Hyperliquid

Hyperliquid’s validators delisted a live market and settled every open position at a price they selected. It took about two minutes. The reason was that the protocol’s own HLP vault — which makes markets using pooled user capital — had gone $13.5m underwater on that market.

Strip the vocabulary away and the sequence is: the house took a position, the house lost, and the house changed the result. Traders who were correctly positioned did not receive what the market said they were owed, because the venue decided the market would end at a different number.

The defence is real and worth stating. An uncontained loss in the vault would have been absorbed by the same pooled user capital, so the validators faced a choice between two groups of users rather than between users and themselves. Acting fast limited the damage. Every centralised venue has an emergency power of some kind, and most would have used it.

It is still the thing to understand before depositing. The market is not the final word here; the validator set is, and it has demonstrated both the capability and the willingness to use it within two minutes. Decentralised in this context describes where the code runs, not who decides what your position was worth.

The structural issue underneath is the HLP vault itself. A protocol that makes markets with user capital is a counterparty to its own users, and the March 2025 event is what that conflict looks like when it resolves.

None of the above is a claim that it does not work. Hyperliquid generated $63.3m of protocol fees over thirty days — more than every other perpetuals venue in our rating put together — and it runs the deepest on-chain order books that exist, on a layer 1 it built for that single purpose.

The design choice behind that is the interesting one. Most on-chain derivatives venues are automated market makers with a trading interface on top, which produces predictable execution and poor price discovery. Hyperliquid built an actual order book and a chain fast enough to run it, and the result behaves like a professional venue rather than an approximation of one.

Ecosystem value stands at $6.84bn. On execution quality, cost and depth there is no serious competition in its category.

So the assessment is unusually clean: it is the best venue in its category by every measurement we can take, and the governance event above is the risk you accept to use it. Both halves are true, and anyone weighing it should hold them at the same time rather than choosing one.