How Hyperliquid works

The whole machine, explained by following a single order through it.

Updated Fact-checked November 4, 2026 9 sources
Short answer

When you place an order on Hyperliquid, it becomes a signed transaction that validators order with HyperBFT consensus, then HyperCore matches it on an onchain order book, checks your margin against an oracle-based mark price, and settles it in the same block. The docs give a median end-to-end latency of 0.2 seconds; funding is then paid hourly and margin is rechecked as prices update about every 3 seconds.

The quickest way to understand Hyperliquid is to follow one trade. Say you want to go long 1 BTC with 10x leverage. Here is everything that happens, in order.

Step 0: your money arrives

Hyperliquid accounts are wallet addresses. You fund one by sending USDC. Circle now mints USDC natively on Hyperliquid and supports transfers from other chains through its CCTP bridge; the older Arbitrum bridge holds less than 10% of HyperCore USDC, per the docs. Once credited, your USDC sits in your account in HyperCore’s clearinghouse, the part of the chain that tracks every address’s balances and positions. Think of it as the exchange’s ledger, except it is public and nobody can move your money without your signature. See how to deposit USDC.

Step 1: you sign an order

When you click “Buy”, your wallet signs a message: buy 1 BTC perp, limit price X, 10x, cross margin. The app sends it to the network. Nothing is “pending at the exchange”: the signed order is a transaction like any other.

Step 2: validators put it in a block

Hyperliquid’s validators (27 active on 4 November 2026) run HyperBFT, a consensus protocol in the HotStuff family. One validator proposes a batch of transactions; once validators holding more than two-thirds of staked HYPE sign it, the batch is committed. There is no waiting for “6 confirmations”: one committed block is final.

Hyperliquid’s mempool is aware of trading. Within each block, actions are sorted into three groups and run in this order:

  1. actions that don’t place orders (transfers, leverage changes and so on),
  2. cancels,
  3. actions that place new limit or immediate-or-cancel orders.

So if a market maker cancels a quote in the same block that you try to hit it, the cancel wins. That rule protects people who post liquidity, and in turn keeps spreads tight for everyone.

Step 3: HyperCore matches it

HyperCore is the exchange engine built into the chain. It keeps one order book per market and matches orders by price, then time: the best price fills first, and at the same price, whoever was there first. Prices must be a multiple of the tick size and sizes a multiple of the lot size, exactly like a stock exchange.

Before your order can open a position, the clearinghouse checks margin: at 10x you need 1/10 of the position’s value (at the current mark price) available. It checks again for the resting order at the moment it fills, because prices may have moved since it was placed.

Step 4: the price that decides your fate

Your position’s profit, loss and liquidation are judged against the mark price, not the last trade. That’s deliberate: a single strange trade shouldn’t be able to liquidate you.

PriceBuilt fromUsed for
Oracle priceEach validator computes a weighted median of spot prices on Binance (weight 3), OKX (2), Bybit (2), Kraken, KuCoin, Gate, MEXC and Hyperliquid spot (1 each); the chain takes the stake-weighted median of validators’ submissionsFunding
Mark priceMedian of: oracle + a 150-second average of Hyperliquid’s premium; Hyperliquid’s best bid/ask/last trade; other exchanges’ perp pricesMargin, liquidations, TP/SL triggers, unrealized PnL

Both update about every 3 seconds. For assets whose main market is Hyperliquid itself (HYPE, PURR) the oracle leans on Hyperliquid’s own spot book.

Step 5: every hour, funding

A perpetual future never expires, so something has to keep its price tied to the real asset. That’s funding: every hour, longs pay shorts (or the reverse) based on how far the perp trades above or below the oracle. When markets are calm the rate sits at a baseline of 0.00125% per hour paid by longs; that’s what BTC, ETH and HYPE showed on 4 November 2026. Funding is peer to peer. Hyperliquid takes no cut.

Step 6: if it goes wrong, liquidation

If the price falls and your account value drops below the maintenance margin (half of the initial margin at the market’s max leverage, so 1.25% for a 40x market), the protocol starts liquidating. It first sends market orders to the book, so you keep whatever margin is left. Only if your equity falls below two-thirds of maintenance margin does a backstop take over the position: the liquidator vault inside HLP. As a last resort, auto-deleveraging closes profitable opposite positions so the exchange never carries bad debt.

Step 7: you close, fees are split

When you close, you pay a fee (0.045% taker or 0.015% maker at the entry tier, less 4% with a referral code and less again if you stake HYPE). Hyperliquid’s docs say fees go to the community side of the system: HLP, deployers of the markets, and the Assistance Fund, which converts fees into HYPE automatically. If you traded through a wallet or app, it may also charge a builder fee.

Where HyperEVM fits

Everything above happens in HyperCore. The chain also runs a second environment, the HyperEVM, where ordinary Ethereum-style smart contracts live: lending markets, liquid staking, DEXs. Both run under the same consensus, and contracts can read HyperCore prices and send it orders through precompiles and CoreWriter.

Why it matters to you

  • Your liquidation price is about the mark, not the chart. Wicks on one exchange rarely liquidate you; a real move across exchanges does.
  • Fills are final and public. No exchange can roll back a trade, and anyone can see your positions by address.
  • Speed is close to a centralized venue. For most people a 0.2-second median feels instant; for high-frequency traders, the block ordering rules matter more.

What can go wrong

  • Oracle and mark depend on outside venues. If major exchanges glitch, Hyperliquid’s prices are built to resist it, but extreme events can still move the mark fast.
  • Leverage plus fast blocks means fast liquidations. There is no margin call phone call. Use stop losses and keep a buffer.
  • Network upgrades pause matching briefly. The docs mention a post-only period during upgrades, when market orders and TWAP suborders don’t fill.

Questions people ask

Where does Hyperliquid get its prices?

Two places. The oracle price, published by validators about every 3 seconds, is a weighted median of spot prices on Binance, OKX, Bybit, Kraken, KuCoin, Gate, MEXC and Hyperliquid itself. The mark price, used for margin and liquidations, is a median that also includes Hyperliquid's own order book and other exchanges' perp prices.

Is the Hyperliquid order book really onchain?

Yes. Every order, cancel, trade and liquidation is a transaction in a HyperBFT block. There is no off-chain matching engine. That is why anyone can replay the full history of the book.

How fast is a Hyperliquid trade?

Hyperliquid's docs state a median end-to-end latency of 0.2 seconds and a 99th percentile of 0.9 seconds for a client located near the validators, with capacity of about 200,000 orders per second.

Why do cancels go before new orders?

Within each block Hyperliquid sorts actions so that non-order actions run first, then cancels, then new orders. That lets market makers pull stale quotes before anyone can trade against them, which tends to keep spreads tighter.

Sources

  1. Hyperliquid Docs — HyperCore overview
  2. Hyperliquid Docs — Order book (HyperCore)
  3. Hyperliquid Docs — Oracle
  4. Hyperliquid Docs — Robust price indices (mark price)
  5. Hyperliquid Docs — Clearinghouse
  6. Hyperliquid Docs — Funding
  7. Hyperliquid Docs — Liquidations
  8. Hyperliquid Docs — Staking (rounds, epochs, quorum)
  9. Hyperliquid Docs — USDC (native USDC, CCTP)