HIP-3: Builder-Deployed Perpetual Markets and RWA Outlook

Hyperliquid quietly changed who is allowed to create a market. For a serious trader the question is no longer what does the exchange list? but who deployed this market, what oracle did they pick, and what happens to me if they get it wrong?

For its first few years, trading a perpetual on Hyperliquid meant trading a market the protocol itself had curated, much like Binance or Bybit, where a listing committee vets spot liquidity before a contract goes live. HIP-3 adds a permissionless route alongside those validator-operated markets. It lets any qualifying builder deploy a perpetual market directly onto Hyperliquid's core engine, choosing the asset, the oracle, and the contract specifications. Hyperliquid still handles execution, margining, and settlement; the builder defines and operates the market.

That change is why, in 2026, Hyperliquid is a venue where you can trade perpetual derivatives referencing equities, commodities, and FX alongside crypto, without owning the underlying asset. It is also why open interest in those real-world-asset (RWA) markets has become one of the most-watched numbers in the ecosystem. And it is why your risk is no longer just directional. This note explains the mechanism precisely, then maps the microstructure traps that come with it.

Developments since June 2026 and current specifications

• Fees are documented as a deployer setting. As of October 9, 2026, Hyperliquid's fees documentation describes configurable deployer fees: an additional fee share of 0–300%, with the deployer keeping up to 50% of fees, plus an optional growth mode that applies a 0.1 fee multiplier, with a 30-day cooldown per asset. When this model was introduced is not established by these sources.

• Deployers consolidated. Ventuals' sunset guide scheduled trading halts for OPENAI and ANTHROPIC on June 15, 2026 and for its commodity and index markets on June 18, with automatic settlement upon each halt; the guide alone does not verify completion. Felix planned to stop trading on its HIP-3 DEX on June 19 and complete delisting on June 20, according to an advance report by PANews (June 16, 2026). Dreamcash scheduled its CASH markets for settlement from June 30 to July 2 while keeping its app running, according to Phemex News (June 26, 2026, citing PANews).

• An oracle incident hit a live RWA market. On July 28, 2026, an abnormal pre-market print on a Korean venue fed into the oracle of trade.xyz's SK Hynix perp, and about $57.4 million of long position notional was liquidated across 960 accounts, according to BeInCrypto via Yahoo Finance (July 28, 2026). Details below.

• Permissioned HIP-3 is on testnet. Hyperliquid's HIP-3 deployer actions documentation describes a testnet-only “HIP-3*” variant with an optional allow-list and proxied user actions. It was reported on September 3, 2026 by crypto.news, which said existing markets are unchanged and no mainnet date had been announced.

• Open interest kept climbing. HIP-3 open interest was $4.03 billion on August 8, 2026, according to Cryptopolitan (August 10, 2026). CryptoDaily (September 9, 2026, citing The Block) reported more than $4.44 billion and over 34% of Hyperliquid's total open interest for August 2026. Both are third-party historical estimates.

From a listing committee to a deployment framework

A centralized exchange treats a new derivative as a curation problem. Before a contract launches, a committee checks that the underlying has deep, multi-venue spot liquidity so the index price is hard to manipulate. That gatekeeping is slow, but it concentrates risk-management expertise in one place.

HIP-3 inverts the arrangement. Per the official documentation, the deployer of a perp market is responsible for market definition (the oracle definition and contract specifications) and market operation (setting oracle prices, leverage limits, and settling the market if needed). The market runs on the same HyperCore stack as every other Hyperliquid perp, so it shares the same execution infrastructure, margining engine, and unified API. That does not make markets equivalent: liquidity, spreads, and execution quality are specific to each market. Nor is the curation inherited. The builder makes the judgment calls a listing committee used to make, and the consequences land on the people who trade the market.

In practice, HIP-3 has so far concentrated around one deployer rather than spreading across many. trade.xyz (DEX name xyz) held about 97.6% of HIP-3 open interest according to PANews (June 16, 2026), and about $4.01 billion of $4.03 billion on August 8, 2026, according to Cryptopolitan (August 10, 2026). Other launches have come and gone. Dreamcash's USDT0-collateralized CASH markets launched in February 2026 in collaboration with Selini Capital as liquidity provider and Tether as a strategic investor, according to The Block (February 2026). Dreamcash held under 1% of open interest according to PANews (June 16, 2026), and its CASH markets were scheduled for settlement from June 30 to July 2, 2026 while the app continues, according to Phemex News (June 26, 2026). Market-structure figures move quickly, so treat them as context, not constants.

The mechanism that actually matters

If you only remember the marketing line, permissionless perps, you will misprice the risk. The specifics determine whether a market is safe to size into. These are from Hyperliquid's official HIP-3, fees and deployer actions documentation, as of October 9, 2026. Where those pages disagree, the table says so.

Two implications fall straight out. First, the economics favor the builder: keeping up to half the fees on a market you control is a real revenue line, which is why sophisticated teams deploy RWA venues. But the fee level is now a deployer choice, and growth mode lets a deployer trade near-zero fees for volume, so check the fee setting market by market instead of assuming one rate. Second, the 500k HYPE stake is the protocol's alignment lever, and it is backed by a slashing regime that makes deployer competence your problem too.

Where builder markets break: the oracle

On Hyperliquid's validator-operated perps, the headline safety design is that the oracle is built from external spot feeds rather than Hyperliquid's own order book, which neutralizes the classic push-the-perp-book attack. On a HIP-3 market that protection is only as good as the deployer's choices, because the deployer sets the oracle and mark inputs. The official spec allows any oracle, but warns that perps make the most sense when the underlying is well-defined and difficult to manipulate, notes that most price indices are poor perp oracle sources, and reminds deployers that slashing applies to every market on their DEX.

That caveat exists because HIP-3 deliberately opens the door to assets with thin or non-existent CEX spot markets. When there is no deep external reference, the deployer must construct one, and the cost to move a shallow reference can be lower than the open interest riding on the perp. Two failure shapes follow:

• Manipulation premium. If a market's index leans on a few shallow venues, price can be pushed to force liquidations. A time-weighted feed blunts instant flash-loan moves but not sustained pressure. Before you trust a market, look at what the oracle is made of, not just where the price is.

• Stale-feed disconnects. If the external source stops updating, the reference goes stale while the contract keeps trading. A stale reference does not necessarily freeze the mark: the deployer actions documentation says a mark price with no updates for 10 seconds falls back to the local mark price, and it tells deployers not to rely on that fallback. How a given deployer prices through a stale feed has to be checked separately. When the reference catches up, the correction can be abrupt, and on an hourly funding cadence a persistent premium can be expensive well before the snap-back.

This is no longer hypothetical. On July 28, 2026, an abnormal pre-market print on NXT, a thinly traded South Korean venue, became the reference for trade.xyz's xyz:SKHYNIX market; the perp fell 17.9% within its discovery bounds and about $57.4 million of long position notional was liquidated across 960 accounts, according to BeInCrypto via Yahoo Finance (July 28, 2026). Hyperliquid said the market is deployed and operated by trade.xyz, not by Hyperliquid. trade.xyz said its oracle worked as intended and pledged to reimburse affected traders, according to Cryptopolitan as republished by Mitrade (July 29, 2026). That figure is liquidated position notional, not trader losses. No slashing vote was reported. The documented 50% start-of-day move rule is a cross-margin eligibility and validator-review trigger; it is not a general slashing exemption for smaller moves. Any compensation came from the deployer's own decision, not from the protocol.

RWA perpetuals: the calendar is a microstructure risk

The most distinctive HIP-3 use case is the RWA perpetual: equities, commodities, FX, and similar. These collide with a mismatch that does not exist for crypto, namely that traditional markets close, and the perp does not.

Hyperliquid settles funding on an hourly cadence from a one-hour premium average. When an RWA's underlying spot market closes on Friday, its reference can stop updating for the weekend while the perpetual keeps trading around the clock. If demand bids the perp above the static reference, the premium, and the funding paid by longs to shorts, can accumulate hour after hour. Collecting that funding looks attractive until Monday, when the traditional market reopens and the reference can re-price in a single gap. A short that harvested weekend funding now faces that gap with little room to react. The same logic applies at a daily open: the SK Hynix episode began at a session open, not a weekend.

Liquidity, liquidations, and auto-deleveraging

A perp with no liquidity is untradeable at size. A centralized exchange solves this by paying designated market makers; HIP-3 has no such guarantee. Depth on a builder market depends on whichever market makers the deployer attracts or incentivizes, and those providers allocate capital on risk-adjusted terms. If a market shows toxic or one-sided flow, they can widen quotes or step back, exactly when you would want them present.

Per the official liquidations documentation, a liquidation is first attempted by sending market orders to the order book. On HIP-3 assets with cross margin enabled, eligible positions can then go to the DEX's onchain backstop liquidator (HIP-3 docs). Auto-deleveraging (ADL) triggers when an account's value or an isolated position's value turns negative. Selected profitable positions on the opposite side are then closed at the previous mark price, ranked by (mark_price / entry_price) × (notional_position / account_value), so the most profitable and most leveraged positions go first. The uncomfortable corollary for a successful strategy in a long-tail market is that being right makes you a candidate for ADL. Lower effective leverage may reduce your ADL exposure, but it does not eliminate it. Withdrawing collateral while keeping the position raises leverage and can worsen your ranking. Watch order-book depth and open interest, and do not assume you can exit on demand.

Slashing and cross-margin: read these before you size

Slashing falls on the deployer, but shapes your market

To keep markets honest, a deployer's 500k HYPE is slashable by stake-weighted validator vote, and remains slashable through the 7-day unstaking queue. The documentation is explicit that slashing does not distinguish between malice, incompetence, and compromised keys; what matters is the effect on the protocol, and attempted malicious inputs are slashable even if they cause no harm. Slashed stake is burned, not distributed to affected users. Indicative ceilings: up to 100% for inputs causing invalid state transitions or prolonged downtime, up to 50% for brief downtime, up to 20% for network degradation or performance issues. The practical signal: a deployer's stake at risk is a rough proxy for how carefully a market is likely to be run, and burned stake means you should never assume a slashing event will compensate you.

Cross-margin is irreversible and gated

Enabling cross-margin on a HIP-3 asset is, per the docs, irreversible. Validators only permit it for assets that meet eligibility standards: sufficient observable liquidity, a reliable external oracle, and manipulation resistance. And they review for slashing whenever an asset's external price (externalPerpPx) moves more than 50% relative to the start-of-day price. Assets where 50% daily moves are expected more than once a month are ineligible for cross-margin and subject to slashing. For you, the takeaway is concrete: know whether the HIP-3 market you are trading is cross or isolated. If it shares your cross-margin pool, a dislocation in one long-tail asset can reach the rest of your account.

Strategic implications

HIP-3 is best read as a deliberate transfer of responsibility. The protocol keeps what it is uniquely good at, fast and solvent execution and settlement, and pushes market design and curation out to a competitive layer of builders. That is what makes the RWA expansion possible at speed. It also moves where a trader's edge comes from: here it is operational diligence, not just a price view. Before you size, audit five things: the oracle, the deployer, the margin mode, the liquidity, and the fee setting. Treat an unfamiliar builder market as guilty until proven liquid. For builders, the calculus is the mirror image. Keeping up to half the fees on a market you control can be a genuine business, and the June 2026 exits show how quickly the field has consolidated around one leader. You pay for it with a 500k HYPE stake, held for at least 183 days and exposed to a slashing regime that does not care whether a failure was malicious or merely careless.

How to monitor

• Open interest by HIP-3 DEX and asset — concentration and growth; verify the current figure before citing it.

• Oracle composition — the number and depth of venues behind a market's index, and how the deployer handles session opens.

• Funding versus the underlying — especially the weekend premium on RWA pairs.

• Book depth and ADL exposure — programmatically, via the API, rather than a UI warning light.

• Fee setting and growth mode — per market, since rates now differ by deployer and asset.

• Deployer stake and margin mode — stake maintained, and whether the market is cross or isolated.

Sources

• Hyperliquid Docs — HIP-3: Builder-deployed perpetuals (official)

• Hyperliquid Docs — Fees (official)

• Hyperliquid Docs — HIP-3 deployer actions (official)

• Hyperliquid Docs — Liquidations (official)

• Hyperliquid Docs — Auto-deleveraging (official)

• trade[XYZ] Docs — Fees (deployer documentation)

• Ventuals Docs — Sunset guide (deployer documentation)

• PANews — HIP-3 consolidation and planned Felix exit, June 16, 2026 (third-party)

• Phemex News — Dreamcash to close CASH markets, June 26, 2026 (third-party)

• crypto.news — Hyperliquid expands HIP-3 with permissioned markets, September 3, 2026 (third-party)

• BeInCrypto via Yahoo Finance — SK Hynix perp liquidations, July 28, 2026 (third-party)

• Cryptopolitan via Mitrade — trade.xyz reimbursement pledge, July 29, 2026 (third-party)

• Cryptopolitan — HIP-3 open interest above $4 billion, August 10, 2026 (third-party)

• CryptoDaily — Hyperliquid open interest at $14.3 billion, September 9, 2026 (third-party)

• The Block — Tether invests in Dreamcash, February 2026 (third-party)

Figures as of October 9, 2026 unless dated otherwise. This article is educational information, not financial advice.