Inside Hyperliquid: How One Perp DEX Took Over On-Chain Futures

On September 23, 2026, the token that powers Hyperliquid touched an all-time high just under $98, capping a run that has made HYPE the eleventh-largest cryptocurrency by market value, worth about $20 billion, according to CoinGecko. A week earlier, the parent company of Kraken said it intends to route US-regulated perpetual futures onto Hyperliquid for American customers. And back in May, the chief executive of Intercontinental Exchange, the company that owns the New York Stock Exchange, told a Bernstein audience that Hyperliquid was, in his words, bigger than NASDAQ, and then marveled that it was run by roughly a dozen people.

Perp DEXs, decentralized venues for trading perpetual futures, went from an offshore curiosity to the busiest corner of on-chain finance in the space of about two years, and one venue took most of the traffic. This is a deep dive into that venue: what Hyperliquid is, how its custom blockchain matches and settles trades, who takes the other side of your position, how its HYPE buyback machine works, where it has broken, how decentralized it really is, and why both the US derivatives regulator and Wall Street incumbents are now circling.

From offshore experiment to the venue Wall Street wants in on

Global crypto perpetuals cleared roughly $85 trillion in notional volume in 2025, and almost none of it was reachable by a compliant US trader. On September 16, 2026, Payward, the parent of Kraken, said it would change that, and it named Hyperliquid as the first stop: it plans to bring onchain perpetual futures to US clients starting on Hyperliquid, using the protocol’s builder-deployed HIP-3 markets. Bitnomial, Payward’s CFTC-regulated exchange and clearinghouse, would create and own the markets, while NinjaTrader Clearing, a CFTC-registered futures commission merchant inside the group, would hold customer accounts, all pending regulatory approval.

The interest is not only regulatory. Jeffrey Sprecher, founder and chief executive of Intercontinental Exchange, spent part of a May 27 fireside chat with Bernstein analyst Chinedu Bolu praising the upstart. Hyperliquid, he said, is bigger than NASDAQ, before adding, in a line that has followed the company around ever since, that it is run by only a handful of very smart people. What impressed him was mundane and telling at once: Hyperliquid runs oil derivatives around the clock, including weekends, when ICE’s own markets are closed, and traders show up for that off-hours exposure. As CoinDesk reported, the bigger-than-NASDAQ line referred to trading volume, where Hyperliquid dominates more than 70% of decentralized perpetual futures, not to market cap, where HYPE still trails Nasdaq Inc. The through-line is hard to miss: the product most associated with no-KYC offshore leverage is being wired into the regulated US futures stack, and Hyperliquid is the venue everyone starts with.

What Hyperliquid is, and why this perp DEX matters

Start with the two words. A perpetual future, or perp, is a derivative that tracks an asset’s price with no expiry date, held in line by a recurring funding payment between the two sides of the market. A decentralized exchange, or DEX, lets you trade from your own wallet, with no company holding your coins. A perp DEX marries the two: leveraged, no-expiry contracts that you open and close without ever handing custody to an intermediary.

Hyperliquid launched its own blockchain in late 2024 and, by 2026, clears the majority of on-chain perpetual volume. Sprecher’s own framing put it above 70% of decentralized perpetual futures, and CoinDesk pegged the platform at roughly 9% of all global perpetual positions, with open interest reaching a 2026 high near $11 billion. One point of vocabulary matters for the rest of this article: in the United States, a perpetual future is a derivative, and derivatives are the domain of the Commodity Futures Trading Commission (CFTC), not the Securities and Exchange Commission (SEC). The SEC only enters the frame when the question shifts from the contract to the coin, that is, whether HYPE itself is a security.

The venue’s rise is unusual in one more way. Hyperliquid grew without the venture-capital backing that funded most of its rivals, and in late 2024 it seeded its community with one of the largest token airdrops in crypto history, handing HYPE to tens of thousands of early users rather than to funds. Its founder, Jeff Yan, has framed the goal as building financial infrastructure that is neutral, programmable and globally open, closer in spirit to a public utility than to a private broker. That origin story is part of why the token trades on real cash flows rather than on a promise, and part of why the decentralization questions later in this article carry a sting.

Under the hood: HyperBFT, HyperCore and HyperEVM

Most decentralized exchanges rent space on a general-purpose blockchain and live with its limits. Hyperliquid did the opposite: it built the chain around the exchange. According to a technical overview by Figment, the network runs on HyperBFT, a Byzantine-fault-tolerant consensus derived from the HotStuff family, and it produces blocks in about 70 milliseconds, which is what gives the venue its centralized-exchange feel.

The heart of the system is HyperCore, the matching engine implemented as protocol logic rather than as a smart contract sitting on top. It is a full central limit order book running natively at the base layer, capable of about 200,000 orders per second, and every order, fill and liquidation is written into the consensus record. Order cancellations are prioritized over new market orders during block construction, which gives market makers a sliver of protection when they need to pull stale quotes, a design meant to reduce toxic flow. Crucially, the validators that run the chain are also the ones that publish the oracle prices used to mark positions and trigger liquidations. On top of all this sits HyperEVM, a general-purpose Ethereum-compatible layer with a dual-block design, fast two-second blocks for latency-sensitive work and slower one-minute blocks for heavier computation, connected to HyperCore through precompiles that read and write native state. The payoff is a venue with no per-order gas auction and exchange-grade speed. The cost, as we will see, is that all of this runs on a bespoke chain guarded by a small set of validators.

Where Hyperliquid sits among perp DEXs

Hyperliquid is not the only way to build a perp DEX; it is the most successful example of one particular design. Broadly, on-chain perpetual venues fall into three families. The first is the on-chain order book, where a full central limit order book runs as protocol logic; Hyperliquid is the flagship, and dYdX, which moved to its own Cosmos-based chain, is the other well-known example. The second is the oracle-and-pool model, where there is no order book at all: traders take the other side of a shared liquidity pool at a price imported from an oracle. GMX pioneered it, and on Solana the largest example is Jupiter, whose JLP pool holds a basket of assets and pays most of its fees to depositors. The third is the hybrid, where matching happens off-chain for speed while settlement and custody stay on-chain; Aster, backed by figures from the Binance orbit, is the loudest recent entrant.

Each design trades something away. The oracle-and-pool model gives zero slippage and works well for long-tail assets, but it imports its prices, which is exactly the weakness attackers probed in GMX’s early history. The pure order-book model gives the tightest markets and true price discovery, but it demands a bespoke, high-performance chain, which is why Hyperliquid built its own and why its validator set is small. The hybrid buys centralized-exchange speed at the cost of a matching layer you have to trust. Hyperliquid’s bet was that traders would forgive a concentrated validator set if the product felt like a major centralized exchange, and, at more than 70% of decentralized perp volume, that bet has paid off so far. The rest of this article is really the fine print on that trade.

HLP: the vault that plays the house

On a centralized exchange, a company and its insurance fund are the named backstop when a trade goes wrong. On Hyperliquid, that role belongs to a smart contract and a pool of strangers called the Hyperliquidity Provider vault, or HLP. Anyone can deposit USDC into HLP; the vault then market-makes across the order book, collects funding, and, most importantly, absorbs liquidations that cannot be filled elsewhere. There is no performance fee, and depositors share the profit and loss pro rata, which is another way of saying they are the counterparty of last resort.

The vault’s own numbers tell the story of how volatile that job is. According to a CoinGecko analysis, HLP’s assets peaked near $604 million in September 2025 and had fallen to roughly $269 million by June 2026, a drop of about 55%, even as it booked cumulative profit of around $137 million since inception in May 2023. Its single best day was the market-wide flash crash of October 10, 2025, when it earned tens of millions of dollars filling liquidations that everyone else was running from. That is the deal in one line: on a calm day HLP earns a modest spread, and on a violent day it either wins big or, as one notorious episode showed, comes within minutes of a serious loss.

For a depositor, HLP is best understood as writing insurance on the whole exchange. In calm markets you collect steady premiums in the form of spreads and funding; in a crash you may be handed positions nobody else wanted, which is how the vault booked its best-ever day during the October 2025 flash crash and how, at other times, it has stared down large unrealized losses. There is no performance fee skimming your upside, but there is also no promise you come out ahead, and no operator standing behind the vault if it draws down. Anyone can join, which is the point, and anyone who does becomes, in a small way, the house. Grasping that is the difference between chasing a yield and knowingly taking on a risk.

How a trade actually works: funding, leverage and liquidation

To open a position you post USDC as margin and take a long or short with double-digit leverage. From that moment your position is marked continuously against the oracle price the validators publish. Because a perpetual has no expiry, something has to keep its price glued to the underlying market, and that something is the funding rate: an hourly payment that changes hands between longs and shorts. When the perp trades richer than the index, longs pay shorts, which nudges buyers to close; when it trades cheaper, shorts pay longs. The mechanism is elegant precisely because no central operator sets it; it falls out of the imbalance between the two sides.

If your margin ratio falls through the maintenance threshold, your position is liquidated. If the liquidation cannot be filled cleanly in the order book, the HLP vault steps in as backstop, and in the worst case the protocol resorts to auto-deleveraging, or ADL, which force-closes profitable traders on the opposite side to keep the system solvent. That last mechanism is worth internalizing before you ever click buy: a large enough move can close a winning trade you never wanted to exit. And because the price that marks you, liquidates you, and settles your funding is produced by the same validators who run the chain, oracle quality is not a footnote; it is the thing your money rests on.

A concrete example makes the moving parts visible. Say you post $1,000 of USDC and open a 10x long on Bitcoin, giving you $10,000 of exposure. If Bitcoin rises 5%, your position gains about $500, a 50% return on your margin; if it falls 5%, you lose about $500, and a move against you of roughly 10% wipes out the margin and triggers liquidation. While the trade is open, you either pay or receive funding every hour depending on which way the market is leaning, so a crowded long that you are part of slowly bleeds funding to the shorts. None of this involves a broker deciding your fate: the oracle price, the funding rate and the liquidation engine are all protocol logic, running the same way for everyone. That is the appeal, and, when an oracle wobbles or a market gets squeezed, also the danger.

HIP-3 and the pivot to real-world assets

The biggest structural change of 2026 was HIP-3, which turned Hyperliquid from a venue with a fixed menu into an open platform for launching markets. Under HIP-3, anyone who stakes 500,000 HYPE, a slashable bond worth roughly $46 million at late-September prices, can deploy their own perpetual market. As CoinGecko’s explainer lays out, the deployer runs its own oracle feed, sets leverage and margin, attracts its own market makers, and keeps a fixed 50% of the trading fees. Validators can vote to slash up to 100% of the staked HYPE, and the stake stays slashable through a seven-day unstaking window, so a deployer cannot simply walk away. The catch that every trader should note: HIP-3 markets are not backstopped by HLP, so their liquidity and risk management live and die with the deployer.

What poured through that door was real-world assets. Builder-deployed markets now cover crude oil, gold, Nvidia, Tesla, a Nasdaq-100 tracker, and even pre-IPO names, and the specialist deployer Trade.xyz dominates the segment. The scale is startling. CoinDesk reported that these RWA perps hit a record $3.6 billion in open interest and overtook Bitcoin as Hyperliquid’s largest market by that measure; in July 2026, RWA perpetual volume reached 99.2% of Bitcoin perp volume on the platform; and builder-deployed markets grew from about 2% of Hyperliquid’s perp volume at the start of 2026 to roughly half of it. The pre-IPO corner is the strangest and most novel. Tanay Ved, Senior Research Associate at Coin Metrics, defined the instrument plainly in a research note: pre-IPO perpetual futures, he wrote, are synthetic derivative contracts designed to create a continuous market for the implied valuation or share price of a private company. In other words, a way to trade the price of a company that has no public shares at all.

The mechanism has quietly turned Hyperliquid into a marketplace of marketplaces. Because deploying a market is open to anyone willing to post the bond, specialist teams have sprung up whose entire business is running markets on top of Hyperliquid, with Trade.xyz the dominant one and the eventual home of Kraken’s regulated US markets. That is a very different shape from a traditional exchange, which decides its own product line from the top down. It also concentrates a new kind of risk: the deployer, not the protocol, runs the oracle and manages the liquidity, so a badly run builder market can fail in ways the core Hyperliquid markets, backstopped by HLP, cannot. The table below sets the two side by side.

HIP-4, prediction markets and outcome trading

If HIP-3 opened the platform to real-world assets, HIP-4, which launched on May 2, 2026, opened it to real-world questions. HIP-4 introduces fully collateralized, expiry-based contracts that settle at either 0 or 1 depending on whether a specific event happens, which is the technical shape of a prediction market. The composability is the point: a trader can hold a perpetual position and hedge it with an outcome contract inside the same account, without shuffling funds between platforms. It puts Hyperliquid into direct competition with dedicated prediction venues, and it also inherits their single hardest problem, which is deciding what actually happened and who gets to say so. That resolution question, the weak point of every on-chain betting market, is one HOGE Wire has examined in detail in our look at prediction markets and the oracle problem, and it applies just as sharply here: an outcome contract is only as trustworthy as the process that settles it.

HYPE, the Assistance Fund and the buyback machine

The reason HYPE trades where it does is a buyback loop that few tokens can match. Between 97% and 99% of the fees Hyperliquid’s markets generate flow into the Assistance Fund, which then buys HYPE on the open market and takes it out of circulation. According to crypto.news, by May 2026 the fund had deployed more than $1.3 billion and held tens of millions of HYPE, and at its peak the buyback ran at close to 7% of the token’s market cap each year, several times what most large tokens return to holders. Where a company might pay a dividend, Hyperliquid does something closer to a permanent, protocol-funded share repurchase.

It is worth being precise about what that buyback is and is not. It is not a dividend, because nothing is paid directly to holders; instead, fees are used to shrink the float, which pushes value into the remaining tokens if demand holds steady. It is also not risk-free, because the fuel is trading fees, and trading fees fall when volumes fall or, as detailed below, when a growing share of them is routed to third-party deployers. In good months the effect is a powerful tailwind that few tokens can match; in bad months it simply slows, and the market notices quickly.

In September 2026 the Hyper Foundation proposed going further still, moving to permanently remove about 13% of the circulating supply, close to $1 billion in HYPE at recent prices, by locking it away in the Assistance Fund, as The Defiant reported. All of this concerns the token, and this is exactly where the SEC, rather than the CFTC, would eventually have a say: the perpetual contracts are derivatives, but any spot HYPE exchange-traded fund, and any judgment on whether the token is a security, sits with securities regulators. That distinction is central to the wider ETF story we cover in our piece on 2026 crypto ETF approvals and the options-rule test.

The revenue paradox: when RWAs eat the buyback

Here is the tension that will define Hyperliquid’s next year. The RWA boom is powered by HIP-3, and HIP-3 hands 50% of each market’s fees to its deployer. So the very thing driving record volume also diverts fees away from the buyback that supports HYPE. The math is visible in the accounts. CoinDesk reported that gross protocol revenue fell from about $357 million in the third quarter of 2025 to roughly $202 million in the second quarter of 2026, a decline of about 43%, even though trading activity was booming, precisely because builder-deployed markets pass half their fees through to their operators.

That leaves the foundation with a genuine strategic fork. It can keep courting RWA volume, accepting that a growing share of fees leaks to deployers and dilutes the buyback, or it can protect the buyback and cede the fastest-growing part of its own market. The proposed supply burn reads, in part, as an answer to the first horn of that dilemma: if the flow of fee-funded buybacks slows, a one-time reduction in supply is another lever to support the token. Whether that is prudent capital return or financial engineering is exactly the debate a HYPE holder should be having.

There is a subtler point buried in the revenue slump. A falling headline revenue number does not necessarily mean the business is shrinking; it can mean the business is changing shape, from a first-party exchange that keeps every fee into a platform that takes a smaller cut of a much larger and faster-growing base. Whether that trade is good for HYPE holders depends entirely on how big the real-world-asset opportunity turns out to be. If tokenized equities, commodities and pre-IPO names become a serious venue for global capital, half of that is worth far more than all of a crypto-only casino. If they stall, the buyback took a haircut for nothing. That uncertainty, more than any single month’s figure, is the real story behind the price.

The JELLY squeeze: the day the house intervened

The clearest window into how Hyperliquid really works under stress is the JELLY episode of March 26, 2025. A trader opened a large short on a thin token called JELLYJELLY, then pushed the token’s price up more than 400%, forcing the short into the HLP vault, where the unrealized loss ballooned past $12 million. Within minutes, Hyperliquid’s validators voted to delist the market and force-settle it at a favorable price of $0.0095, which turned HLP’s near-loss into a small gain of a few hundred thousand dollars. The attacker still managed to withdraw several million dollars before the freeze, as CoinDesk documented.

Read one way, it was a stress test passed: the system protected its depositors and moved faster than any court or committee could. Read another way, it was the warning label made visible. The same power that rescued HLP, a small set of validators rewriting a live market in real time, is the power critics point to when they ask how decentralized this venue truly is. The decision to pull a market and settle it by fiat is a governance action, and governance actions are the subject of our broader explainer on how and why exchanges delist a token. On Hyperliquid, that lever can be pulled in minutes.

How decentralized is Hyperliquid, really?

For all the on-chain settlement, the base layer is guarded by a strikingly small group. A governance audit by crypto.news counted 34 registered validators as of late August 2026, of which 27 were active, with the foundation’s own nodes holding close to half the stake and the ten largest validators controlling roughly three-quarters of the active stake. Earlier, a January 2025 operator letter had flagged that five foundation validators controlled more than 81% of staked HYPE across the set at the time. The node software is closed-source, the foundation can jail operators, and, as JELLY showed, governance can rewrite a market in minutes.

The scale gap is the part worth sitting with. Ethereum is secured by hundreds of thousands of validators and Solana by several hundred, while Hyperliquid’s active set numbers in the dozens. That is not a like-for-like comparison, since Hyperliquid’s validators do specialized, latency-critical work, but it is the reason Sprecher’s admiring line, that the venue is run by only a handful of people, cuts both ways: efficient and centralized are two descriptions of the same fact. Readers who want the contrast in concrete terms can see what a genuinely permissionless validator set demands of an individual in our guide to Ethereum solo staking in 2026. Hyperliquid is a different animal: faster, and far more concentrated.

None of this has dented usage, which is the uncomfortable part for purists. Traders have largely decided that speed, deep liquidity and a product that works matter more, day to day, than the number of independent parties who could in theory collude or be coerced. That is a defensible choice, but it is a choice, and it means the honest label for Hyperliquid is not decentralized exchange in the maximalist sense so much as a transparent, on-chain exchange with a small operator set. The JELLY episode proved that operator set can act decisively; whether it can be trusted never to act against its users is a question the design cannot fully answer.

The regulated on-ramp: Kraken, Bitnomial and the CFTC

Because a perpetual future is a derivative, the US gatekeeper is the CFTC, not the SEC. The agency opened the door in late May 2026 by approving KalshiEX’s BTCPERP, the first true perpetual on a registered US exchange. CFTC Chairman Michael Selig called it a watershed moment, noting that the agency, as CoinDesk reported, has not approved a new type of derivative in over a decade. CME Group sued in June, arguing that perpetual contracts belong under the swaps framework rather than the futures one. In September, the CFTC moved to dismiss, calling the case, in a memorable phrase, much ado about nothing, and pointing out that nothing stops CME from listing the same contracts; CME’s response was due in early October, with Judge Colleen Kollar-Kotelly presiding, and no ruling had been issued as of this writing, per CoinDesk.

Kalshi’s approval was the opening move in a broader onshore race. Coinbase secured a path to route certain perpetuals through an affiliated foreign venue under a foreign-futures framework, and other regulated firms have lined up their own filings, each picking a slightly different legal route to the same product. The common thread is that the CFTC, not the SEC, is writing the rulebook for the contracts themselves, a division of labor that could still be redrawn if Congress passes comprehensive market-structure legislation. For now the practical takeaway for a US trader is simple: a compliant, brokered version of the perp is arriving, and it will look and feel very different from opening a permissionless app in a browser.

Payward’s plan to route regulated perps onto Hyperliquid through Bitnomial-deployed HIP-3 markets is that same door, one step further on-chain: a CFTC-licensed exchange, clearinghouse and broker wrapped around a permissionless protocol. But the gap between the two Hyperliquids is the single most important thing for a US reader to understand. The public, permissionless protocol that anyone can reach from a wallet is not the regulated wrapper. A US retail trader who uses the open venue directly sits outside the perimeter, with no broker, no clearinghouse, and no recourse if a market is delisted, an oracle misfires, or a position is auto-deleveraged. The convergence of offshore perps and onshore rules is the running theme of our coverage of crypto’s shifting regulatory deadlines after the CLARITY Act stalled, and Hyperliquid is now the sharpest test case for it.

The risks a Hyperliquid trader carries

None of the above is a reason to avoid the venue or a reason to pile in. It is a map of what you are actually holding when you open a position. The specific risks stack up like this.

• Oracle risk: the price that marks, liquidates and funds your position is published by the validators; on HIP-3 markets it is published by the deployer.
• Governance risk: a market can be delisted and force-settled by validator vote within minutes, as JELLY showed.
• Liquidation and ADL risk: a violent move can force-close even a winning position through auto-deleveraging.
• Counterparty and vault risk: HLP is the backstop, and its assets have swung by more than half in a year.
• Deployer risk on HIP-3: builder markets have no HLP backstop, so their safety depends entirely on the operator.
• Concentration risk: a small, partly foundation-controlled validator set runs closed-source software.
• Collateral risk: positions are margined in USDC, so the whole structure rests on a stablecoin holding its peg.
• No-recourse risk: using the permissionless protocol directly puts a US trader outside the regulated perimeter.

Put those together and a clear-eyed picture emerges. Hyperliquid is genuinely impressive engineering that has out-executed better-funded rivals, and it settles more on-chain than almost any venue of its size. It is also more concentrated, more discretionary and more legally exposed for a US retail user than its polished interface suggests. Neither of those truths cancels the other. The right posture is the one any derivatives trader should hold anyway: size positions as if the worst-case mechanism, an oracle glitch, a governance freeze, a forced deleverage, will eventually fire, because on a long enough timeline one of them will. The venue’s own short history is already the evidence.

What to watch from here is a short list. Whether the CME suit is dismissed and Payward’s regulated HIP-3 markets actually open to US clients; whether RWA perps keep eating the buyback and how the foundation resolves the revenue paradox; whether the validator set genuinely decentralizes; whether the supply-burn proposal passes; and whether HYPE’s price-discovery run holds after a record near $98. Hyperliquid spent two years proving that an order book can live entirely on-chain and still feel like a centralized exchange. The next two will decide whether that is a feature Wall Street adopts, a concentration risk regulators eventually rein in, or both at once.

Frequently Asked Questions

What is Hyperliquid, and is it a perp DEX?

Hyperliquid is a decentralized exchange for perpetual futures that runs on its own purpose-built blockchain. It is a perp DEX, meaning you trade leveraged, no-expiry contracts directly from your own wallet rather than depositing with a custodian, and by 2026 it handles most on-chain perpetual trading volume.

How does Hyperliquid make money, and what is the HYPE buyback?

The protocol charges trading fees, and between 97% and 99% of them flow into its Assistance Fund, which buys HYPE on the open market and removes it from circulation. That buyback is the main way value flows back to the token, though the growth of HIP-3 builder markets, which pass half their fees to third-party deployers, has cut into it.

What are HIP-3 real-world-asset perps?

HIP-3 lets anyone who posts a large, slashable stake of HYPE deploy their own perpetual market, run their own oracle, and keep half the fees. It has been used mostly to launch real-world-asset perps on things like oil, gold, Nvidia, Tesla and pre-IPO names, and by mid-2026 those markets overtook Bitcoin as Hyperliquid’s largest by open interest.

Is Hyperliquid safe and decentralized?

It settles every trade on-chain, but its validator set is small and concentrated, its node software is closed-source, and governance has shown it can delist and force-settle a market within minutes, as it did with JELLY in 2025. That makes it fast and resilient in a crisis, but it is far less decentralized than networks like Ethereum, and traders carry oracle, liquidation and governance risk.

Can US traders use Hyperliquid legally?

A perpetual future is a derivative, which in the US falls under the CFTC rather than the SEC. The open, permissionless protocol sits outside that regulated perimeter, so a US retail trader using it directly has no broker or recourse; the compliant path is arriving separately, through Kraken’s parent Payward routing CFTC-regulated perps onto Hyperliquid via a licensed deployer.

By Liam Brennan, Markets and DeFi desk, HOGE Wire.