Kinetiq risk

TL;DR

Kinetiq issues kHYPE, a liquid staking token for HYPE on HyperEVM. Deposited HYPE is delegated to Hyperliquid validators and kHYPE accrues yield through a rising exchange rate, not through rebasing. This is our first report on it, opened because the $10k showcase holds 69.275 PT kHYPE on Pendle maturing 2026-09-24 and had never assessed the issuer behind it.

Disambiguation first, because it is a live hazard: kHYPE (Kinetiq) is not stHYPE (StakedHYPE / Thunderhead), which DeFiLlama's price feed lists as staked-hype at $57.05 on 2026-08-05. Both trade on HyperEVM and both have their own Pendle market on chain 999, and so do vkHYPE and haHYPE. The vkHYPE market even shares our exact maturity date of 2026-09-24. Nothing below is sourced from any of them.

The mechanics check out better than most things we review. I reproduced the exchange rate from its own components: totalStaked 51,945,824.61 plus totalRewards 524,256.95 minus totalClaimed 38,436,958.88 gives 14,033,122.68 HYPE backing 13,723,433.37 kHYPE, which is 1.0225664594511592, and kHYPEToHYPE(1e18) returns 1.022566459451159247 at the same block. That is agreement to the last wei from an independent path. The secondary market agrees too: the four deepest kHYPE/WHYPE pools quote 1.02241 to 1.02246, about 0.012% under redemption. The yield is real staking yield, and the number reconciles to Hyperliquid's validator APR less the documented fee.

The verdict is caution and the red is admin control. One 4-of-8 Safe owns the ProxyAdmin of the kHYPE token itself and of five more core contracts, holds the only DEFAULT_ADMIN_ROLE and MANAGER_ROLE on all four managers, holds all four pause roles, and holds DEFAULT_ADMIN_ROLE on kHYPE. There is no timelock anywhere in that path. Four signatures replace the implementation of the token our PT redeems into, and the change lands in one block. All eight signers are plain EOAs; how those keys are held is not visible on-chain and is not published.

Checklist

Audits and contracts. The audit record is genuinely strong and, unusually, still current. I read Kinetiq's contracts page rather than trusting a summary of it, and all eight reports are there with the dates as claimed: Pashov Audit Group 2025-03-06, Zenith 2025-03-17, Code4rena 2025-04-16, Spearbit 2025-06-24, Pashov again 2025-11-23 on the instant-unstake work, Zenith 2025-11-24 and Spearbit 2025-11-27 on kmHYPE, Spearbit 2026-01-15 on sKNTQ. Four distinct firms, one of them a competitive audit, no invented names. The Cantina bounty is live, dated 15 Sep 2025, pays up to $5,000,000 for a Critical against a stated $1,000,000 pool, and scopes exactly the seven contracts that matter: kHYPE, StakingManager, StakingAccountant, ValidatorManager, PauserRegistry, OracleManager and OracleAdapter. The two bounty figures are stated separately on that page and I am reporting both rather than reconciling them.

Three things keep this yellow. Every core contract is an OpenZeppelin transparent proxy, so an audit certifies bytecode that can be swapped. And two rows on the published contracts page do not describe the live system.

The first is PauserRegistry, and the detail matters. In the kHYPE table the displayed address is 0xac03CABA51e17c86c921E1f6CBFBdC91F8BB2E6b, which on HyperEVM has no code, no balance and nonce 0. The hyperlink on that same row points at 0x752E76ea71960Da08644614E626c9F9Ff5a50547, which does have code and is what the live StakingManager returns from pauserRegistry(). So the link is right and the text is wrong. Click it and you land on the real contract; copy the address and you get a dead one. Every other row on that table has link and text agreeing.

The second is the oracle adapter. The page lists DefaultOracle and OracleAdapter as the same address, 0xefbcCc6E33DA1C1ef638cBc0F044968D0f590fED. On the live OracleManager that address returns false from both isAuthorizedOracle and isActiveOracle. The single authorized and active adapter is 0x26fe3019b55f65f454c6b5f8baeea1ed405544c9, which is not on the page at all.

Neither is a fund risk on its own. But on a page whose entire job is letting an outsider verify who controls what, two of the kHYPE rows send that outsider to the wrong contract, and one of them is the pause control.

Admin control. Read directly from HyperEVM rather than from any docs claim, which is how it should be done. The address 0x18a82c968b992d28d4d812920eb7b4305306f8f1 is a Gnosis Safe, VERSION() returns 1.3.0, getThreshold() returns 4 and getOwners() returns 8 addresses. It has executed 62 transactions and has no Safe modules enabled, so there is no side door around the threshold. That one Safe is:

the owner of the ProxyAdmin behind the kHYPE token (0x9c1e8db0), behind StakingManager (0x8194aa9e), behind ValidatorManager (0xf8b427d3), behind StakingAccountant (0x7957de69), behind OracleManager (0x4c0a83d7) and behind the InstantUnstakePool (0x1ab8a96e); the only holder of DEFAULT_ADMIN_ROLE on StakingManager, ValidatorManager, StakingAccountant and OracleManager, one member each; the only holder of MANAGER_ROLE on all four, one member each; and the sole holder of DEFAULT_ADMIN_ROLE, PAUSER_ROLE, UNPAUSER_ROLE and PAUSE_ALL_ROLE on the live PauserRegistry.

There is no TimelockController anywhere in that chain of custody. Each ProxyAdmin's owner is the Safe directly. The delay between four signatures and a new kHYPE implementation is zero.

Now the part that is easy to get backwards, and worth stating carefully because the reassuring version of it is wrong. kHYPE's MINTER_ROLE and BURNER_ROLE do each have exactly one member, the StakingManager contract, so no key mints kHYPE today. That is not the same as saying a key cannot. The Safe holds DEFAULT_ADMIN_ROLE on the kHYPE token, and getRoleAdmin(MINTER_ROLE) returns DEFAULT_ADMIN_ROLE. So the Safe can call grantRole(MINTER_ROLE, anyAddress) and mint immediately. No upgrade, no delay, one transaction. The single-member role list looks like a constraint and is really just a current state.

What MANAGER_ROLE can do short of that is the day-to-day surface, and I read it from the audited source rather than inferring. setDelegation, which chooses which validator a staking manager's HYPE goes to, is onlyRole(MANAGER_ROLE). So are rebalanceWithdrawal, activateValidator, deactivateValidator, setWithdrawalDelay, pauseStaking and pauseWithdrawal. setUnstakeFeeRate is the same role and is bounded: require(newRate <= 1000, "Fee rate too high"), so the standard unstake fee is capped at 10%. It currently reads 0 against a BASIS_POINTS() of 10000. A 10% ceiling is a real bound and worth crediting, but it is a large haircut, not a small one.

Read the StakeHub page next to all this. It says validator selection is "Quantitative validator selection" with "No human bias or subjective decision-making." The scoring may well be quantitative. The on-chain write that actually moves stake is a multisig call. Those are not the same claim.

Oracle. There is no price oracle in the redemption path, which removes the failure mode people usually worry about. What exists is a performance and slashing oracle, and it matters because of one sign in one formula. StakingAccountant computes, verbatim from source, uint256 totalHYPE = totalStaked + rewardsAmount - totalClaimed - slashingAmount; where slashingAmount is validatorManager.totalSlashing(). That value is written by reportSlashingEvent, which is onlyRole(ORACLE_MANAGER_ROLE). On-chain that role has one member and it is the OracleManager contract, 0x192826e4. A contract rather than a key is good design.

The draft I reviewed left the next question open, so I closed it. OracleManager's generatePerformance(address) is onlyRole(OPERATOR_ROLE), and that function is the path that calls validatorManager.reportSlashingEvent. OPERATOR_ROLE has one member, 0x23a4604c, an EOA with 2.49 HYPE and 13,974 transactions, a hot key in an operating loop. So yes, the operator can trigger a slashing report. What it cannot do is choose the number: the magnitude is derived from authorized oracle adapters, of which there is exactly one active today, and it is filtered by a ValidatorSanityChecker that is set and live at 0x8b641d61 with 8,934 bytes of code.

Two qualifications keep this from being clean. setMaxPerformanceBound is itself onlyRole(OPERATOR_ROLE), so the operator can widen the bound its own reports are checked against. And setSanityChecker is onlyRole(DEFAULT_ADMIN_ROLE), held by the Safe, while the code says "Skip sanity check if sanityChecker is not set (address zero)". Setting it to zero disables the check rather than failing closed.

Three facts pull the other way and are why this is yellow rather than red. totalSlashing() reads exactly 0. The protocol has never recorded a slashing event. And Hyperliquid's own docs say "There is currently no automatic slashing implemented," reserving it in principle for double-signing. Jailing is the live penalty, and the docs are clear on its effect: "A jailed validator does not produce rewards for its delegators." That costs yield, not principal. None of Kinetiq's 14 validators is jailed or inactive today.

Liquidity and exit. Three exits, and the middle one is the one the first pass missed.

The queue is the real one and it works. Kinetiq documents "~8-9 days total: includes 1-day delegation lockup + 7-day unstaking queue", and both legs are confirmed against Hyperliquid's docs, which state "Delegations to a particular validator have a lockup duration of 1 day" and that "transfers from staking account to spot account have a 7 day unstaking queue". The 7-day leg also reads on-chain as withdrawalDelay() = 604800. Neither leg is Kinetiq's to shorten; the queue is a chain parameter. Right now 203,730.60 kHYPE sits in totalQueuedWithdrawals(), roughly 1.5% of supply, and Hyperliquid reports 62,372.68 HYPE in flight with four pending withdrawals for the StakingManager. Ordinary flow, not a run.

There is also a protocol-native instant exit, which the Kinetiq kHYPE docs page does not mention and which I found by probing the live contract. StakingManager.instantUnstakePool() returns 0x665b6779, a proxy holding 85,967.32 HYPE, roughly $4.9M, with instantUnstakeFeeRate() reading 10 basis points, so 0.10%. This is the feature Pashov audited in November 2025 and the FAQ's 0.10% withdrawal fee. It is a genuine third door and it should be counted. It is also, note, a sixth contract whose proxy admin is the same Safe.

The DEX book adds about $7.42M across the four meaningful kHYPE/WHYPE pools: nest $4.18M, prjx $1.52M, ramses $1.33M, hyperswap $0.40M. Roughly $7.5M if you count all fifteen kHYPE/WHYPE pairs, but the rest are dust. Combined 24h volume is about $1.38M and 98% of it is on ramses.

Add the instant pool to the DEX book and same-day capacity is about $12.3M against $814.3M of kHYPE outstanding. That is roughly 1.5% of supply. Better than the picture without the instant pool, and still thin. Anyone needing out in size on a bad day is choosing between a week and a half of waiting and a bad fill. September 2025 is the proof rather than the hypothesis, and it is under track record below.

For our own position the size question is not close. The Pendle kHYPE market 0xb48b0c95b2ddc464484305b7363fad5bd5b7a683 holds $3,210,116 of liquidity and our 69.275 PT marks at about $3,950, which is 0.12% of the pool. We can leave whenever we want. What the thin underlying book does is move the mark, not block the door.

Yield: real versus emission. Real, and the arithmetic closes from two directions. Pendle reports underlyingApy of 1.9821% for kHYPE with underlyingRewardApy at 0, meaning no reward-token component. Our own Hyperliquid refresh the same day put the median predicted validator staking APR at 2.18%. Kinetiq documents "There is a 10% fee on staking rewards, of which 70% is used to purchase KNTQ on the open market and 30% is sent to the Kinetiq treasury for operations." Take 2.18% less 10% and you get 1.96% against an observed 1.98%. Nothing here is printed. Note what the fee does, though: 7% of gross staking yield is diverted into buying the protocol's own governance token. That is a transfer from kHYPE holders to KNTQ holders, disclosed and modest, but not free. Absolute yield is low, which is a return question rather than a risk one.

Holder concentration. For an LST the concentration that can lose you money is validator concentration, so start there, and answer the question our Hyperliquid report forces: does Kinetiq spread the chain's stake concentration or deepen it? Both, in different directions, and it is worth being precise.

At the network level it spreads. Kinetiq delegates 13,817,998.57 HYPE across 14 validators, about 3.17% of the 435,722,050 HYPE of active stake, and not one unit of it goes to a Hyper Foundation validator. All five foundation validators, which together hold 48.87% of active stake, are absent from Kinetiq's delegation set. I checked each of the 14 individually: all active, none jailed, and the stake-weighted commission across them is 2.31%, the highest being Anchorage By Figment at 10%. That is a real, deliberate counterweight to the concentration our Hyperliquid report rates red.

Internally it concentrates. HyperStake takes 4,700,248 HYPE, 34.02% of everything Kinetiq delegates. Kinetiq x Hyperion takes 3,454,495, exactly 25.00%. Those two are 59.02% of the book, the top three are 63.24%, the top five are 70.09%, and ValidatorManager contains no cap on per-validator allocation. Kinetiq is also the dominant delegator to both: it supplies about 47% of HyperStake's total stake and about 54% of Kinetiq x Hyperion's. And the second-largest allocation goes to a validator carrying the protocol's own brand. I am not claiming that is self-dealing, and a protocol operating a validator is common practice. I am claiming that a quarter of user stake sitting on a house validator, chosen by a multisig call, is not what "No human bias or subjective decision-making" leads a reader to expect.

Token-holder concentration is the part I did not finish. Spot checks show Pendle's SY-kHYPE holding 2.29% of supply and the StakingManager escrow 1.45%, with the Safe and treasury holding effectively nothing. No dominant holder turned up in what I sampled, but that is not a ranking and I am not presenting it as one.

Track record. Twelve and a half months, first DeFiLlama datapoint 2025-07-17. In that window I found no exploit, no hack, no contract incident, no emergency pause and no slashing, across web search and a targeted incident search. stakingPaused() and withdrawalPaused() both read 0 today and totalSlashing() reads 0. That is a clean record as far as it goes, and twelve months is not far.

One event is on the board and it is the relevant one. PeckShieldAlert reported on 2025-09-28 that between 24 and 27 September, kHYPE deviated from its peg and bottomed at 0.8802 against WHYPE before the peg was restored. I could not open the original post this run, so I am reporting it on two independent secondary sources that quote the same figure and dates rather than on the post itself. Read what that is and is not. It is a secondary-market dislocation on those thin pools, not a failure of redemption; the queue kept paying the exchange rate throughout. But for anyone holding a claim that marks off the kHYPE price rather than off redemption, which describes a Pendle PT before maturity, the secondary price is the price. The gap was at least 12% against a WHYPE par of 1.0, and wider than that against the redemption rate, which sat above 1.0 at the time. It stayed open for three days on an asset whose contract-level backing never moved.

The TVL trend needs one correction that the USD number hides. Parent TVL is $892.1M against $1,141.6M thirty days ago, which reads as minus 21.9%. But HYPE was about $70.59 then and about $57.12 now. In HYPE units the protocol went from roughly 16.2M to roughly 15.6M, a few percent. Over 90 days the USD figure looks flat, up 3.7%, and that is misleading in the other direction: HYPE rose from about $42.76, so in HYPE units Kinetiq shrank from roughly 20.1M to 15.6M, about 22%. These unit conversions move with whichever intraday price mark you take, so treat them as approximate. The shape is what matters: steady, unhurried outflow, no cliff.

Worst case

The one that costs principal: four of the eight Safe signers are compromised or coordinate. There is no timelock, so the next block carries whatever they want. They do not even need an upgrade for the fastest version, since the Safe already holds DEFAULT_ADMIN_ROLE on kHYPE and can grant MINTER_ROLE to an address of its choosing in a single call. Or it replaces the implementation and rewrites the accountant so kHYPEToHYPE returns whatever it likes. Everything downstream that prices kHYPE off the protocol's own exchange rate, our Pendle PT included, is then priced off a number the attacker writes. Nothing warns you first, because the only thing between the signature and the effect is the signature. There is no insurance fund. Position sizing is the mitigation, and for us it already is: about $3,950 of PT.

The one that is far likelier and has a rehearsal: HYPE sells off hard, leveraged kHYPE loops on HyperEVM unwind, the $4.9M instant pool drains, and the remaining $7.4M of DEX depth absorbs the rest at a discount, exactly as in September 2025. The redemption path still works and still pays 1.0226 or whatever the rate is by then, but it takes 8 to 9 days you may not have. Our PT would mark down with it. Since we hold this as a hedge leg rather than a directional bet, the failure is not the loss, it is the hedge going quiet at the moment it is supposed to work.

The slower one: generatePerformance produces a bad slashing figure, whether from a compromised operator key, a widened performance bound, or a sanity checker set to zero. totalSlashing goes above zero and the exchange rate steps down permanently for every holder. Hyperliquid has no automatic slashing today, so a genuine slash is not currently on the table. A reported one is a software path, and software paths do not require the underlying event to be real.

Bottom line

Caution. This is the best-constructed liquid staking token on Hyperliquid and the research kept confirming that rather than undercutting it: four audit firms across eight reports with the most recent in January 2026, a live bounty scoped to the right seven contracts, an exchange rate I reproduced independently to the last wei, real staking yield that reconciles to the validator APR net of a disclosed fee, a secondary price sitting 0.012% off redemption, a working instant-exit pool, and a validator set that pointedly excludes the five Hyper Foundation validators our Hyperliquid report rates red.

It is caution because of one thing, and worst-flaw-wins means that is enough. A single 4-of-8 Safe with no timelock holds proxy upgrade rights over six core contracts including the token itself, every management role on every manager, every pause role, and the ability to grant itself minting rights in one transaction. Eight EOA signers, custody undisclosed. Everything good above is revocable by four signatures in one block.

On the external read: Yearn Curation scores Kinetiq kHYPE 2.3 on a 1 to 5 scale, band "Low Risk", in a report dated 2026-05-19. I checked their breakdown, and it is more interesting than the headline. Their five weighted categories are Audits and Historical 2.00, Funds Management 1.75, Liquidity Risk 2.00, Operational Risk 2.00, and Centralization and Control 3.30. Centralization is by a wide margin the worst thing on their own scorecard, and it carries their heaviest weight alongside Funds Management. So we do not actually disagree about the protocol. We disagree about what to call it. "Low Risk" is not a fair label for instant, timelock-free multisig upgradeability over the asset itself plus a one-call path to minting, and it does not price a $12.3M fast exit against $814M of token that has already printed 0.8802. One notch more conservative is the right call.

Practical rules for the showcase this month. The PT position is fine at its current size and the exit is not a problem: about $3,950 against $3.21M of Pendle liquidity, maturity 50 days out on 2026-09-24, and the 8 to 9 day unstake queue fits comfortably inside that if we ever need to take delivery and redeem. Do not size up on this leg without a reason. And carry the jitoSOL lesson across: one kHYPE is 1.0226 HYPE today and rising, so PT units, kHYPE units and HYPE units are three different things. Anything that compares raw counts across them will lie, the same way the hedge-drift alert lied when it read jitoSOL against SOL one for one.

Data appendix

• TVL: $892,122,255 for parent slug kinetiq and $814,312,212 for child slug kinetiq-khype (DeFiLlama, both re-verified this run). The $77,810,043 gap is fully accounted for by the other children: Kinetiq Earn $40,784,399, Kinetiq kmHYPE $34,309,177, Kinetiq Launch $2,716,470, summing to $77,810,045, agreeing to about $2 of rounding. Our watchlist slug is kinetiq, so $892.1M is the headline and $814.3M is the kHYPE line that our PT actually depends on. Parent 90d $860,243,847 (2026-05-08), 30d $1,141,602,507 (2026-07-07), 7d $850,554,292 (2026-07-30). First datapoint 2025-07-17 at $458,111,649. Parent ATH $2,653,266,457 on 2025-10-04, so 66.4% below peak.
• TVL in HYPE terms: HYPE was about $70.59 on 2026-07-06, $42.76 on 2026-05-07, and $57.12 spot on 2026-08-05 (DeFiLlama coins API). So the 21.9% USD decline over 30 days is a few percent in HYPE units, roughly 16.2M to 15.6M HYPE, and the 90-day USD gain of 3.7% is about a 22% decline in HYPE units, roughly 20.1M to 15.6M. Sensitive to the intraday mark chosen; approximate.
• Exchange rate: StakingAccountant.kHYPEToHYPE(1e18) = 1,022,566,459,451,159,247, so 1 kHYPE = 1.022566459451159247 HYPE. Inverse HYPEToKHYPE(1e18) = 0.977931547389818248. Reproduced from components: totalStaked 51,945,824.60640136 + totalRewards 524,256.952325888 - totalClaimed 38,436,958.88248355 = 14,033,122.6762437 HYPE over a kHYPE supply of 13,723,433.373490151, which is 1.0225664594511592, matching the contract to the wei. Formula verbatim from source: uint256 totalHYPE = totalStaked + rewardsAmount - totalClaimed - slashingAmount;. totalSlashing() = 0. At the live HYPE mark of $57.12 that implies kHYPE fair value about $58.41; Pendle marked kHYPE at $58.4294 the same minute.
• Secondary peg, live: kHYPE/WHYPE native quotes 1.02244 (nest), 1.02241 (prjx), 1.02242 (ramses), 1.02246 (hyperswap), against an on-chain redemption rate of 1.0225665. Market is about 0.012% under redemption, effectively at par.
• Depeg history: PeckShieldAlert, 2025-09-28: kHYPE briefly deviated from its peg between 24 and 27 September, bottoming at 0.8802 against WHYPE, peg since restored. The original X post returned HTTP 402 this run, so this is carried on two independent secondary sources quoting the same figure and dates. Measured as the kHYPE/WHYPE ratio, not USD, and not against the redemption rate, which was above 1.0 at the time. No other depeg found, and no exploit, hack, pause or slashing found through 2026-08-05.
• Audits: eight reports, verified against Kinetiq's contracts page this run. Spearbit 2026-01-15 (sKNTQ), Spearbit 2025-11-27 (kmHYPE), Zenith 2025-11-24 (kmHYPE), Pashov Audit Group 2025-11-23 (kHYPE LST instant unstake), Spearbit 2025-06-24 (kHYPE LST), Code4rena 2025-04-16 (kHYPE LST), Zenith 2025-03-17 (kHYPE LST), Pashov Audit Group 2025-03-06 (kHYPE LST). Bug bounty on Cantina, live since 15 Sep 2025, maximum $5,000,000 for Critical against a stated $1,000,000 pool, scoped to kHYPE, StakingManager, StakingAccountant, ValidatorManager, PauserRegistry, OracleManager and OracleAdapter.
• Admin and governance (all read on-chain): upgrade authority is Safe 0x18a82c968b992d28d4d812920eb7b4305306f8f1, VERSION() 1.3.0, getThreshold() 4, getOwners() 8, nonce() 62, getModulesPaginated returns an empty module list. It owns the ProxyAdmin of kHYPE (0x9c1e8db004d8158a52e83ffdc63e37eabea8304c), StakingManager (0x8194aa9eca9225f96a690072b22a9ad0dd064f64), ValidatorManager (0xf8b427d31bf581d56dc50d5d40d861a40b2374f2), StakingAccountant (0x7957de697ea42238f53f60ca7ad55b7b03cc0bda), OracleManager (0x4c0a83d7244efb4843e2dc849068389292d1f441) and InstantUnstakePool (0x1ab8a96edae0d39ce90e6d845de3173f4d4e9859). It is the sole DEFAULT_ADMIN_ROLE and sole MANAGER_ROLE holder on StakingManager, ValidatorManager, StakingAccountant and OracleManager (one member each), and sole holder of DEFAULT_ADMIN_ROLE, PAUSER_ROLE, UNPAUSER_ROLE and PAUSE_ALL_ROLE on the live PauserRegistry. No TimelockController in the path; each ProxyAdmin's owner is the Safe directly. All eight signers are EOAs with zero code (nonces 0, 27, 0, 10, 13, 23, 12, 12). kHYPE MINTER_ROLE and BURNER_ROLE each have exactly one member, the StakingManager, but the Safe holds DEFAULT_ADMIN_ROLE on kHYPE and getRoleAdmin(MINTER_ROLE) returns DEFAULT_ADMIN_ROLE, so the Safe can grant minting rights in one call without upgrading. OPERATOR_ROLE on StakingManager and OracleManager is a single EOA, 0x23a4604cdfe8e9e2e9cf7c10d7492b0f3f4b4038, zero code, 2.49 HYPE, nonce 13,974. ORACLE_MANAGER_ROLE on ValidatorManager is the OracleManager contract. TREASURY_ROLE and SENTINEL_ROLE on StakingManager currently have zero members, so rescueToken has no holder today, though the Safe can grant it.
• Contracts: kHYPE 0xfD739d4e423301CE9385c1fb8850539D657C296D, StakingManager 0x393D0B87Ed38fc779FD9611144aE649BA6082109, ValidatorManager 0x4b797A93DfC3D18Cf98B7322a2b142FA8007508f, StakingAccountant 0x9209648Ec9D448EF57116B73A2f081835643dc7A, OracleManager 0x192826e470bd65FDC2CB472eDd834D096233049b, InstantUnstakePool 0x665b67793594fc5c251a3c95cbeb4b6245cd2123, live PauserRegistry 0x752e76ea71960da08644614e626c9f9ff5a50547, live oracle adapter 0x26fe3019b55f65f454c6b5f8baeea1ed405544c9, ValidatorSanityChecker 0x8b641d61f1e8811038bc99ada1685b261da2797d, Operator 0x23A4604cDFe8e9e2e9Cf7C10D7492B0F3f4B4038, treasury 0x64bd77698ab7c3fd0a1f54497b228ed7a02098e3 (confirmed by StakingManager.treasury()). Two documentation divergences on the kHYPE table of the contracts page: the PauserRegistry row displays 0xac03CABA51e17c86c921E1f6CBFBdC91F8BB2E6b, which has zero code, zero balance and nonce 0, while the same row's hyperlink points correctly to 0x752E76ea...; and DefaultOracle / OracleAdapter are both listed as 0xefbcCc6E33DA1C1ef638cBc0F044968D0f590fED, which returns false from both isAuthorizedOracle and isActiveOracle on the live OracleManager.
• Parameters: withdrawalDelay() 604800 (7 days), unstakeFeeRate() 0 against BASIS_POINTS() 10000 and a source-level cap of require(newRate <= 1000) so 10% maximum, minStakeAmount() 5 HYPE, stakingLimit() 0, stakingPaused() 0, withdrawalPaused() 0, totalQueuedWithdrawals() 203,730.60 kHYPE, instantUnstakeFeeRate() 10 bp on the InstantUnstakePool. Documented fee: "There is a 10% fee on staking rewards, of which 70% is used to purchase KNTQ on the open market and 30% is sent to the Kinetiq treasury for operations."
• Withdrawal paths: Kinetiq docs, "~8-9 days total: includes 1-day delegation lockup + 7-day unstaking queue", via queue, wait, confirm, plus a documented 24-hour lock immediately after staking. Corroborated by Hyperliquid docs: "Delegations to a particular validator have a lockup duration of 1 day" and "transfers from staking account to spot account have a 7 day unstaking queue". Instant path: InstantUnstakePool holding 85,967.32 HYPE, about $4.91M at the live mark, fee 0.10%. Live at the L1 for the StakingManager: 13,817,998.57 HYPE delegated, 62,372.68 pending withdrawal, 4 pending withdrawals.
• Validator set (Hyperliquid API, 2026-08-05): 34 validators total, 27 active, 7 inactive, 435,722,050 HYPE of active stake. Kinetiq delegates to 14 of them, 13,817,998.57 HYPE, 3.17% of active stake. HyperStake 4,700,248 (34.02%), Kinetiq x Hyperion 3,454,495 (25.00%), Nansen x HypurrCollective 583,110 (4.22%), Purrposeful x HyBridge x PiP 481,795 (3.49%), Bitwise Onchain Solutions x FalconX 465,114 (3.37%), infinitefield.xyz 460,694 (3.33%), then Anchorage By Figment, Liquid Spirit x Hydromancer x Rekt Gang, CMI, Hyperdash, USDT0 x Luganodes, HypurrCorea Spacebar x DeSpread, ASXN and B-Harvest at about 3.32% each. Top two 59.02%, top three 63.24%, top five 70.09%. Kinetiq supplies about 47% of HyperStake's total stake and about 54% of Kinetiq x Hyperion's. Zero delegation to any of the five Hyper Foundation validators, which hold 212,929,945 HYPE, 48.87% of active stake. Stake-weighted commission 2.31%, highest is Anchorage By Figment at 10%. None jailed, none inactive.
• Slashing and the rate: reportSlashingEvent is onlyRole(ORACLE_MANAGER_ROLE) on ValidatorManager and totalSlashing enters the exchange-rate numerator negatively. The reachable caller path is OracleManager.generatePerformance(address), onlyRole(OPERATOR_ROLE), whose magnitude derives from authorized adapters (exactly one authorized and active) and is filtered by the live sanity checker. setMaxPerformanceBound is onlyRole(OPERATOR_ROLE); setSanityChecker is onlyRole(DEFAULT_ADMIN_ROLE) and the code skips the check when it is address zero. Hyperliquid docs: "There is currently no automatic slashing implemented"; "A jailed validator does not produce rewards for its delegators"; validator commission "cannot be increased unless the new commission is less than or equal to 1%". setDelegation, rebalanceWithdrawal, activateValidator and deactivateValidator on ValidatorManager are all onlyRole(MANAGER_ROLE). No per-validator allocation cap in the contract.
• Our exposure (Pendle, chain 999): market 0xb48b0c95b2ddc464484305b7363fad5bd5b7a683, PT-kHYPE-24SEP2026 0x50fc4edc6346f36993bb30fe60e932504ed17391, YT 0xe45162d03e0475dab556ef1cccefeea2ec27ade2, SY-kHYPE 0x57fc55dff8ceca86ee94a6bf255af2f0ed90eb9e, maturity 2026-09-24, protocol field "Kinetiq" (confirmed). PT $56.9943, kHYPE $58.4294, SY $58.4578. Market liquidity $3,210,116, total TVL $17,185,568, 24h volume $84,544. impliedApy 2.2745%, underlyingApy 1.9821%, underlyingRewardApy 0, ptDiscount 0.3035%. Our 69.275 PT is about $3,950, roughly 0.12% of pool liquidity, 50 days to maturity. Pendle labels the instrument "PT kHYPE (HYPE)" and the market's accountingAsset is native HYPE.
• Secondary liquidity: four meaningful kHYPE/WHYPE pools total about $7.42M, nest $4,175,724, prjx $1,517,284, ramses $1,329,864, hyperswap $397,725, against $814.3M of kHYPE. About $7.52M across all fifteen kHYPE/WHYPE pairs. Combined 24h volume roughly $1.38M, 98% of it on ramses. Adding the $4.91M instant pool gives about $12.3M of same-day capacity, roughly 1.5% of supply.
• Holder concentration: kHYPE supply 13,723,433. Sampled holders: Pendle SY-kHYPE 314,909.92 (2.29%), StakingManager escrow 199,640.97 (1.45%), Safe 0, treasury 84.56. Wallet-level top-holder ranking: refresh.
• Name collision: kHYPE (Kinetiq, 0xfD739d4e...) is a different asset from stHYPE (StakedHYPE / Thunderhead), listed as staked-hype (STHYPE) at $57.05 on 2026-08-05. Chain 999 also carries live Pendle markets for stHYPE (0xcc681585a2763fb1624d42135509fc5427a92d49, maturing 2026-08-27), vkHYPE (0x2257e604c7cbfe6cebeb0d196990bdf430f6676b, maturing 2026-09-24, the same date as ours) and haHYPE (0x5fae3ef835d47ab73dde5b9c6b01cc0bc54e9199, 2027-01-28). No figure in this report is sourced from any of them.
• External cross-reference (not adopted): Yearn Curation rates Kinetiq kHYPE 2.3 of 5.0, band "Low Risk", report dated 2026-05-19 (curation.yearn.fi/report/kinetiq-khype, verified this run). Their weighted breakdown: Audits and Historical 20% / 2.00, Centralization and Control 30% / 3.30, Funds Management 30% / 1.75, Liquidity Risk 15% / 2.00, Operational Risk 5% / 2.00. We land one notch more conservative at caution, on the same dimension their own scorecard rates worst.
• Recent-news scan (2026-07-01 to 2026-08-05): nothing material found. No exploit, incident, pause, governance action, slashing or depeg in the window. Kinetiq appears in our Hyperliquid refresh as a HIP-3 deployer under "Markets by Kinetiq", two dexes, each under $10M of open interest and effectively dormant. That is the kmHYPE product line, not the kHYPE LST.

First report on this protocol. Maintained monthly. Methodology: DeFi Research Instruction v2.