Writing Essay

How Hyperliquid changed perpetuals — and why the US is preparing anyway

Hyperliquid did not invent perpetual futures. It changed who has to be trusted for the order book — and who gets paid when someone else builds the front door.

~9 min read

I am a product leader and founder working across fintech, digital assets and blockchain. My bias is simple: complex technology should become useful at the moment a real person needs it.

Perpetual futures are not new. Centralized exchanges made them familiar to millions of traders years ago. What changed is the trust model around the book itself — and the product surfaces that sit on top of it.

Hyperliquid is the clearest example I’ve watched closely. Not because the screens look dramatic, but because the jobs users and builders can do are different from a CEX deposit flow and different from an AMM-first DEX demo. The order book lives onchain. Custody stays with the user by default. Listing and distribution became builder products. And in the United States, the story is not that Americans can open the Interface — they still cannot under the Terms — but that the asset and the category are already being priced and argued about in regulated channels.

The interaction model changed

On a typical centralized perp venue, the product asks you to deposit into an intermediate balance. You trust the exchange’s books, its matching engine, and its operational controls. That model works at scale when the operator is trustworthy and solvent. It also makes custody part of the product whether or not the marketing says so.

Hyperliquid’s product bet is different. The docs describe HyperCore as a fully onchain order book: every order, cancel, trade and liquidation happens onchain with one-block finality from HyperBFT. The claimed throughput is 200,000 orders per second. The Interface at app.hyperliquid.xyz is wallet-native. The company that runs the website does not take custody of user funds, and the Terms of Use are explicit that the Interface is not the exclusive way to reach the chain.

From a product lens, that rewrites the first job. Instead of “trust this exchange’s books,” the user is asked to trust their keys and a transparent state. Optional qualified custodians still exist for teams that need them — Anchorage, BitGo, FalconX, Fireblocks and Komainu are listed on the public site — but custody is no longer the default assumption baked into the funnel.

The craft question is whether that trust model can feel as usable as a CEX screen. Hyperliquid’s answer has been to keep the familiar CLOB surface while changing what settles underneath. That is a harder product problem than shipping another AMM pool with a nicer chart.

Builder codes and HIP-3 turned the venue into infrastructure

The second change is distribution. Builder codes let a front-end attach a per-order fee after a user approves it — up to 0.1% on perps and 1% on spot, processed onchain. Wallets and terminals can own the relationship and monetize routing into shared liquidity without rebuilding matching. That is closer to an AWS-style unbundling than to a single branded exchange.

HIP-3 goes further on listings. A deployer can stake 500,000 HYPE for at least 183 days and launch a perp DEX with its own oracle and margin settings, keeping a fee share. Permissionless markets reverse the old CEX listing funnel. Equities, commodities and FX-style exposures show up as builder products rather than as a committee decision inside one venue.

CoinDesk’s August 2026 reporting on RWA perps is useful here as a product consequence, not as a scoreboard. Builder-deployed markets moved from a small share of volume early in 2026 toward roughly half by mid-year in that reporting, and weekend trading in tokenized stocks and commodities continued when traditional venues were closed. Whether you like every listing or not, the job changed: Hyperliquid is less “one exchange” and more matching-and-liquidity infrastructure that other products can sit on.

There is also a quieter product surface in the Hyperliquidity Provider vault. HLP packages market-making and liquidation strategies into a community deposit product with positions visible onchain. On September 26, 2026, DefiLlama and the vault API both put HLP account value near $183 million. That is not the headline volume story, but it matters for craft: “be the house” became a deposit experience with auditability, rather than an opaque internal desk.

Numbers I’m watching

Scale claims in this category move fast, and public counters do not always agree. I prefer dated sources over homepage marketing numbers. On the evening of September 26, 2026, Hyperliquid’s public API showed roughly $3.2 billion in 24-hour notional across all perp DEXes on the network, with open interest on the order of $13–17 billion depending on whether builder-deployed books are included. The homepage has displayed higher 24-hour volume figures in the same window; when those conflict, I trust the API and independent dashboards more than a marketing counter.

A few facts still stick when you date them carefully:

  • All-time volume: the community Dune dashboard by 0xpanatagama showed roughly $6.2 trillion cumulative volume when I checked on September 26, 2026.
  • Category share: late-August 2026 snapshots from CryptoBriefing and BuiltOnBulk put Hyperliquid near 58% of tracked onchain perp DEX volume over 30 days, against a category around $423 billion.
  • Calendar year 2025: Bitwise cited about $2.9 trillion in trading volume for that year when it launched its spot HYPE ETF.
  • Genesis design: the Hyper Foundation’s genesis note allocated 31% of HYPE (310 million tokens) to the genesis distribution, with no allocations for private investors, centralized exchanges or market makers.
  • Fee destination: protocol fees are directed to community surfaces; docs and the public site describe roughly 99% of protocol revenue flowing to the Assistance Fund, which buys HYPE and burns it.
  • HIP-3 stake: deploying a builder perp DEX requires staking 500,000 HYPE for a minimum of 183 days.

None of those figures is a reason to treat the product as finished. Volumes are highly variable day to day. Open interest ranges depend on which books you count. Revenue and fee share shift as HIP-3 deployers keep more of the take. The point of watching the numbers is to stay honest about scale without confusing a live API pull with a permanent ranking.

What “US ready” actually means

This is the part that gets overstated in conversation. Hyperliquid’s Terms still define Restricted Persons to include anyone who resides in, is located in, is incorporated in, or has a registered office in the United States. Circumvention tools are prohibited. The Interface is not open to US persons under those Terms, and I am not writing as if it were.

What is real in the United States is engagement with HYPE as an asset and with the broader perpetuals category as a policy object.

In May and June 2026, three US-listed spot HYPE ETFs came to market: 21Shares THYP on Nasdaq, Bitwise BHYP on NYSE, and Grayscale HYPG on Nasdaq. Those products give regulated price — and in some cases staking — exposure to HYPE. They do not give access to Hyperliquid’s perp order book. That distinction matters if you care about product truth rather than headline proximity.

Policy infrastructure showed up as well. The Hyperliquid Policy Center launched on February 18, 2026 as an independent 501(c)(4), with Jake Chervinsky as CEO and a Foundation contribution of 1,000,000 HYPE. Separately, institutional custody rails listed on the public site make it easier for teams that need qualified custody to interact with the ecosystem without pretending the Interface is a US-licensed venue.

Broader US perpetual products at registered venues are a separate story. Progress in that category does not rewrite Hyperliquid’s Terms, and it should not be narrated as if it did. The Interface remains closed to Restricted Persons. The interesting product fact is the split: Americans can get HYPE exposure through spot ETFs and watch policy debate intensify, while still being blocked from the Hyperliquid Interface under the current Terms.

So the accurate US story, as I see it, is narrower and more interesting than “Hyperliquid opened to America.” The United States is already pricing the token and arguing about the category — ETFs, advocacy, incumbent pressure — while the product Interface remains closed to Restricted Persons. Readiness here means rails and policy attention, not a green light on the trading screen.

What this means for product work

I keep returning to three craft lessons.

First, custody is a product feature, not an afterthought. If your default is an intermediate balance, you have chosen a trust model whether you named it or not. Hyperliquid’s self-custody default forces a different conversation about keys, state and recovery — and it also forces clarity about when a qualified custodian is the right tool. Users notice when funding, balances and settlement feel coherent. They also notice when the trust story on the marketing page does not match the flow they are asked to complete.

Second, matching and liquidity can be unbundled from the front door. Builder codes and HIP-3 are the clearest signal that Hyperliquid wants to be infrastructure other products route through. If you are building a wallet, a terminal or a market-access surface, that changes make-versus-buy for execution. The fee baseline on perps is competitive with centralized venues — 0.045% taker and 0.015% maker at tier 0, with volume tiers on rolling 14-day weighted volume — but the more interesting product difference is where fees go and who can attach a builder cut.

Third, access policy and asset exposure are not the same product. Spot ETFs and advocacy can make an asset legible in the United States without opening the Interface. Product leaders who blur those layers end up promising a journey the Terms do not allow. Clarity here is not legal theater. It is part of the same craft standard I use on onboarding: do not ask a user to believe a path the product cannot deliver.

I am not arguing that every market should look like Hyperliquid, or that an onchain CLOB is always the right architecture. I am arguing that the useful question is no longer only “can we list perps?” It is who is trusted for the book, who gets paid for distribution, who can create markets, and which jurisdictions are engaging the asset even when the Interface stays closed.

That is the standard I carry into product conversations now: measure the work by what a real user or builder can actually do — and by what they still cannot.

Sources

For more on the work across Lume, Sonic, IOHK/IOG and NEM, see the selected work on sunilvallath.com or the writing hub.