Hyperliquid Gossip Priority: Wintermute Won 76% of the Top Data Slot Auctions
A Wintermute-linked wallet won 76% of Hyperliquid's top gossip priority auctions. How read priority works, who pays for it, and what it means for fills.
$ Stop reading delayed data. Read live order book depth on the 100 most liquid Hyperliquid pairs right now.
Launch Free Terminal →A wallet linked to Wintermute won about 76% of Hyperliquid's top-priority data slot auctions over the past month, according to Loris Tools analysis cited by Bloomberg on November 12. The slot buys an early look at incoming buy and sell orders before the exchange processes them, giving a trading algorithm milliseconds to adjust quotes or hedge elsewhere.
The mechanism is Hyperliquid gossip priority, also called read priority. It has been live since April, and it has drawn criticism that it hands larger firms an edge, The Block reported. For anyone trading Hyperliquid from a screen, the useful question is narrower: what exactly is being sold, how concentrated the buyers are, and what that changes about your own fills.
How Hyperliquid gossip priority auctions work
The official priority fees documentation describes three independent Dutch auctions synced to the same three-minute schedule. Each auction index is a priority slot, and lower indices are strictly served first. Nodes that respect the ordering send mempool transaction data to the winning IPs ahead of other peers, and each auction's result applies during the following auction period.
Several details limit who can use it. A bid is tied to an IP address, and the onchain IP must exactly match the IP the sending peer sees. Slots are not additive, so an IP holding several slots is ranked by its lowest one. The Foundation's non-validator node opts into the ordering, but each hop in a chain of nodes may or may not respect it.
Pricing is aggressive by design. Each auction resets to ten times the previous winning price for that slot and decays toward a floor of 0.1 HYPE, paid from spot balance and burned. The docs put the current empirical effect at roughly 20 milliseconds of latency reduction per auction slot.
The parameters have moved since launch. Dwellir's April explainer described five auctions running on the three-minute cycle. The docs now describe three, so any analysis of slot economics should be checked against the current documentation.
Read priority vs order priority on Hyperliquid
Gossip priority only affects reading data. Sending data is a separate product called order priority, and the same documentation sets it out in detail. For IOC orders, a priority rate between 0 and 8 basis points of filled notional has a linear effect on latency, worth about 45 milliseconds per basis point on mainnet. Fees come from undelegated staking balance and are burned.
The key protection is unchanged: all cancels are processed before any immediately executable order, priority or not. A maker who sees adverse flow coming can still pull a quote ahead of a paid taker. ALO orders have their own priority, which reorders the tail of a price level's queue over a 400 millisecond window, a mechanism covered in an earlier Buildix article.
Put together, the system sells three kinds of speed: seeing pending flow first, getting aggressive orders in first, and holding a better place in the passive queue. The Wintermute story concerns the first.
How concentrated the Hyperliquid data auction is
Bidding is open, but participation is thin. Loris Tools data cited by TronWeekly shows about $7.2 million in cumulative read-priority fees from 40 unique payers as of November 10, against more than 1,050 unique payers of order-priority fees. The same dashboard showed the three largest read-priority payers accounting for almost all recorded fees.
A Quicknode guide citing Hyperscreener data gives a 30-day snapshot: $2.90 million in total priority fees, of which $1.08 million was gossip priority across about 28,000 reads, an average near $38.67 per read. It also notes that one address accounted for more than half of all-time read-priority spend.
Winning the auction is not the whole cost. The founder of Loris Tools said the full speed advantage requires direct access to nodes operated by the Hyperliquid Foundation, and that this access depends on trading volume and staking requirements, per a ChainCatcher summary of the Bloomberg report.
The fairness argument on both sides
DRW founder Don Wilson had already questioned whether the mechanism is fair and whether it could meet US regulatory requirements, PANews reported. That matters because Hyperliquid is working toward a US route: Payward's proposed Bitnomial HIP-3 venue was still awaiting approval at the end of Q3, according to the Hyperliquid Research Collective.
Wintermute's Evgeny Gaevoy argued that the auctions are more transparent than the low-latency networks and colocation deals of traditional finance, and said the firm mainly uses the service for crypto trading, according to ChainCatcher. Hyperliquid Labs said anyone can bid and that every priority fee payment is recorded onchain.
Both points can be true at once. Traditional venues sell speed through rack space and cross-connects whose buyers are mostly private. Hyperliquid sells it through an auction anyone can read. Transparency shows you who holds the edge. It does not spread the edge around.
What a 20 millisecond head start means for your fills
A slot winner sees pending orders before most of the network. If a large market buy is about to hit the book, the winner can cancel or reprice asks before it lands, and cancels always execute first. The taker then fills against a book that has already adjusted. That is the practical cost to everyone else: worse marginal fills on aggressive orders into thin liquidity, not stolen trades.
The response for a non-HFT trader is mostly about order type. Passive limit orders are not exposed in the same way, while large IOC orders swept through several levels are where a faster reader can step in front. Splitting size, working limits inside the spread, and measuring realized slippage against the mid at submission will show whether the effect reaches your trading at all.
None of this touches signals measured over seconds or minutes. Cumulative volume delta, order book imbalance and VPIN describe where aggressive flow is going over many blocks, which is a different game from winning a 20 millisecond race. On Buildix, the pair deep view shows those three metrics together, and entry signals are built on the same orderflow inputs.
The auction makes Hyperliquid's speed market visible, and the first public read on it shows one firm holding most of the top slot. Watch whether the payer count rises as more desks route through Foundation nodes, or whether governance changes the slot design before a US venue goes live.