Bloomberg’s Odd Lots podcast aired a studio interview on Oct. 15, 2026 with Hyperliquid Labs co-founder and CEO Jeff Yan, giving the market one of the most detailed public explanations yet of how the company sees the protocol, its deployment model, its U.S. compliance path and its longer-term ambition.
Hyperliquid, now the largest decentralized perpetuals venue by on-chain trading volume, has drawn broader attention this year after users traded crude oil around the clock during the Iran war and after Donald Trump publicly referred to efforts to bring the platform into compliance in the United States.
Not an exchange, but a financial base layer
Asked to explain what Hyperliquid is, Yan rejected the idea that it should be described as an exchange.
He said Hyperliquid is a protocol, and described the broader vision as one that can "house all finance." In his framing, the project is meant to serve as a neutral foundation for financial activity, much as electronic trading once replaced open outcry as a structural upgrade for markets. Hyperliquid, he said, is trying to deliver a similar step change for today’s fragmented electronic trading environment.
Yan pushed back on the common shorthand that casts Hyperliquid as "on-chain Binance." In his view, that misses the point. He said Hyperliquid is a blockchain protocol maintained by permissionless validators in a globally unified state network, while Hyperliquid Labs is only a small team building technology for that chain.
He also addressed Trump’s earlier public remark that U.S. regulators were working to bring Hyperliquid into compliance domestically. Yan said he did not know about that beforehand and was surprised when he heard it. As things stand now, U.S. users trading derivatives on Hyperliquid are violating the terms of service, he said, while adding that the current U.S. administration has taken a relatively constructive stance toward on-chain financial infrastructure and appears more interested in finding a way to fit the technology into the existing financial system than in banning it outright.
How markets are launched: 500,000 HYPE and a Dutch auction
The hosts then turned to the growing list of perpetual markets already available on Hyperliquid, including Nvidia, SK Hynix, the S&P 500 and crude oil, and asked who gets to launch them.
Yan said the preferred term is "deploy," not "issue," because "issue" carries too much traditional finance baggage.
To qualify as a deployer, a participant must first stake 500,000 HYPE tokens. Yan stressed that this means 500,000 HYPE, not $500,000 worth of HYPE. After staking, the deployer unlocks the on-chain deployment function and can define the oracle system used to publish off-chain pricing data to the chain for a perpetual contract.
That matters because, in Yan’s telling, a perpetual contract ultimately depends on just one number: the index price of the underlying asset. The oracle acts as the bridge that carries that off-chain price onto the chain. He added that if spot trading eventually moves onto Hyperliquid as well, oracle prices could in theory be derived directly from on-chain properties rather than submitted manually.
Could 10 different people all deploy a perpetual tied to Nvidia at the same time? At the protocol level, Yan said, there is no rule stopping that. The protocol is a toolkit, and deployers can use it as they choose.
What is scarce is deployment capacity. Every order book and every liquidation system consumes network compute resources, so access has to be allocated somehow.
Hyperliquid handles that through an on-chain Dutch auction. Each auction runs once every 31 hours, with the price declining linearly from a higher starting point until the first bidder takes the slot. Yan said the choice of 31 hours is deliberate: 31 and 24 are coprime, so the start time shifts across time zones instead of staying fixed each day.
He contrasted that structure with more conventional auctions, which he said tend to encourage late bidding and strategic information withholding. In a Dutch auction, the first bidder wins. After each round, the next auction’s opening price is automatically set at twice the previous clearing price, creating what he described as a natural price discovery process for the market value of a scarce deployment slot.
Oracle design and risk: a hard 1% constraint, not subjective judgment
The discussion moved to oracle and manipulation risk, with the hosts bringing up the Mango Markets attack as an example of the danger that can arise when perpetual trading volume overwhelms spot trading and an attacker can move the spot market, pass that move through an oracle and profit in derivatives.
Yan first referenced the incident involving SK Hynix, saying that was a problem specific to Trade.xyz. He said the deployer later published a detailed postmortem and compensated users. In his view, that did not point to a flaw in Hyperliquid itself; he compared it to a software bug that gets patched.
More broadly, Yan said he does not agree that a large imbalance between perpetual volume and spot volume automatically creates a fatal weakness. He pointed to S&P 500 futures, whose volume far exceeds that of the constituent cash equities, yet still function as a major price discovery venue.
For him, the key variable is not the ratio itself but whether the market contains "smart money" willing to fade prices that drift too far from fair value. If market makers are prepared to step in and trade against a 1% move they believe is wrong, then price discovery can happen in the derivatives market, with the spot market serving more as a reference point.
Asked whether the protocol would reject the launch of a market with very poor liquidity, Yan said the protocol does not understand concepts such as "bad liquidity." That kind of subjective assessment is not encoded at the protocol layer.
What the system does understand is a hard rule: the relative deviation between two oracle price updates cannot exceed 1%. If someone tries to submit a 2% move in one shot, the system will only accept 1%, and the remaining 1% has to be submitted again. Yan said that is a mathematical constraint, not a value judgment. He compared it to buying order book software: the software provider does not decide whether the asset being traded makes sense, it only supplies the trading infrastructure.
24/7 oil trading and funding rates
The hosts asked how Hyperliquid was able to support weekend overnight trading in WTI and Brent crude during the Iran war earlier this year, when traditional markets were closed and no standard cash reference was available.
Yan said the answer depends on how the deployer designs the contract, not on a protocol-wide rule.
He then made a broader point about perpetuals design. Most perpetual contracts today use a 24/7 oracle price as the reference, but he said that is not strictly necessary. The real anchoring mechanism for perpetuals is the funding rate. Even if prices move around, periodic funding payments can push the derivative price back toward the price of the underlying asset over time.
That means a different design is possible. During regular market hours, a contract could track spot. When the traditional market is closed, the derivative could float more freely and settle the difference later through periodic funding rate payments. Convergence would likely be slower, Yan said, but the mechanism would still work.
U.S. compliance: regulate the deployer, not the protocol
When asked what would need to happen before U.S. users could legally trade on Hyperliquid without a VPN, Yan repeated his warning that Americans using a VPN to access the product are violating the terms of service and should not do it.
He argued that the Commodity Futures Trading Commission’s core principles are technology-neutral by definition. In his view, the main issue is that regulators still need time to understand what this category of infrastructure is and whether it satisfies the principles they already apply elsewhere.
Yan said regulators want to support innovation, but they also carry real responsibility. The existing financial system is functioning, serves many people and comes with established protections, so regulators naturally have to work through whether this technology actually fits within those core principles.
As for where regulation should land, Yan’s answer was clear. He said he can imagine a world in which the protocol itself is treated more like TCP/IP — a technology used to build a financial stack, but not itself the regulated entity. Regulation, he said, should attach to the party with control. In Hyperliquid’s case, that means the deployer, the party setting rules and making decisions for a market, while the protocol remains the network and liquidity infrastructure layer.
He also argued that on-chain systems have a built-in compliance advantage because they are openly auditable. The sum of all user balances on-chain can be checked directly against the amount of USDC held at the relevant address. In traditional finance, banks have to produce books and submit to an audit process. On-chain, he said, that kind of provable solvency is native to the system.
From Hudson River Trading to a bootstrapped crypto build
The interview also covered Yan’s own path into crypto. He said he joined Hudson River Trading, or HRT, right after college and had a very positive experience there. Even though HRT was already a successful company, he said the day-to-day environment still felt like a technology startup. He said he learned a great deal about markets and met many talented people.
When he left to start Hyperliquid in 2023, he said the initial ambition was narrow: build something on-chain that people genuinely needed.
At the time, he said, crypto already had plenty of centralized perpetual exchanges with strong product experiences, which showed clear user demand. The weakness was structural. If users had to park funds on the exchange, they were still exposed to counterparty risk.
That is where he saw the opening: combine the product quality users liked in centralized perpetual venues with the self-custody model and lack of counterparty risk offered by on-chain systems. If that combination worked, he said, it would amount to a real product-level jump.
Yan said the original idea was not to build a system that could "house all finance." At first, the team simply wanted to build on-chain perpetuals. But as work progressed, they concluded that no existing technology stack could support that larger goal through small fixes or incremental improvements alone.
He compared the process to an internet startup that begins by building an application, then discovers it needs to buy servers, and later finds that even the network layer itself has to be rethought. Hyperliquid went through a similar progression, he said: what started as an effort to build an exchange eventually forced the team to build its own consensus layer.
Yan likened that stage to the early history of Amazon Web Services. AWS started as Amazon’s internal infrastructure before expanding outward. Hyperliquid, he said, is now in a similar phase. The scope is large, the technical challenge is large, and that is exactly why it is worth attempting. Looking across the blockchain sector today, he added, other chains may have their own strengths, but none besides Hyperliquid has pursued the very specific vision of a globally unified financial system in the same way.
Nodes, decentralization and hardware needs
The hosts also asked what kind of infrastructure Hyperliquid requires and whether the network depends on large data centers spread across multiple locations.
Yan said the reality is almost the opposite. He said there are now more than 30 validator nodes, with the top 27 by stake actively running the network.
He emphasized that the hardware requirement is lower than many people would expect. Referring to a post from one validator on X, Yan said the validator realized that a fairly ordinary machine — one that people can actually buy, without signing a three-year exclusive deal with Nvidia or assembling an extreme setup — is enough to run this global financial system. He added that the network has not yet reached the full scale implied by the vision.
Even so, he said the core principle is that the cost of operating the network should not become absurdly high no matter how many applications run on top of it.
There is still a limit. Yan said you cannot run a node on a phone. The current requirement is 128 GB of RAM. That is not the cheapest laptop configuration on the market, but it is a specification users can buy today, and he described that as the balance the team wants.
Anonymous founders, ecosystem scale and the end goal
Near the end of the interview, the hosts noted that the Hyperliquid team has kept a relatively low profile and that one co-founder remains fully anonymous. They asked whether that could eventually change, especially if users want to know exactly who is running a financial system of this scale.
Yan responded by saying the broader Hyperliquid team is much larger than Hyperliquid Labs itself. He described the system as more of a constellation, including participants such as Perp, which he called a Nasdaq-listed digital asset Treasury company, as well as deployers like Trade.xyz and many other actors helping run the network.
He said people may mistakenly assume the network is owned by its developers. In his view, it is not. The network should represent a system that is fair, transparent and supportive of these tools. Developers still play a role by turning community ideas into code, he said, but over time even that intermediary function could become less central.
When the hosts suggested that the endpoint sounds like TCP/IP — a system the world uses without most people knowing who created it — Yan agreed. Hyperliquid does not want to be remembered, he said. It wants to be used.