Deployer economics and risk

A deployer trades a large HYPE lock for a share of its markets’ fees. It also carries an unpriced slashing risk and a duty to run or settle markets correctly (trader fees). At an illustrative 91.25 USD per HYPE and a 10 per cent hurdle, a HIP-3 deployer earning half the 0.045 per cent base perp taker fee (scale 0.5) breaks even at about 55.6M USD of volume a day.

What you earn: the fee share

Standard	Deployer’s share	Rule
HIP-1	fees collected in the token’s base asset	0 to 100%, starts at 100%, then decrease-only; the rest is burned
HIP-3	scale per asset, formula below	0 to 3, or 0 to below 10 in growth mode; changes once per 30 days on mainnet
HIP-4	deployerFeeScale times the base outcome fee	0 to 10; the trader pays base x (scale + max(scale, 1)), the protocol keeps the rest, never under 1x base; charged when closing or settling, never when opening, at the spot schedule (7,256 mainnet fills); no maker rebates

On HIP-3, with x the normal fee rate and y the rate after any aligned quote asset discount (else x = y), the trader pays P to the protocol plus D to the deployer:

x > 0, scale < 1:   P = y            D = scale * x
x > 0, scale > 1:   P = y * scale    D = x * scale
x < 0, scale < 1:   P = y / (1 + scale)   D = y * scale / (1 + scale)
x < 0, scale > 1:   P = y / 2        D = y / 2

With a normal fee of 1 unit and no aligned collateral:

growthMode	scale	To protocol	To deployer	Trader pays
off	0	1	0	1
off	0.5	1	0.5	1.5
off	1	1	1	2
off	3	3	3	6
on	1	0.1	0.1	0.2
on	9.99	0.999	0.999	1.998

Growth mode (see the numbers) at a scale just under 10 pays the deployer about what scale 1 pays outside it, hence the HIP-3 page’s 0 to 300 per cent (0 to 100 per cent in growth mode). From scale 1 up the deployer gets exactly half of what the trader pays (D = P), below it less. The general fees page’s “keep up to 50% of trading fees” for spot and HIP-3 deployers fits HIP-3 only in that sense; for HIP-1 the spec allows 100 per cent. On HIP-4 at tier 0 (0.070 per cent spot taker), a closing taker at scale 1 pays 0.14 per cent of notional before discounts, 0.07 per cent to the deployer.

What the stake costs you

The stake earns only staking rewards and cannot fall below the requirement while live, so it costs its forgone return. The opening example in full:

capital          = 500,000 HYPE * 91.25 USD          = 45,625,000 USD, about 45.6M
carry per year   = 45,625,000 * 0.10                 = 4,562,500 USD
carry per day    = 4,562,500 / 365                   = 12,500 USD
deployer take    = 0.00045 * 0.5                     = 0.000225 of volume
breakeven volume = 12,500 / 0.000225                 = 55,555,556 USD per day, about 55.6M

That volume is needed every day the stake is held. Inputs scale linearly:

Hurdle	Deployer keeps 25% of 0.045%	50%	100%
5%	55.6M USD a day	27.8M	13.9M
10%	111.1M	55.6M	27.8M
20%	222.2M	111.1M	55.6M

• It counts only the taker fee: a tier-0 maker’s 0.015 per cent adds a third to the take. It also ignores staking rewards (delegated stake keeps earning them) and slashing.
• Live markets mostly run growth mode at scale 1, where the deployer keeps 0.1 of the normal fee: breakeven at 10 per cent is then about 277.8M USD of taker volume a day. 156 of 168 listed HIP-3 assets do, with about 98 per cent of 3.93B USD of 24-hour HIP-3 volume; the other 12 run scale 1 outside growth mode, one 0.5.
• The HIP-3 page expects the requirement to fall; at 50k HYPE breakeven falls tenfold.
• Stakes stack across HIP-3 and HIP-4; bands on another’s HIP-3 perp skip the HIP-3 stake (composition).

How long the capital is gone

The lock ends at the later of 183 days and a settled book; the queue then adds 7 days (about 5 minutes on testnet).

• Lock. A HIP-3 stake frees once every asset is settled, a HIP-4 one once no outcome is active; an unsettled book extends the 183-day lock indefinitely. The 5 deployers whose dexes list no asset hold 0.002 to 3,876 HYPE.
• Queue. 7 days, a HIP-3 stake still slashable; one mainnet release took 7 days and 1 second.
• Headroom. Only stake above the requirement can leave. On testnet, 102 HYPE against 100: 1.5 withdrawn, the next 1.0 refused (Insufficient balance to withdraw from CStaking), same signers.

Slashing burns the stake

HIP-3 slashing (rules, tiers in the numbers) is a stake-weighted vote that burns the HYPE, regardless of intent or funder. Per the docs, HIP-4 markets contradicting semanticRestriction are slashable; the announced design added wrong settlement and markets unsettled over a week, which the testnet release disclaimed.

No slash is on record: all 12 mainnet deployer balances checked reconcile with their deposits, rewards and requested withdrawals. Any slashing probability is a prior, and a burn leaves nothing to recover.

Running the market is a liability too

The second exposure is operational: a HIP-3 oracle drives every position’s margin and liquidation (operating), and a HIP-4 deployer settles each outcome with no deadline or documented amend (settling).

Why the stake cannot be delegated

A holder cannot lend a deployer its stake and keep custody:

• Delegation goes to validators. Precompile structs record a delegation’s validator, never a beneficiary; a third party can only give HYPE to the deployer outright or delegate to a validator.
• Only the deployer’s own staking account counts, undelegated HYPE included (measured).
• Staking links are refused. The one documented way to count another’s stake (towards a fee tier) got an empty account linked to a fully staked one refused: Deployer cannot have staking link.
• Operation separates from the bond, one way. Sub-deployers and fee recipients hold no stake (0 of 34 HIP-3 ones on mainnet); agent-signed activation and withdrawal were refused, naming the agent.

What does work: co-operating through a quorum

The deployer can itself be a native multi-sig. On testnet, a fresh account was converted while empty to 2-of-2 between an operator and an underwriter:

Step	Operator alone	Quorum
Staking deposit	refused, Multi-sig threshold not met	accepted
Activation	refused	accepted
Withdrawal of headroom	refused	accepted, into the unstaking queue
setSubDeployers grant		sent; the operator then registered and settled alone; read back on chain

The operator runs the venue at one signature, neither party moves the bond alone, and the protocol escrows the requirement against both, so the funder can lose only to slashing. Revoking a grant needs the deployer, making the grant set a term of the deal.

CoreWriter does not work for multi-sig, so custody cannot be a HyperEVM contract. 8 of 13 mainnet deployers are multi-sig (14 venues, km and mkts sharing one), thresholds 2-of-3 to 4-of-8.

The numbers

Parameter	Value
Stake, HIP-3 and HIP-4 mainnet	500k HYPE each, stacking
Stake, testnet	100 HYPE
Minimum lock	183 days
Unstaking queue	7 days mainnet, slashable; about 5 minutes testnet
HIP-1 fee share	0 to 100%, decrease-only
HIP-3 fee scale	0 to 3; 0 to below 10 in growth mode; change once per 30 days
HIP-3 growth mode	protocol fees, rebates, volume contribution and L1 user rate limit contribution cut 90%; only markets disjoint from validator-operated perps; 30-day cooldown per asset
HIP-4 fee scale	0 to 10; trader pays base x (scale + max(scale, 1))
HIP-3 slashing guideline	up to 100%, 50%, 20% by severity
Worked example	45.6M USD locked; 12,500 USD a day at 10%; 55.6M USD a day breakeven
Multi-sig authorised users	at most 10

Monitor the covenant, not the delegation list

• The protocol tests delegatorSummary.delegated + delegatorSummary.undelegated, readable at HyperEVM precompile 0x805.
• totalPendingWithdrawal above zero is the only warning of a deployer pulling its bond, 7 days ahead.
• A funded deployer must never carry a staking link; test exits against mainnet timing, not testnet’s 5 minutes.

Limits of the public data

• Do not model a shared bond: one mainnet address deployed 2 HIP-3 dexes (km, mkts) and holds 500,454 HYPE, but km lists no asset, so its requirement may already be released.
• Rewards accrue to the bond: reconciled mainnet deployer balances held surpluses of 0.0029 to 17,562 HYPE.
• HIP-4 fee income is invisible: no endpoint reports it, and the info API has no volume per venue.

For the slashing rules in full, see HIP-3; for the custody research, see collateral underwriting.