Hyperliquid

Decentralised derivatives spent five years being a worse version of the thing they were replacing. Order books were kept off-chain because on-chain matching was too slow, liquidity was rented with unsustainable emissions, and the user experience asked traders to accept latency, slippage and wallet friction in exchange for a principle. Hyperliquid is the first project our desk has reviewed that resolves that trade-off rather than apologising for it.

The approach was uncompromising: rather than deploying an exchange onto a general-purpose chain and inheriting its constraints, the team built a layer 1 whose entire reason to exist is running a central limit order book. Everything in the architecture — block cadence, state model, execution pipeline, fee design — is subordinated to that single application. The result is a venue where orders place, fill and cancel with the responsiveness traders expect from a centralised platform, while every order, fill and liquidation is publicly verifiable.

HyperBFT consensus produces blocks fast enough that order placement feels immediate rather than queued, and the order book lives in consensus state rather than in an operator's database. This is the distinction that matters. On most so-called decentralised exchanges, matching happens on a server the user cannot inspect and only settlement is on-chain. Here the matching engine is the chain, which means front-running by the operator, phantom liquidity and undisclosed internalisation are structurally impossible rather than contractually discouraged.

HyperEVM extends the platform into general smart contracts while giving those contracts native access to the exchange's state, so a lending protocol can read live oracle marks and margin balances directly rather than through a bridge. That composability is the genuine platform play: it turns a single exceptional application into an ecosystem substrate without diluting the specialisation that made the application good.

Liquidation and margin machinery has been stress-tested by real market events rather than by simulation. The vault system that backstops liquidations is publicly auditable, its performance is visible to anyone, and losses have been socialised transparently when they occurred. We deduct half a point at this line only because validator-set breadth and client diversity remain thinner than the architecture ultimately warrants, and a fast BFT chain concentrates more trust in its operator set than a slower, more redundant design.

Hyperliquid's distribution is the most user-favourable of any major venue we have assessed. There was no venture round, no private allocation sold at a discount to funds, and no exchange-listing tranche. The genesis distribution went overwhelmingly to people who had actually used the product, and subsequent emissions have continued to favour users and market participants over insiders.

Fee flow is equally clean. Trading fees are directed to buybacks and to the assistance fund rather than to a corporate entity, which means the token accrues value from the same activity it secures. Our supply-forensics workflow looks for the moment when an insider tranche meets a thin bid; on Hyperliquid there is no such tranche to model.

This is what a functioning revenue-linked token looks like: real fees, generated by real volume, from users who chose the venue on merit rather than on incentive farming. Very few tokens in this asset class can make that claim without a footnote.

Here is the honest deduction. Hyperliquid's performance comes from a tightly engineered, relatively concentrated validator set and a small core development team that ships fast. Those are the same properties that make it quick, and they are also the properties that make it more centralised than a general-purpose base layer. Validator count, geographic distribution and independent client implementations all have room to grow, and governance today is more informal than a mature protocol should be.

The team is candid about the trade-off rather than pretending it away, and the direction of travel has been toward broader participation. But readers should size their exposure with the current reality in mind: this is a high-performance venue with a decentralisation roadmap, not a maximally decentralised settlement layer.

The interesting question for the next twelve months is whether HyperEVM turns Hyperliquid from an exchange into a financial operating system. Early signs are encouraging: lending markets, structured-product vaults, liquid staking and automated strategy managers have deployed and are using native access to exchange state rather than approximating it through oracles. That composability is worth more than raw throughput, because it lets third parties build products that would require a licence and a prime broker anywhere else.

Spot markets, a permissionless listing mechanism and builder-code revenue sharing extend the same logic outward: front-ends can route order flow and earn a share of fees, which turns competitors into distribution. It is the single smartest business decision we have seen a decentralised venue make.

The risks to monitor are concentration of open interest in a small number of markets, the resilience of the backstop vault during a genuine gap-down event, and regulatory attention to on-chain leverage aimed at retail users. None of those are hypothetical in this asset class, and none of them are unique to this venue.

Hyperliquid took the single hardest product in crypto — a competitive derivatives exchange — and built it on-chain without asking users to accept a worse experience for the privilege. It funded itself with fees rather than with a private round, and it gave the upside to the people using it.

Four and a half out of five. The half point comes back when the validator set and client diversity catch up with the engineering.