Jeff Yan's Plan to Bring the Entire Financial Industry Onto Hyperliquid

Jeff Yan’s ambition is to bring finance onto Hyperliquid, a goal he says remains a long way off: even putting 90% of finance on the network would be a major success. The premise is that financial markets form an “inverse pyramid,” with underlying assets supporting a much larger range of derivatives and other instruments. Yan sees Hyperliquid as a neutral, shared layer for those markets, and as technology that could improve traditional firms’ offerings rather than displace firms whose customers value their relationships, brands, and services.

Hyperliquid is a blockchain where permissionless validators agree on one network state, including account balances and prices. Yan says the safety assumption is that two-thirds of the stake is non-Byzantine. The team chose to build a network, rather than just an exchange, because existing systems had technological gaps; its vision of a global financial system with one global state required work at the consensus layer, including rethinking how nodes communicate. The result is intended to be a set of composable financial primitives. The validator set has “30 something” members, with the top 27 by stake operating the network at any given time. Running a node requires 128 GB of RAM, a constraint Yan says is meant to remain manageable on consumer hardware, though the system cannot run on a phone and has not reached the scale of its full vision.

Anyone can deploy a perpetual market, but deployment slots are auctioned because markets consume network computing resources. Every 31 hours, an on-chain Dutch auction starts at a price twice the previous clearing price and falls linearly until a bidder wins. A deployer currently must stake 500,000 HYPE to unlock deployment functionality. Under HIP-3, the market builder specifies the oracle, leverage limits, and settlement, while the protocol provides the order book and clearing infrastructure. Nvidia and SK Hynix perpetual markets already exist; Yan expects more firms to deploy markets and cited Kraken’s announced deployment. He said established firms such as VanEck becoming deployers would be important, not surprising.

For perpetual markets, deployers or publishers provide oracle prices that track an off-chain index, and traders supply collateral-backed liquidity through the order book; the protocol does not take the other side. The builder is responsible for making its oracle represent the stated reference, and the protocol cannot make subjective judgments about an oracle’s suitability. It limits relative oracle-price changes to 1% between updates, clamping larger moves so they can be submitted in smaller steps. Yan says perpetual prices can still converge toward an underlying that is not available around the clock: trading and funding based on the discrepancy can pull prices together, although less frequent funding may slow convergence. If spot markets move onto Hyperliquid, oracle prices could eventually draw on on-chain data instead.

HyperCore executes clearing autonomously as part of blockchain state. Liquidations are triggered by system rules rather than a person setting a subjective price; Yan compares the process to the default waterfall at a US regulated futures exchange, implemented on-chain with different terminology. He argues that high-volume S&P 500 derivatives can contribute to price discovery by allowing market makers to trade against dislocations, potentially helping a thinly traded underlying reflect value. The interviewer raised the risk that derivatives could outweigh and manipulate an underlying price through an oracle, citing Mango Markets, but the discussion did not establish that this happened there. Yan described the Hynex quote incident as an implementation issue: the system behaved as designed but unintuitively, and Trade XYZ published a post-mortem and issued refunds.

Yan traces the project’s 2023 vision to building an on-chain product people actually needed: centralized perpetual venues had demonstrated demand, and combining a similar experience with self-custody could be a step-function improvement. He says self-custody matters because exchange counterparty risk exists in both crypto and non-crypto markets. Permissionless, around-the-clock access could benefit more people, but broader access is not automatically a right or an improvement for everyone. A perpetual tracking private-company shares, for example, would not confer dividends or future cash flows, though futures also lack that claim; education cannot compensate for opportunities being entirely gated off. Yan says deployers are a natural point for regulation because they exercise controls, while the protocol is more like TCP/IP for building financial systems. Derivatives trading through Hyperliquid interfaces is against the terms of service for US users, including Americans using a VPN. Yan believes the current administration wants to explore incorporating on-chain infrastructure into the existing system, but said regulators must weigh the needs of a financial system serving many people against technology they do not yet fully understand.