cryptonist
A printed circuit board photographed flat, its traces and mounted components in rows

Explainers

What Hyperliquid is, and what its on-chain data shows

Hyperliquid is a layer 1 blockchain whose main application, an exchange for perpetual futures and spot tokens, is part of the chain itself, not a contract deployed on it. Every order, cancel and liquidation is a transaction that validators order and execute. That one design choice explains most of what is unusual about it: where its fees go, what its token is for, and why, in March 2025, its validators voted to close a market.

A chain built around one exchange

Hyperliquid’s documentation describes the chain’s state in two parts. HyperCore holds the perpetual and spot order books, the margin system and the clearing of positions. The HyperEVM is a general-purpose smart contract environment of the kind Ethereum introduced; it is not a separate chain but runs under the same consensus, so a contract deployed there can read prices from the HyperCore books and send orders to them through a system contract called CoreWriter. Orders sent that way are held back on-chain for a few seconds, so that a contract cannot use the HyperEVM to get ahead of orders sent to HyperCore directly.

The order in which those parts arrived is worth knowing. An April 2023 post from the team described a chain custom-built to run a derivatives exchange, with no plans then for general-purpose smart contracts, operating in a mainnet closed alpha with no gas or trading fees and a consensus layer based on a tuned version of Tendermint. The Hyper Foundation’s review of 2024 lists HyperBFT consensus and a testnet EVM among that year’s releases. When read on 22 November 2026, the HyperEVM overview page still described the environment as being in an alpha stage.

Two organisations appear in the documentation. Hyperliquid Labs is described as a core contributor that is self-funded and has taken no external capital. The Hyper Foundation published the plan for the token’s genesis and runs a programme that delegates HYPE from its own budget to validators it selects.

What an on-chain order book means

Decentralised exchanges mostly set prices in one of two ways. An automated market maker holds a pool of two tokens and prices each trade from a formula over the pool’s reserves; Uniswap’s documentation gives the constant-product rule, under which a trade that is large against the pool moves its own price. An order book has no formula. Traders leave limit orders at prices they choose, and an incoming order fills against the best of them.

Hyperliquid’s order books work as a centralised exchange’s do: prices sit on a tick grid, sizes on a lot grid, and orders match by price and then by time. What differs is where the book is kept. The team’s 2023 case for building a chain at all, in a post on why an on-chain order book matters, was that whoever runs an off-chain matching engine can refuse cancels or favour some users in ways nobody else can verify. On HyperCore the book is chain state, so its rules are the chain’s rules and any node can replay them.

One of those rules shows what that makes possible. Within a block, the documentation says, the actions in each consensus batch (usually one or two per block) are sorted into three groups: first those that send no good-til-cancelled or immediate-or-cancel orders, then cancels, then the actions that do send such orders. A January 2024 post gives the purpose: with cancels and post-only orders processed ahead of the rest, a market maker’s cancel does not lose a race against a faster taker. A centralised venue could adopt the same rule privately; here it is part of how every node executes a block.

The product those books were first built for is the perpetual future, a contract with no expiry that is held near its underlying by payments between longs and shorts. Hyperliquid’s funding is paid every hour, peer to peer, from a rate built on the gap between the contract’s price and an oracle price. Validators publish that oracle every three seconds as a weighted median of spot prices from seven centralised exchanges and Hyperliquid’s own spot market (assets whose main spot market is elsewhere, such as BTC, leave out Hyperliquid’s price, and assets that trade mainly on Hyperliquid, such as HYPE, can leave out the outside venues), and the clearing house uses the stake-weighted median of what the validators submit.

When an account’s equity falls below its maintenance margin, its positions are sent to the book as market orders. If equity falls below two-thirds of maintenance margin without the book absorbing them, a liquidator vault that is part of HLP, the protocol’s community-owned market-making vault, takes the position over. Behind that sits auto-deleveraging: if an account’s value goes negative, profitable positions on the other side are closed against it, so that the platform carries no bad debt.

How HyperBFT reaches agreement

HyperBFT is Hyperliquid’s own consensus algorithm, which the documentation describes as a variant of HotStuff, a Byzantine fault-tolerant consensus protocol. Blocks are produced by validators in proportion to the stake delegated to them. Consensus runs in rounds, and the staking page defines a quorum as any set of validators with more than two-thirds of total stake; the chain is safe for as long as a quorum is honest. The validator set and its stakes are fixed for epochs of 100,000 rounds, about 90 minutes on mainnet.

Validators can vote to jail a peer that answers consensus messages too slowly or too rarely. A jailed validator stops taking part and earns nothing for its delegators until it unjails itself. Jailing is not slashing: the documentation reserves slashing for provably malicious behaviour such as signing two blocks in the same round, and says no automatic slashing is implemented.

According to the Hyper Foundation, the validator set became permissionless on 21 April 2025, with the active set then made up of the twenty-one validators with the most stake. Validators do more than order transactions. They publish the oracle prices that funding and liquidations use, and they vote on whether to delist the perpetual markets they operate, which then settle at a one-hour time-weighted oracle price. Since HIP-3, a deployer that stakes 500,000 HYPE can also run a perpetual exchange of its own on HyperCore, defining its markets, setting their oracle prices and settling them, with that stake exposed to slashing over the markets it lists.

What HYPE does, and where fees go

HYPE is staked in a delegated proof-of-stake system. A validator needs to delegate 10,000 HYPE to itself, locked for a year, before it can become active; a delegation is locked for a day; and HYPE leaving the staking account passes through a seven-day queue. The reward rate falls as more HYPE is staked, being inversely proportional to the square root of the total, and rewards come out of a future emissions reserve, not out of fees. The genesis post set the maximum supply at one billion HYPE, with 31% allocated to the genesis distribution of 29 November 2024 and 38.888% reserved for future emissions and community rewards. On the HyperEVM, HYPE is also the gas token.

Fees work differently in the two halves of the chain. On HyperCore, the fee schedule charges on volume; for perpetuals in the lowest volume tier, it was 0.045% for orders that take liquidity and 0.015% for orders that add it when read on 22 November 2026, with discounts for staking HYPE. The same page says trading fees go to HLP, to an assistance fund and to deployers of spot and HIP-3 markets. The assistance fund converts its fees into HYPE automatically, and the HYPE it holds is burned. On the HyperEVM, EIP-1559 is enabled and base fees are burned as on Ethereum; so are priority fees, which on most EVM chains go to the block producer. The first transaction sent to a new HyperCore account also carries a one-off activation fee of one unit of a quote token, such as 1 USDC. How fees like these appear in protocol data, and why fees and revenue are not the same number, is set out in protocol fees and revenue.

Documented incidents

One episode concerned a single market. On 26 March 2025 Hyperliquid said that, after evidence of suspicious market activity, the validator set had convened and voted to delist the JELLY perpetual market. All users apart from flagged addresses would be made whole by the Hyper Foundation, automatically and from on-chain data, and making the voting system sturdier and more transparent was a priority. The statement did not describe the trades involved, and this article does not either: the detailed accounts in circulation come from the press and from analysts, not from Hyperliquid. What the statement documents is the decision itself, taken by a group of validators who also produce the chain’s blocks and publish its prices.

In July 2025 the problem was in front of the chain. Hyperliquid’s status page records that between 14:10 and 14:47 UTC on 29 July 2025, orders sent through its API servers reached the nodes late because of a spike in traffic, and that there was no hack or exploit. Because the API servers sit outside consensus, the incident shows a dependence of another kind: on the infrastructure most users send orders through, where consensus is not involved at all.

Earlier halts were on the chain itself. The same page’s history lists network outages in January, February and March 2024, one attributed to a thirtyfold spike in load and another to traffic heavy enough that no transactions went through for three minutes.

The site has no hack page for Hyperliquid L1, because DefiLlama’s record does not yet hold the twenty incidents a chain page needs; how crypto gets hacked reads that record as a whole.

What the on-chain data shows

As of 22 November 2026, per DefiLlama, Hyperliquid L1 had $2.2B of total value locked, 8.7% lower than thirty days earlier, with 206 protocols listed and $5.9B of stablecoin supply on the chain (data page). The total is the value of assets deposited in DeFi protocols on the chain at the snapshot’s prices, leaving out liquid staking, bridges and deposits DefiLlama flags as double-counted. Price moves change it with nobody depositing or withdrawing, so a thirty-day change mixes flows and prices in one number (what TVL measures).

The list of largest protocols is on a different basis, and here the difference is large. Its first entry, Hyperliquid Bridge, stands at $5.6B on the chain, more than twice the chain’s own figure. DefiLlama files it under Bridge, a category the chain total leaves out, and describes native USDC as minted on Hyperliquid through Circle’s transfer protocol, with USDC also bridged from Arbitrum. Hyperliquid’s USDC page says the legacy Arbitrum bridge holds less than a tenth of the USDC on HyperCore.

After it come Hyperliquid Earn at $1.4B, which DefiLlama files under Lending and describes as collateralised lending native to HyperCore; Kinetiq kHYPE at $592.1M, a liquid-staking token for staked HYPE and therefore outside the chain total; HyperLend Pooled at $231.1M, a lending market on the HyperEVM; and Hyperliquid HLP at $148.7M, filed under Derivatives. The exchange’s spot book has an entry of its own, Hyperliquid Spot Orderbook, at $98.8M under Dexs.

Stablecoin supply on the chain is about 2.7 times its total value locked (computed). The two figures come from different DefiLlama feeds and measure different things, supply circulating on a chain against deposits in listed protocols, and the snapshot does not say how much of the supply is held as trading margin on HyperCore and how much elsewhere. Where stablecoin supply lives explains why its own chain table and the chain pages do not agree.

What to check

Three questions follow from all this.

  1. Who operates the market? A validator-operated perpetual is delisted by validator vote and settles at a time-weighted oracle price; a HIP-3 market is defined, priced and settled by its deployer.
  2. How concentrated is the active set? Validators publish the prices liquidations use and vote on delistings, so their number and independence bear on every position, and the Foundation’s delegation programme adds its own stake to the validators it chooses.
  3. Which figure is being quoted? The chain total, the bridge balance and the stablecoin supply are three different measures, and the chain total, the headline figure on the data page, is the smallest of them.