Hyperliquid HIP-3 and HIP-4 Explained: Perps to Predictions

Two protocol upgrades turned Hyperliquid from a crypto perpetuals exchange into something closer to an operating system for markets. HIP-3 lets anyone with enough staked HYPE launch a perpetuals exchange for stocks, oil, or gold. HIP-4 adds prediction markets that settle without a token vote. Here is how both work, what they have built so far, and where the risks sit.

Hyperliquid spent its first two years being described as the fastest decentralized perpetuals exchange in crypto. The description was accurate and incomplete. Since late 2025, the network has been executing a more ambitious plan: turning its core trading infrastructure into a platform that other builders deploy markets on top of, the way developers deploy apps on cloud infrastructure. Grayscale Research made the comparison explicit in a June 2026 note, writing that Hyperliquid now looks less like a stock exchange and more like Amazon Web Services.

Two upgrades carry that transformation. HIP-3, live on mainnet since October 13, 2025, opened perpetual futures listing to outside builders and brought tokenized stocks, commodities, and indices onto the platform at scale. HIP-4, live since May 2, 2026, added a second market primitive built for prediction markets and other event contracts. Together they explain why seven of the top ten markets by volume on a crypto exchange are now things like Nvidia stock and gold, and why the platform is picking a direct fight with Polymarket and Kalshi.

This guide walks through what each proposal does, how the mechanics work, what has happened since launch, and what can still go wrong.

First, the basics: what a HIP is

Hyperliquid is a layer 1 blockchain built around a fully on-chain central limit order book. Its core engine, HyperCore, processes around 200,000 orders per second and handles matching, margining, and liquidations for every market on the chain. A separate component, the HyperEVM, runs Ethereum-style smart contracts on the same consensus layer. The native token, HYPE, secures the network through staking, pays fees, and absorbs most protocol revenue through a continuous buyback program. Cumulative protocol revenue passed $1 billion in late June 2026, with an annualized run rate near $840 million.

Changes to the protocol arrive through Hyperliquid Improvement Proposals, or HIPs, which the community debates and HYPE stakers weigh in on before the core contributors ship the code. The first two set the pattern. HIP-1 created the standard for launching spot tokens, with ticker slots sold through recurring Dutch auctions, so listing a token became a market process instead of an application form. HIP-2 added a protocol-native liquidity mechanism that seeds order books for new tokens automatically, solving the empty-book problem that kills most new listings on other venues. Both dealt with spot markets, and both introduced ideas that return later: auctions as the allocation mechanism for scarce listing slots, and protocol-level guarantees standing behind builder-created markets. The third and fourth proposals took those ideas after the two bigger prizes: perpetual futures on everything, and event contracts on anything.

HIP-3: builder-deployed perpetuals

Before HIP-3, listing a new perpetual market on Hyperliquid worked the way it works on most exchanges: the core team decided. That created a bottleneck and a gatekeeper, two things the platform’s own community had complained about as the asset universe stayed narrow while demand for stock and commodity exposure grew.

HIP-3, called Builder-Deployed Perpetuals, removed the gatekeeper. Since October 2025, any builder who stakes 500,000 HYPE can deploy an independent perpetuals exchange on HyperCore, without core team approval. At current prices near $64, that stake represents roughly $32 million, a number that matters for reasons covered below.

The deployer controls nearly everything about their market. They choose the assets, the oracle that sets the mark price, the collateral token, margin requirements, leverage limits, funding parameters, and the front-end experience. The first three assets in any HIP-3 exchange deploy without an auction. Additional assets go through a Dutch auction shared across all HIP-3 deployers, similar to the HIP-1 ticker auctions.

What the deployer does not control is the plumbing. HIP-3 markets inherit the full HyperCore stack: the same matching engine, the same order types, the same margining and liquidation logic, and the same solvency guarantees as the validator-operated markets. A trader interacting with a builder-deployed market gets the same execution quality as on the flagship crypto perps.

The economic design has three pillars:

• The stake is a bond, not just a ticket. The 500,000 HYPE can be slashed if the deployer misbehaves, for example by manipulating an oracle or breaking market rules, and the requirement holds for 30 days even after a deployer halts all markets.
• Fees split down the middle. HIP-3 markets charge users twice the fee of validator-operated perps, and the deployer keeps 50%. The protocol collects the same revenue per trade either way, so builder markets grow the pie without cannibalizing it.
• Cross margin has eligibility standards. Validators only allow cross margin on HIP-3 assets with sufficient observable liquidity, a reliable external oracle, and resistance to price manipulation, and any 50% intraday move in the reference price triggers a review.

The design goal is alignment: builders with $32 million at stake and a 50% revenue share have every reason to run clean, liquid, well-oracled markets, and a slashing mechanism waits for the ones who do not.

What HIP-3 actually built

The proposal would be a footnote if nobody used it. The opposite happened. The first market, a synthetic Nasdaq-style index called XYZ100, went live within days of activation. Its deployer, TradeXYZ, then built out United States equities including Nvidia, Tesla, Google, and Amazon, plus gold and silver contracts benchmarked to COMEX front-month futures, and later secured official licensing rights to the S&P 500 ticker, a landmark moment for a DeFi protocol.

The numbers followed. Open interest across HIP-3 markets passed $1.43 billion within months of launch. By spring 2026, seven of Hyperliquid’s top ten markets by volume were tokenized equities or commodities, not crypto pairs. During the West Asia crisis earlier this year, when traditional commodity venues closed for the weekend, traders moved to Hyperliquid to trade oil, gold, and silver around the clock, and HIP-3 markets drove up to 40% of the platform’s total volume. Non-crypto assets showed 60% trader retention in late March, a signal that around-the-clock access to traditional markets is a durable product, not a novelty. At peak HIP-3 activity the platform generated $2.3 million in daily fees, funding $11 million in HYPE buybacks.

Other deployers took different angles. Kinetiq built around its liquid staking token. Liminal used HIP-3 markets to run fully on-chain delta-neutral yield strategies across equities, FX, and commodities, including markets collateralized with yield-bearing assets like Ethena’s USDe. In June, Hyperliquid and TradeXYZ launched the FOMO app, a single interface for trading equities, pre-IPO stocks, crypto, indices, and commodities. Access also spread through consumer wallets: HIP-3 markets can be traded through any Hyperliquid-compatible front end, including Phantom.

The listing economics also flipped in a way worth pausing on. Under the old model, and on centralized exchanges generally, a new asset waits for an exchange’s business development calendar, and projects have long complained about the cost and opacity of the process. Under HIP-3, listing latency collapsed from a governance or negotiation timeline to a deployment transaction plus an auction, and the gatekeeping moved from relationships to capital. A pre-launch project that wants a perpetual market for hedging no longer needs a major venue’s blessing; it needs a deployer willing to run the market. Comparable systems show how unusual this is: dYdX v4 still routes every new market through a governance vote with a week or two of latency, and GMX listings run through its core team. Hyperliquid is the first chain-level implementation where market creation itself carries no approval step.

The concentration is the caveat. TradeXYZ accounts for more than 90% of all HIP-3 open interest, and Blockworks Research has flagged the deployer economics as a structural risk: with a roughly $30 million lockup, auction costs, and stiff competition, a smaller deployer’s break-even period can stretch to four years. Blockworks has proposed lowering the stake for small builders and letting them keep 100% of revenue until they recover their costs. Hyperliquid’s own documentation says the 500,000 HYPE threshold is expected to fall as the infrastructure matures. Until it does, HIP-3 is permissionless in principle and an oligopoly in practice.

HIP-4: outcome markets

HIP-3 covered continuous markets, things with a price that moves all day. It could not cleanly handle discrete events. A perpetual future needs an oracle that updates continuously with limits of roughly 1% deviation per update, a design suited to leveraged trading on a live price and incompatible with questions that jump from uncertainty to a hard answer in one instant, like an election call or an inflation print.

HIP-4, announced on February 2, 2026 and live on mainnet since May 2, added a purpose-built primitive for exactly that. Outcome markets are fully collateralized contracts that settle to exactly 0 or 1 at expiry. Each market has two sides, typically Yes and No, and the order books for the two sides are merged: an order to buy Yes at a price of 0.62 is the same order as one to sell No at 0.38, so all liquidity concentrates in one book. Positions are collateralized in USDH, the network’s native stablecoin, and because every position is fully backed, there is no liquidation risk.

The market lifecycle has a distinctive opening. Each new outcome market starts with a single-price clearing auction lasting around 15 minutes, during which traders submit limit orders but nothing executes. The auction clears at the price that matches the most volume, and unfilled orders roll into continuous trading on the standard order book. The mechanism exists to concentrate early liquidity and produce a fair opening price instead of a thin, gappy first print. It borrows a page from how traditional exchanges open trading each morning, which is fitting for a protocol that keeps hiring ideas from the market structure it wants to replace.

The architecture runs natively inside HyperCore, sharing the matching engine, order types, and throughput of every other market on the chain. That matters for one under-discussed reason: liquidity providers can quote prediction markets with the same tooling and speed they use on perps, instead of the bespoke market-making setups that thinner prediction venues require. Deep books were always the missing ingredient on long-tail event markets, and Hyperliquid’s bet is that professional liquidity follows familiar infrastructure.

The fee structure is openly aggressive. Opening or minting an outcome position costs nothing. Fees apply only on closing, burning, or settling, and makers pay zero. That pricing targets Polymarket and Kalshi, which processed a combined $44.8 billion in June on the back of the World Cup, and the community reaction at announcement made the intent plain. When the proposal dropped in February, crypto.news covered the market pricing in exactly that ambition, with traders framing HIP-4 as Hyperliquid trying to house all of finance.

Initial markets are curated and validator-deployed, starting with recurring daily Bitcoin price threshold contracts that reset each day, run by the prediction platform Outcomexyz. Planned categories include politics, sports, macro data releases, crypto events, and entertainment. A later phase opens permissionless deployment: builders will stake 1,000,000 HYPE per market slot, slashable and burned if validators find oracle manipulation, invalid state transitions, or prolonged downtime. One slot supports rolling and recurring markets, recycling after each settlement.

Settlement without a token vote

The deepest difference between HIP-4 and the incumbent on-chain prediction markets is not fees. It is how truth gets decided.

Polymarket outsources contested resolutions to UMA’s optimistic oracle, where token holders vote on disputed outcomes, an architecture that has produced repeated controversies in 2026, including a $60 million market on a Strategy Bitcoin sale that resolved against the documented facts. The full mechanics and failure modes of that system are covered in our companion guide to how prediction markets resolve.

HIP-4 replaces the token vote with the chain itself. Settlement runs through Hyperliquid’s validator set executing automated resolution against pre-specified, objective data sources. There is no dispute window, no escalation, and no path for a token holder with a position in the market to also vote on its outcome. The trade-off is scope: deterministic settlement works for objective questions with a clean data source, which is why the first markets are price thresholds. Ambiguous questions, the kind that generate the worst oracle disputes elsewhere, are exactly the kind HIP-4’s design avoids listing.

What all of this looks like from the trader’s side

For a user, the machinery above mostly disappears. HIP-3 markets sit in the same interface as the flagship crypto perps, trade through the same API, and settle against the same margin account. A trader shorting gold on a builder-deployed market places the order the same way they would short Ethereum, and the differences show up in three places worth knowing.

Fees are higher on builder markets. The headline rate on a HIP-3 perp is twice the validator-operated rate, which at base tiers works out to roughly 3 and 9 basis points for makers and takers before discounts, with the deployer keeping half. Staking discounts, referral rebates, and collateral-based reductions still apply on top, so an active HYPE staker narrows the gap considerably.

Oracle quality varies by deployer. On validator-operated markets, the network itself maintains the price feed. On a HIP-3 market, the deployer chooses and operates the oracle, which is why the mark price on a weekend oil contract can drift from where Monday’s COMEX open eventually prints. During the West Asia crisis, Hyperliquid’s oil market traded on its own oracle through days when no traditional reference price existed at all. That independence is the product and the risk in one feature.

Collateral differs by market. Most markets margin in stablecoins, but HIP-3 supports alternative collateral where the deployer enables it, including yield-bearing assets, and HIP-4 outcome positions collateralize in USDH. Settlement demand for outcome markets flows through the stablecoin into the same fee-and-buyback loop that already routes nearly all protocol revenue toward HYPE, which is why analysts treat HIP-4 volume as a direct token catalyst rather than a side business.

The practical entry points have multiplied too. Beyond the native app, HIP-3 and HIP-4 markets surface through Phantom, through the FOMO app for the equities lineup, and through any front end built on the public API, since every builder market shares the unified HyperCore order flow.

The risk column

Every part of the story above has a counterweight, and an honest explainer lists them.

Deployer concentration is the loudest one. A permissionless system where one builder holds 90% of open interest has recreated a gatekeeper one level up, and the $32 million entry stake keeps it that way for now. Regulatory exposure is the second. Hyperliquid operates without KYC in most of the world, the United Kingdom’s FCA has declared the platform unauthorized, and pending United States market structure legislation could either validate or constrain synthetic stock perpetuals, a product category regulators have barely begun to examine. Institutional ceilings are the third: a June JPMorgan report saw limited institutional demand for perpetual futures generally, citing unbounded basis risk and missing clearing protections, which matters for a token whose valuation leans on volume growth. And the products themselves are dangerous instruments. Leveraged perpetuals on any underlying can liquidate a position in minutes, and cross margin across markets adds its own failure modes.

There is a subtler risk in the oracle layer that the slashing design only partially covers. A deployer’s oracle is a single point of interpretation for its markets, and unusual conditions expose the gap: when traditional venues close and a HIP-3 commodity market keeps trading, the mark price is whatever the deployer’s methodology says it is, with no external reference to check against until markets reopen. Validators review any 50% intraday reference move and slashing punishes proven manipulation, but a subtly mispriced weekend, honest or otherwise, transfers money between longs and shorts without tripping any threshold. Traders in builder markets are underwriting oracle methodology whether they think about it or not.

None of that has slowed the platform yet. Hyperliquid controls an estimated 70% of on-chain perpetuals volume, spot HYPE ETFs drew $111 million in inflows in late June while Bitcoin and Ethereum funds bled, and the ecosystem is spending on the long game, including a $29 million policy center in Washington. Whether the moat holds is a different question from whether it exists.

The bigger picture for L1 competition

HIP-3 and HIP-4 also reframe what layer 1 blockchains compete on. Ethereum and Solana fight over DeFi liquidity, users, and fees, a race with its own 2026 scoreboard. Hyperliquid opted out of the general-purpose contest and vertically integrated one thing: markets. The bet is that an exchange-shaped blockchain with permissionless market creation captures more value than a general-purpose chain hosting exchange apps. dYdX tried a dedicated appchain with governance-gated listings. GMX built on someone else’s layer 2. Hyperliquid is the first to make market creation itself permissionless at the chain layer, and the early evidence, an order of magnitude expansion in what can be traded on-chain, suggests the design space was bigger than the industry assumed.

What to watch from here

Three markers will tell the story over the next year. First, whether the HIP-3 stake requirement drops and the deployer set widens beyond one dominant builder. Second, whether HIP-4 volume becomes measurable against Polymarket and Kalshi once permissionless deployment opens and categories expand past crypto prices. Third, whether regulators treat builder-deployed stock perpetuals as an innovation to license or a loophole to close. The upgrades themselves are shipped and working. The open question, as always in this industry, is what survives contact with scale.

Frequently asked questions

Disclaimer: This article is for informational purposes only and does not constitute investment advice. Digital asset markets are volatile and you can lose your entire investment. Always do your own research. Information current as of July 3, 2026.