypto in our discord, discord.gg/oddlots. And for even more online content, check out our daily newsletter, which you can find at bloomberg.com/oddlots. And if you enjoyed this conversation, if you like it when we talk about the new rails of finance, then please leave us a positive review on your favorite podcast platform. And if you watched the video and you enjoyed the show, then please hit the like button or leave a comment. Or better yet, subscribe. And remember, if you’re a Bloomberg subscriber, you can listen to all of our episodes absolutely ad free. Just find the Bloomberg Channel on Apple Podcasts and follow the instructions there. Thanks for joining.