Hyperliquid Sets Up Toll Booth for Execution Priority

Traditional exchanges have spent years trying to internalize that spending rather than watch it accrue to third parties. The colocation racks, the microwave towers, the fibre runs measured in metres – a genuine wealth transfer to a handful of speed firms that flawed market design forces everyone else to subsidize.

Priority fees on Hyperliquid, a Layer-1 blockchain built for perpetual-futures trading, have run on the mainnet since 13 Apr in an attempt to route some of that money back through the front door. The idea is old, and the venue is new. The fees were split into two auctions solving two different problems. Order priority – "write" fees – buys preferential inclusion for immediate-or-cancel orders on HIP-3 assets, the permissionless markets outside builders can deploy by staking Hyperliquid's HYPE token. Gossip priority – "read" fees – is a Dutch auction for the right to receive raw transaction data before it executes, a roughly 25ms head start on order flow that liquidation bots and market makers turn into edge.

Both fees, paid in HYPE, are burned. That last detail is what makes this more than a fee tweak: every dollar of priority spend permanently removes HYPE from supply, converting the high-frequency trading (HFT) latency arms race directly into a deflationary force on the token.

Solana benchmark

Since priority fees appeared in the dataset on 13 Apr, Hyperliquid has generated $3.16mn in cumulative priority fees. April was basically a warm-up, with just $96k in priority fees from 13 Apr onward. May did $762k. June did $2.09mn, or roughly two-thirds of all cumulative priority fees so far.

Against this headline, the obvious benchmark is Solana, where priority fees are a mature, multi-hundred-million-dollar market. Over the past 30 days through 2 Jul, Solana averaged $289,703 a day in priority fees to Hyperliquid's $73,357; the latter already running at a quarter of Solana's daily rate three months in. That does not mean Hyperliquid is "catching Solana." It means a central limit order book (CLOB) can grow a meaningful latency market rapidly when the users are professional and the edge is direct.

(Source: Coin Metrics, Hydromancer)

The Solana comparison is useful, but only if framed correctly. Solana priority fees are a general-purpose blockspace mechanism. They are priced through compute-unit limits and compute-unit prices, and the priority fee goes to the validator. The problem Solana is solving is open-state contention: high throughput does not remove execution uncertainty when hot accounts, liquidations, mints, swaps or arbitrage routes all compete at the same time.

Hyperliquid is narrower. It is not pricing generic compute. It is pricing certainty inside a purpose-built trading venue. That gives it a smaller surface area, but a more concentrated use case. Perpetual markets do not need priority fees because the chain cannot process transactions. They require them because milliseconds decide fills, queue position, stale-quote risk and liquidation edge.

From rounding error to revenue share

The share of trading fees captured by priority fees has moved from noise to signal in 11 weeks. In the opening week, priority fees were 0.04% of Hyperliquid's trading fees – a rounding error. By the week of 22 Jun, they hit 5.93%, before settling to 4.84% into the 2 Jul cutoff. Across the full run, priority fees have averaged 1.97% of trading fees, but the aggregate flatters nothing: the line is up and to the right, and the 22 Jun week alone generated $838,500 against a trading-fee base that barely moved.

(Source: Token Terminal, Hydromancer)

That divergence is the tell. Hyperliquid's weekly trading fees have oscillated in a $10–26mn band, while priority fees compounded through it. This is not a market getting pricier to trade. It is a new fee surface that went from negligible to material in less than three months, being discovered by the exact users it was built for, and their willingness to pay is scaling far faster than volume.

Concentration story

Two facts define priority fee flow, and they point in the same direction. The first is where the money sits by instrument. Since inception, equity and commodity perpetual futures – Hyperliquid's HIP-3 markets – have accounted for 59.4% of the $1.43mn generated by the top 40 markets, against 25.5% for native crypto markets. A single market, the SpaceX (SPCX) perpetual – a synthetic contract tracking the private rocket company's implied share price – pulled $265,100, or 18.6% of that top-40 total on its own. The top three markets took 37.5%; the top 10, 65.9%. The demand for speed is not spread across the book. It is stacked on the newest, least liquid, most information-sensitive instruments on the venue – a strategy-specific spend concentrated on the places where execution certainty is most valuable. Mature books with deep liquidity can absorb more flow before certainty becomes expensive. Newer, less stable markets expose the edge faster.

(Source: HL.eco)

The second fact is who pays. Of the $3.16mn in total priority fees, order-priority "write" fees make up $1.73mn (54.7%) and gossip "read" fees $1.43mn (45.3%). Both are concentrated, but the gossip auction is concentrated to the point of absurdity. Write fees follow a recognizable power law – the top 10 wallets account for roughly two-thirds of the total.

(Source: Hydromancer)

The read auction does not follow a power law so much as collapse into a monopoly: a single wallet accounts for more than half of all gossip bids, a second wallet makes it blow past the entire top-10 share of the write side, and five wallets represent 99.9% of gossip priority fees. This is not a market. It is a handful of latency shops and a rounding error of everyone else.

Follow those wallets and the structure resolves. The gossip bidders are, in the main, cost centres – treasury or management addresses bidding on behalf of servers that trade elsewhere, exactly as the mechanism permits. Some show no linked trading at all; the execution leg lives on separate infrastructure, consuming the raw stream. Others carry subaccounts with portfolios in the tens of millions, running automated market-making strategies across crypto, commodities and equities on HIP-3, some crossing 4,000 transactions a day – a cadence no human produces. The write side is more legible: block-inclusion buyers with eight-figure books, several active in outcome markets, running multi-leg market-making strategies where landing on the block is the entire game.

Path forward 

Priority fees become dangerous if they turn into a tax that degrades the book. If takers need to pay more just to interact, spreads can widen. If makers feel they can be picked off, they quote worse or leave. If every market becomes an auction for queue position, the venue can accidentally sell its own microstructure.

Hyperliquid's key protection is that priority fees do not override cancel priority. Cancels still execute before taker orders. That matters because it protects market makers from a pure pay-to-snipe model. A taker can buy better sequencing against other takers, but cannot simply pay to beat a stale-quote cancel. That design choice is the difference between monetizing latency and poisoning liquidity.

If priority fees stay concentrated in a few aggressive strategies, they are a useful but narrow burn stream. If they scale across HIP-3 markets and eventually broader Hyperliquid flow while preserving tight books, they become something much bigger.