Hyperliquid's official base-perp API reported $2.7845 billion in 24-hour notional volume on August 11, 2026. Its official audit page publishes two 2023 Zellic reports for the bridge. Audit coverage does not guarantee protocol safety. Hyperliquid runs its own L1 and uses HLP as a market-making and liquidation-backstop vault.
Key takeaways
- The official audit page publishes two 2023 Zellic reports for the bridge. These reports do not cover every protocol or economic risk.
- The March 2025 JELLY squeeze put the HLP vault $13.5 million underwater before validators force-settled the market. HLP ended up about $703,000 in profit, but the episode exposed real gaps.
- Hyperliquid responded with tighter open-interest caps, patched auto-deleveraging logic, and fully on-chain validator votes for delistings.
- Trading costs are competitive but not zero: 0.015% maker and 0.045% taker at base tier.
- Wallet control reduces centralized custody exposure. Validator, bridge, protocol, and interface risks remain.
Use the live tools
24h perp volume — all tracked venues (2026-07-23)
$151.58B
CEX $134.74B · DEX $16.85B
PerpFinder live dataset · verified 2026-07-23
The architecture: a chain built around one exchange
Hyperliquid does not run on Ethereum, Arbitrum, or Solana. It is its own L1, purpose-built so that a full order book can live on-chain: order placement, matching, funding, and liquidations all execute as chain state with sub-second finality. The public API makes fills and liquidations available for inspection. This data helps users review execution and market events.
Hyperliquid does not inherit security from Ethereum or another settlement chain. Its validator set and node software secure the L1. Deposits use a bridge from Arbitrum. The bridge needs signatures from more than two-thirds of validator staking power. Withdrawals also have a dispute period.
Incident history: what has actually gone wrong
The useful question is not whether incidents happened. They did. The question is what broke, who paid, and what changed.
March 12, 2025: the ETH whale unwind
A trader ran an ETH long of roughly $300 million at high leverage, withdrew margin while the position showed a profit, and let the position cross its own liquidation price on purpose. Exiting through the liquidation engine avoided the slippage a market order of that size would have paid. HLP, which backstops liquidations, absorbed a loss of about $4 million. No code was exploited; the margin rules priced large exits badly. Hyperliquid tightened margin requirements for leveraged withdrawals and cut headline leverage on the biggest pairs in direct response (BTC to 40x and ETH to 25x at the time; caps have since been retuned per asset).
March 26, 2025: the JELLY squeeze
Two weeks later came the sharper test. One actor opened three fresh accounts: two holding long positions in JELLY, an illiquid memecoin perp, and one holding a deliberately over-leveraged short of roughly $4 million notional. The short was designed to fail. When it was liquidated, the backstop logic handed the position to HLP. The attacker then bought JELLY aggressively on thin external spot markets, driving the oracle price up and pushing HLP's unrealized loss to about $13.5 million, with the vault facing worse if the pump continued.
Validators voted, within minutes, to delist the JELLY market and force-settle every position at $0.0095, the attacker's entry price, rather than the manipulated oracle price near $0.50. The settlement flipped HLP's position to a profit of about $703,000 and stranded the attacker's longs. HYPE still fell around 20% that day, and the criticism wrote itself: a supposedly decentralized exchange overrode its own oracle by committee. In an extreme event, your PnL on Hyperliquid can be decided by governance rather than the open market. That cuts both ways. It saved depositors here, and it is a form of intervention risk you should know exists.
What changed afterward
The team shipped three concrete fixes. Open-interest caps now account for an asset's real market cap and liquidity, so an oversized position on a thin token like JELLY should be rejected at entry. The auto-deleveraging trigger was reworked to assess the liquidator's account on its own rather than the pooled balance, closing the bypass the attacker used, and the liquidator strategy is now capped to a small share of HLP. Delistings moved to fully on-chain validator voting instead of off-chain coordination. None of this makes manipulation impossible. It does mean the specific 2025 playbooks no longer work as written.
HLP vault vs. a classic insurance fund
Most CEXes keep a passive insurance fund: a pot of money, topped up by liquidation fees, that only pays out when a position closes worse than its bankruptcy price. Hyperliquid runs a different model. HLP (Hyperliquidity Provider) is an active market-making vault, open to public deposits, that quotes markets, earns fees, and also inherits underwater liquidations as the backstop of last resort.
The practical differences: HLP stands in the line of fire first, its full positions and PnL are visible on-chain in real time, and depositors share both the trading returns and the tail risk. A CEX insurance fund discloses balances on the exchange's own schedule and its size is a policy choice. HLP has remained profitable across its history, including through both March 2025 events. The model is not automatically worse than an insurance fund, but you should understand that the backstop is a live trading book, not idle cash.
Audits and transparency
The official audit page publishes two 2023 Zellic reports for the bridge contract. The reports cover defined bridge components and do not guarantee future safety. The Hyperliquid dashboard shows volume, open interest, funding, and liquidation data from the official API.
The honest caveat is that audits cover code. Both 2025 incidents were economic attacks on market structure, a category no audit certifies against. Judge the protocol on how it handled them: user funds were not lost in either event, and the fixes shipped fast.
What it costs to trade there
Fees are part of any safety math, since hidden costs push traders toward riskier sizing. Hyperliquid charges 0.015% maker and 0.045% taker at base tier, with no gas on the L1. Volume tiers lower those rates, and signing up through a referral code takes 4% off. That puts it near the cheap end for order book DEXes, though venues like Lighter run zero-fee models. Model your own numbers with the fee calculator.
Who should size down
Size each position for validator, bridge, market, and liquidation risk. Thin markets can have greater manipulation risk. Check current location limits and terms before you use the interface. Keep only the trading margin that your risk plan needs on the venue.
Has Hyperliquid ever been hacked?+
No. Since mainnet launch in early 2023 there has been no smart-contract exploit of the L1 or core contracts. The notable incidents, the March 2025 ETH whale unwind and the JELLY squeeze, were market-structure attacks. User funds were not lost in either, and both led to risk-engine changes.
What exactly happened with JELLY?+
An attacker used an over-leveraged short on an illiquid memecoin to force Hyperliquid's HLP vault to inherit the position, then pumped the spot price to put the vault $13.5 million underwater. Validators voted to delist the market and settle all positions at the attacker's entry price, which left HLP up about $703,000 and defeated the attack, at the cost of overriding the oracle.
Is money on Hyperliquid safer than on a centralized exchange?+
The risks are different. Wallet control reduces centralized custody exposure. Hyperliquid adds validator, protocol, and Arbitrum bridge risk. A centralized exchange adds operator custody and solvency risk.
Is Hyperliquid audited?+
Zellic audited the bridge contract, and both 2023 reports are public. Audits do not cover economic attacks, which is why the 2025 incident history and the fixes that followed are the more useful safety signal.
Contains affiliate links — we may earn a commission. This does not affect rankings.
