Hyperliquid is the largest futures exchange on a blockchain. On 15 October 2026 it held open positions worth $11.0 billion, and $8.7 billion traded in 24 hours, a tenth of Binance's futures volume. Its token HYPE rose 105 percent in twelve months while Bitcoin fell 28 percent.
The promise behind it is written in its own documentation: "A permissionless set of independent validators secures the network." No company that can freeze accounts or set prices. And the platform has survived stress. During the liquidation cascade of 10 October 2025, Hyperliquid ran, according to co-founder Jeff Yan, with "100% uptime with zero bad debt".
Up to that day, Hyperliquid keeps its promise. What matters is who decides when things get tight.
On Hyperliquid, the exchange is the blockchain
Hyperliquid runs its own blockchain. Its core, HyperCore, is a complete order book: every order, every cancellation, every liquidation is a transaction on the chain. Alongside it runs HyperEVM, where programs run as on Ethereum.
A pool called HLP plays a central role. Users deposit money into it, and the pool quotes prices and takes over positions that are liquidated. It is market maker and backstop in one.
Most of the fees flow into the so-called Assistance Fund, which uses them to buy HYPE on the market. It now holds 48.0 million HYPE, 15.9 percent of the circulating supply.
On Hyperliquid the exchange is the blockchain. Whoever decides about the chain decides about the exchange.
So who decides about the chain?
Two addresses supply two thirds of the stake
Validators confirm new blocks. According to the documentation, every block needs signatures from validators holding more than two thirds of total stake. It follows that anyone with more than a third can halt the chain.
On 15 October 2026 there were 35 validators, 27 of them active. Six were jailed, including the exchange Kraken and the staking provider Figment. The distribution:
| Share of stake | |
|---|---|
| five Hyper Foundation validators | 47.3 % |
| the three largest, all Foundation | 36.0 % |
| needed to halt the chain (over a third) | 3 validators |
| needed to confirm blocks (over two thirds) | 8 validators |
The Foundation alone cannot confirm a block, but it can stop any.
More important is where the stake comes from. HYPE holders can delegate their tokens to validators. Two addresses that received their HYPE at launch in November 2024 together delegate 67.8 percent of all stake. One received exactly 238 million HYPE, precisely the share the tokenomics assign to core contributors, and today delegates 238.0 million across 23 validators, after staking rewards and after 10.6 million were withdrawn since November 2025. The other received 60 million, the Foundation budget share, and delegates 60.5 million across 17 validators. That these are the team and the Foundation follows from the amounts. The addresses carry no label.
19 validators with a combined 78.4 percent of stake get more than half of their stake from these two addresses. The rules of the delegation program say: "The Foundation reserves the right to cease delegation at any time."
Whoever hands out the stake has a say in who stays a validator.
How did these validators decide when it mattered?
When it got tight, decisions were made
Four cases from 2025 are documented:
| Date | What happened | Who decided | Consequence for users |
|---|---|---|---|
| 12 Mar 2025 | A trader withdrew profits and let a large ETH position be liquidated; HLP lost about $4m | Hyperliquid | Max leverage cut: BTC 40x, ETH 25x |
| 26 Mar 2025 | An attacker pumped the small token JELLY; HLP took over his losing position | "the validator set convened and voted to delist JELLY perps" | Market closed; longs compensated by the Foundation at $0.037555, "apart from flagged addresses" |
| 10 Oct 2025 | Liquidation cascade; HLP could no longer absorb the positions alone | automatic (auto-deleveraging) | 34,983 forced closures across 19,337 wallets in about twelve minutes, at the expense of profitable positions |
| 12 Nov 2025 | Manipulation of the POPCAT token; HLP lost about $4.7m | automatic lock, released with emergency keys | Arbitrum bridge paused twice, 16:08–16:37 and 16:52–16:59 UTC |
The JELLY case shows the mechanics most clearly. According to Hyperliquid, the attacker had built a $4 million position at $0.0095 and then pushed the price. When HLP took over the position, the validators voted and closed the market. The Foundation paid the longs the difference. Hyperliquid wrote: "validators often need to convene to take decisive action as a group to ensure the integrity of the network." Three days later a second market, MYRO, was closed. The votes came from Foundation validators 2 to 5; validator 1 abstained. Their stake alone was enough for the decision at the time.
10 October hit the other side. In auto-deleveraging, when losses can no longer be covered, opposing positions are closed, ranked by profit and leverage. Those who were right had to exit before they could let their gains run. The figures come from a paper by Tarun Chitra (arXiv, 2025). Hyperliquid itself gave no figure, only that it was its first cross-margin auto-deleveraging "in more than 2 years of operation". According to chain data, HLP earned about $41 million in the first half of October.
So far the interventions have tended to protect users. What guaranteed that was not a rule but a decision.
And the money that reaches the platform over the bridge?
$395 million depends on four keys
For a long time most money reached Hyperliquid through a bridge from Arbitrum. On 15 October its contract held 394.8 million USDC.
The source code says withdrawals are approved by two thirds of validator power. On the chain it looks different: the contract knows four signers with one vote each, and three must agree. They have been the same four keys since December 2023. Who owns them cannot be determined on the chain.
Hyperliquid itself now calls this bridge deprecated and recommends the route via Circle's transfer service CCTP. According to the documentation, it holds less than 10 percent of the USDC on the platform.
$395 million depends on four keys whose owners cannot be identified on the chain.
What has all this meant for the HYPE price?
The price survived the interventions, but fell harder on the days themselves
HYPE has traded since 29 November 2024. Monthly returns from Hyperliquid spot prices against the BTC perpetual, December 2024 to September 2026, 22 months:
| Dec 2024 to Sep 2026 | |
|---|---|
| Months | 22 |
| Correlation of monthly returns with BTC | 0.29 |
| Months in which BTC fell | 11 |
| … of which HYPE did better than BTC | 7 |
| … of which HYPE itself fell | 6 |
| $1 in December 2024 became | HYPE $10.69, BTC $0.87 |
| largest drawdown | −68 % (21 Dec 2024 to 6 Apr 2025) |
On the stress days HYPE fell harder than Bitcoin: 12.3 percent on the JELLY day with Bitcoin flat, 14.5 percent against 7.3 percent on 10 October 2025. Each time the price recovered afterwards. A large part of the return comes from the first month: December 2024 at +182.5 percent. 22 months are too few to call anything a lasting property.
Investors have so far forgiven the interventions. There is no evidence that this will last.
Claims about Hyperliquid, checked
| Claim | Who, when | What the data says | Verdict |
|---|---|---|---|
| "A permissionless set of independent validators secures the network" | Hyperliquid documentation | Foundation 47.3 % of stake; 67.8 % of stake from two genesis addresses; delegation tied to Foundation conditions | partly holds |
| 22 independent operators hold just over half the stake | crypto.news, 27 Jul 2026 | 19 validators get more than half of their stake from team and Foundation | partly holds |
| The bridge is secured by two thirds of validator power | comment in the bridge source code | contract: four keys, three of four required | does not hold |
| 97 % of fees go to buybacks | VanEck, Hyperliquid, 2025 | revenue per DefiLlama 80.9 % of fees, both since launch and over the last 30 days; the Assistance Fund's purchases match that revenue | partly holds |
| The team will sell its unlocking tokens | The Defiant, November 2025 | Since November 2025 the team address has withdrawn 10.6m HYPE and distributed it to 126 addresses, the largest tranche so far 3.75m on 7 Oct 2026. That tranche went via five intermediate addresses to one address, which re-staked 3.1m and moved 0.6m to HyperEVM; it shows no sale on Hyperliquid's spot market | not evidenced so far |
| No allocation for venture investors | Alea Research and others | no investor tranche at launch; a private round cannot be ruled out | partly holds |
| No bad debt on 10 Oct 2025 | Jeff Yan, 11 Oct 2025 | no contrary data found; the price was 34,983 forced closures | verified |
The pattern: what Hyperliquid says about its stability holds. What it says about its decentralisation only partly holds.
Open to see does not mean spread out
"On Binance one company decides alone and nobody sees how. On Hyperliquid everything is in the open."
True, and it is a real advantage. Every figure in this text about stake, bridge and HLP comes from the chain itself. But openness does not distribute power. A market closed by vote is just as closed for the trader as one closed by an exchange. Visible does not mean distributed.
In an emergency, a few decide about your money
If you trade on Hyperliquid: your gains can be capped by auto-deleveraging, especially when you are right and use high leverage. Markets in small tokens can be closed by vote, and the validators decide the settlement price. So far the Foundation has compensated users, voluntarily.
If you move money onto the platform: the old route via the Arbitrum bridge depends on four keys. Hyperliquid itself recommends the newer route via CCTP.
If you hold HYPE: you hold a token whose chain is run by a group whose stake comes two thirds from the team and the Foundation.
Trading on Hyperliquid is also a bet on how a few people decide when it counts.
Three numbers show whether that changes: the Foundation's share of stake (47.3 percent), the share of stake from the two genesis addresses (67.8 percent) and the number of validators that together can halt the chain (three today, all Foundation). What remains open is whether the Foundation brings its share below a third.
FAQ
What is auto-deleveraging? When an account loses more than its collateral and the loss can no longer be covered, Hyperliquid closes positions on the other side. According to the documentation, those affected are ranked by profit and leverage. Anyone without open positions bears no losses: "a user who has no open positions will not socialize any losses of the platform."
What is HLP? A pool users can deposit into. It quotes prices and takes over liquidated positions. Its gains and losses go to the depositors. In March and April 2025, after the ETH case and JELLY, deposits fell from about $510 million to $150 million.
Can the Foundation freeze my deposit? Blocking individual accounts is not provided for in the documentation. What is documented are interventions at market level: closing markets, setting settlement prices, pausing the bridge. How validators would vote in a case affecting a single account cannot be read from the past.
How are the monthly returns calculated? Daily close at 00:00 UTC, HYPE from Hyperliquid's spot market, BTC from Hyperliquid's BTC perpetual, month-end close against the previous month-end close. October 2026 is still running and is left out.
Not investment advice, not a recommendation, not a forecast — historical patterns are no guarantee.
Sources: Hyperliquid documentation (hyperliquid.gitbook.io: 101, Staking, Delegation Program, Auto-Deleveraging, USDC) · Hyperliquid on X, 12 Mar 2025, 26 Mar 2025, 27 Mar 2025, 29 Mar 2025 · Jeff Yan on X, 5 Apr 2025 and 11 Oct 2025 · Hyperliquid info API (validatorSummaries, delegatorSummary, delegatorHistory, tokenDetails, spotClearinghouseState), retrieved 15 Oct 2026 · Arbitrum, Hyperliquid bridge contract, retrieved 15 Oct 2026 · Tarun Chitra, "Autodeleveraging: Impossibilities and Optimization", arXiv 2512.01112 · DefiLlama, Hyperliquid fees · The Block and Decrypt, 12 Nov 2025 · crypto.news, 27 Jul 2026
Study the Past — Improve your Future 🥋