The builder fee explained: why we only charge on mirrored volume

What a builder fee is on Hyperliquid

Most trading platforms bolt their fee onto the side of the exchange relationship: a subscription billed to a card, a performance cut computed in an internal ledger, or a spread markup buried in execution. Hyperliquid does something different. It has a first-class concept for third parties that submit orders on a user's behalf — builders — and a native mechanism for those builders to charge for it.

A builder fee is attached at the protocol level to orders the builder submits, and it can only be charged after the account owner has signed an approval that specifies a maximum rate. The user approves, the protocol enforces, and the fee is settled as part of the fill rather than invoiced afterwards. That has one property worth dwelling on: there is no billing relationship to trust. We cannot charge you more than the rate you approved, because the constraint is not our policy — it is Hyperliquid's.

So when we say the fee is 0.1% on mirrored volume, that is not a price list we could quietly revise. It is a permission you granted, visible on-chain, capped at signing time, and revocable with your own key.

How the 0.1% works in practice

The fee is charged on the notional value of mirrored fills. When the system opens a mirrored position in one of your isolated sub-accounts, the fee applies to the notional of that fill. When the position is closed — because the leader closed it, because sizing changed, or because that leader was replaced — the fee applies to the closing notional as well. Round trips are therefore charged twice, once on each side, which is the same convention Hyperliquid's own trading fees follow.

Three consequences follow directly and are worth stating explicitly rather than leaving to be discovered:

• Cost scales with turnover, not with account size. A leader who trades frequently generates more mirrored notional than a leader with the same allocation and a slower holding period, and therefore more fee. Following high-turnover strategies costs more in absolute terms, and this is the single largest driver of your total fee.
• The fee is separate from Hyperliquid's own fees. The exchange charges its own maker and taker fees on every fill according to its published schedule. The builder fee is additive to those, not a replacement for them, and Hyperliquid's fees go to the protocol rather than to us.
• Replacement and rebalancing are not free. Closing a removed leader's positions pays the fee on the closing notional. That is a real reason to prefer gradual weight reduction over frequent removals, and it is why score decay reduces allocation continuously instead of triggering an exit at the first sign of weakness.

Why volume-based beats performance fees and subscriptions

Performance fees are the conventional choice in managed trading, and they are more problematic than their reputation suggests. A share of profits requires an accounting layer — a high-water mark, a crystallisation schedule, a definition of what counts as a gain — and every one of those choices is a judgement call made by the party being paid. That layer also has to be applied to realized and unrealized positions somehow, which creates a well-known asymmetry: a performance fee pays on the upside and shares none of the downside, so it rewards volatility as such. A strategy with wild swings can generate more fees than a steadier one that ends in the same place. Nothing about that structure is transparent from the outside, and its complexity is entirely in the operator's favour.

Subscriptions have the opposite failure. A monthly charge accrues whether the system trades or not, whether you have funds deployed or not, and whether you are getting anything at all. It converts inactivity into revenue, which is the worst possible thing to make profitable for a system whose job is to decide when not to have exposure.

A volume fee removes both problems. There is nothing to compute beyond notional times a fixed rate, and you can verify it yourself from your own fill history — no internal ledger, no reconciliation, no accounting definitions you have to accept. It bills the thing the system actually does for you, which is execute mirrored trades, and it charges nothing in the periods when it does nothing.

What the fee does not include

Absence of charges is as important as their presence, and it is easier to state precisely.

• No performance fee and no cut of your equity. We take no share of your profits and hold no claim on your balance. Gains are entirely yours.
• No subscription or monthly minimum. There is no recurring charge and no floor. If nothing is mirrored in a given period, the fee for that period is zero.
• No inactivity fee. A dormant account costs nothing to keep. We do not charge for the absence of trading.
• No deposit or withdrawal fee from us. Funds stay in accounts you own throughout, and moving them is between you and Hyperliquid. We charge nothing on either direction.
• No spread markup or hidden execution cost. Mirrored orders go to Hyperliquid's order book at market prices. We do not internalise flow, quote against you, or add anything to the execution price. Slippage is a market outcome, not a fee.

Capped and approved upfront

The builder fee approval is a distinct action from the agent approval, and it exists specifically so the fee has a ceiling that does not depend on our conduct. When you approve it, you sign a maximum rate with your own wallet. Hyperliquid will not settle a builder fee above that rate, so the cap is enforced by the venue.

Practically, this means the fee cannot be raised on you quietly. Increasing it would require a new approval signed by you, which is a visible action you would have to take deliberately. It also means you can end the arrangement unilaterally: revoking is done with your own key, without moving funds and without needing anything from us. The two approvals are independent — trading authority and fee authority are separate permissions with separate scopes — which is exactly the property you want in a permission model.

This is the same design principle as the custody model. What matters is not that we promise to behave; it is that the boundary is enforced somewhere we do not control.

Where the incentives align, and where they do not

Volume-based pricing aligns some incentives well and creates one tension. Both deserve to be said out loud.

The alignment: we are paid only when the system is actually working — mirroring trades into your sub-accounts. We are not paid to hold your deposit, not paid during inactivity, and not paid more when your account swings violently. Because we take no share of profits, we have no reason to prefer a leader with a fat-tailed return profile over a steadier one, which is a real distortion in performance-fee structures. And because we hold no equity claim, our revenue does not grow simply because your balance did.

The tension, stated plainly: a fee on volume means more turnover produces more revenue. That is a structural incentive toward activity, and no amount of framing makes it disappear. What bounds it is that turnover is not ours to manufacture. Mirrored volume is generated by the leaders' own trading, not by us deciding to trade — we replicate their fills, and we do not add trades that a leader did not make. The system-initiated trades that do exist are replacements and weight adjustments, and those are governed by the composite score rather than by discretion, which is one of the more concrete reasons the replacement path is rules-based rather than manual. Score-weighting also pushes the same direction: reducing a decaying leader's weight is a repricing, while removing them is a taxable exit, so the default response to weakening evidence is the cheaper one.

You should still hold us to the observable version of this. Your fill history is on-chain and so is every mirrored trade in every sub-account. Turnover that does not correspond to leader activity would be visible.

Transparent monetization

A fee model is a design document. Performance fees say the operator wants exposure to your upside without your downside. Subscriptions say the operator wants revenue whether or not the system does anything. A capped, protocol-enforced fee on mirrored notional says the operator gets paid for executing trades, at a rate you approved, verifiable from your own account history, and revocable when you decide.

It is not a claim that our costs are the lowest possible, and it does not make trading cheap — Hyperliquid's own fees apply, turnover compounds them, and high-frequency leaders cost more to follow than slow ones. The claim is narrower and checkable: there is one fee, you know its rate, you approved its cap, and there is nothing else.

For the complete picture — including Hyperliquid's own fee schedule, how turnover interacts with tracking error, and what replacement costs — the documentation lays out the full cost structure and the risk disclosure covers what fees do not mitigate.