A Crude Oil Perp on a Sunday Night: How a Derivatives Venue Became a Platform

Two venues can quote the same barrel of crude, off the same underlying market data, with the same tick size. On a Sunday night one of them is closed and the other is doing over a billion dollars a day. That gap isn't a technology story. It's a story about who is allowed to list a market, and how long it takes them.

What HIP-3 Actually Is

HIP-3 is Hyperliquid's permissionless market deployment standard, live since 13 October 2025. It lets anyone who bonds 500,000 HYPE stand up an independent perpetual futures market on HyperCore, choosing the underlying asset, the oracle, the collateral, the leverage limits and the fee policy, while inheriting the exchange's matching engine, margining, liquidation system and existing traders.

That last clause is the whole product. Launching a derivatives venue has historically meant solving the cold-start problem twice: building the infrastructure, then convincing market makers and traders to show up. HIP-3 removes the second problem entirely. A team with a view on an asset class and an oracle they trust can be live with a market in weeks, on a book that already has liquidity in it.

The bond is a security deposit rather than a fee. It returns after a 30-day hold and a seven-day unstaking queue, and it stays slashable throughout on a tiered scale that runs to the full amount for critical failures, with slashed funds burned rather than redistributed. In exchange, the deployer keeps up to half the trading fees their market generates. First three markets are free to list; anything beyond that is won through a rolling Dutch auction denominated in HYPE.

Why This Matters More in 2026 Than Ever

The framing that gets this wrong is "a DEX added more listings." What actually changed is the shape of the business:

• The venue stopped being the one deciding what trades. A listings committee is a bottleneck with a roadmap. An auction is a market. Once the exchange isn't picking assets, the range of things that can be listed is bounded by what someone will bond capital to support, not by what the core team has bandwidth for.
• The underlying stopped needing to be crypto. Perps are a synthetic wrapper. Given an oracle and collateral, the contract doesn't care whether the reference price comes from a token, a barrel, an index, or a private company's last round. That was always theoretically true. It became operationally true when deployment opened up.
• Traditional market hours became a product gap you could sell against. Crude futures on NYMEX run a 23-hour session with a daily break and a weekend close. Gold has its own gaps. A perp on the same reference price has none, and that difference alone is a reason to route an order somewhere new.

What Builders Actually Deployed

The market list is the most convincing evidence that this is a platform rather than a feature. As of mid-2026 the deployers include:

• Aura, running WTI and Brent crude, gold, silver and treasuries. Oil and precious metals alone accounted for over 67% of all HIP-3 volume in Q1 2026.
• Trade.xyz, which leads the standard by a wide margin with 24/7 markets on US equities and a synthetic index tracking the S&P 500. It accounts for over 90% of HIP-3 open interest, has passed $100B in cumulative volume, and peaked around $5.6B in a single day. Its S&P 500 contract, XYZ100-USDC, is the largest HIP-3 market by open interest at roughly $213M.
• Felix, with 250+ US stocks and ETFs, starting with a Tesla market in November 2025.
• Ventuals, listing perps on pre-IPO names including OpenAI, SpaceX and Cursor — exposure that has no conventional retail venue at all.
• Trove, running perpetuals on graded Pokémon cards, which is either a joke or the clearest possible demonstration that the platform doesn't care what the underlying is. Both readings are correct.

The concentration numbers are worth reading honestly. This is not a broad ecosystem of hundreds of deployers; it's a handful, with one of them holding most of the open interest. Platform economics with a single dominant tenant behave a lot like a partnership until the tenant list diversifies.

The 24/7 Argument, Concretely

By late March 2026, WTI crude was doing about $1.27B a day on the venue, Brent about $1.04B, and silver about $1.01B. Total HIP-3 open interest was around $1.74B and growing 25% week over week, and only seven of the top thirty markets by open interest were crypto pairs. By June, HIP-3 open interest had passed $3.2B.

The mechanism behind those numbers is unglamorous. When a geopolitical headline lands at 8pm on a Saturday, a trader holding oil exposure has, in traditional markets, no way to act on it until Sunday evening at the earliest, and a gap risk they can only watch. A perp that never closes converts that dead window into a tradable one. Round-the-clock availability made the venue a default place to trade commodities on a weekend, and weekend liquidity is self-reinforcing: the market makers show up because the flow is there, and the flow arrives because the makers are.

There's a real risk attached that the growth numbers don't show. A perp market is only as good as its oracle, and outside crypto the reference data is licensed, lower-frequency, and sometimes closed exactly when the perp is most active. A deployer running a weekend oil market is quoting against a price source that isn't updating the way it does midweek. That's a design constraint to underwrite deliberately, not an implementation detail.

Read the Market Share Numbers Carefully

This is where most coverage of perp DEXs falls apart, and it's worth being precise because the numbers in circulation differ by a factor of nearly twenty.

Depending on the source, Hyperliquid's "market share" in 2026 has been reported as roughly 3.9%, 31%, 39.5%, 44% and 70%. All of those are published figures. None of them are wrong. They're measuring against different denominators, and some of them aren't measuring the same venue at all.

The two you're most likely to see quoted interchangeably are the ones furthest apart. Around 3.9% is Hyperliquid's April 2026 volume of $190.28B set against all perpetual futures volume, centralized venues included. Around 31% is that same $190.28B set against the $611.57B monthly average across the top twelve perp DEXes. The first says on-chain perps are still a minority of the derivatives market. The second says one venue dominates the on-chain slice. Both are true simultaneously, and a deck that quotes the second without the first is selling something.

The trend line matters more than any single figure, and it cuts in two directions at once. On-chain perps have taken durable share from centralized venues over two years. But perp DEX volume as a whole peaked at $1.36T in October 2025 and fell for five straight months to $699B in March 2026, a 49% decline, with 4 April 2026 marking the first day under $10B of volume since September 2025. And Hyperliquid's dominance within that shrinking pool halved, from over 77% in early 2025 to under 40% by April 2026, as edgeX, Aster, Lighter and Grvt took share.

So the accurate version of "perp DEXs are eating CEX share" is narrower than the headline: the category has grown structurally over two years, it contracted sharply through the first half of 2026, and the leader is less dominant than it was. The growth story in 2026 is not crypto perp volume. It's what got built on top.

The Part the Volume Headlines Hide

Here's the detail that separates this from a hype roundup. Builder-deployed markets went from roughly 2% of Hyperliquid's perp volume at the start of 2026 to nearly half of it by June, and in the week of 13–19 July, real-world-asset perps were about 52% of weekly volume — around $25B, the first week tokenized assets led every category on the venue. RWA open interest hit a record $3.6B in July, passing bitcoin to become the largest market on the platform by open interest. RWA perp volume in July reached 99.2% of bitcoin perp volume.

By any product measure, that's a successful platform strategy executed in nine months.

Gross protocol revenue over the same stretch fell every single quarter: roughly $357M in Q3 2025, then $295M, then $217M, then about $202M in Q2 2026. That's a 43% decline from the peak, during a period that included record open interest.

The mechanism is not mysterious, and it's the thing to understand if you're designing platform economics of your own. Deployers keep up to 50% of the fees their markets generate. Builder markets grew from 2% of volume to about half of it. So an increasing share of the venue's flow now arrives with half the revenue already spoken for. Roughly 97% of trading fees route into the Assistance Fund, which buys HYPE on the open market and retires it, and that fund has taken about 44.5M tokens out of supply. It bought roughly $290M of HYPE in Q3 2025 and about $149M in Q2 2026, close to half as much.

Growth in volume and growth in protocol revenue stopped being the same thing. That's not a flaw in the design so much as the price of it: the venue traded margin per trade for a category of flow it could not have listed itself. Whether that trade pays off depends entirely on whether the new markets are additive over a full cycle or simply cheaper versions of flow the venue already had.

HIP-4: The Same Trick, Applied to Outcomes

HIP-4 arrived on mainnet on 2 May 2026 and applies the platform pattern to prediction markets. The instruments are fully collateralized, expiry-based contracts that settle at either 0 or 1, bought at decimal prices between the two, with losses capped at the initial stake. The first market came from Outcomexyz, a daily bitcoin price prediction.

Phase one is curated rather than permissionless. The stated plan is for builder deployment to follow the HIP-3 staking model later in 2026, which would put outcome markets on the same footing as commodity perps: anyone with a bond, an oracle and a resolution source can list one.

If that ships as described, the interesting question stops being about crypto trading entirely. It becomes whether a single margin account, a single collateral pool, and a single matching engine can host crypto perps, commodity perps, equity index perps and event contracts at once — and whether regulators in each of those asset classes take the same view of that as the engineers do. That last part is genuinely unsettled, and anyone building in this direction should treat jurisdictional exposure as a primary design input rather than a compliance review at the end.

What This Means If You're Building

The transferable lesson here has very little to do with perps:

• Permissionless deployment is a distribution strategy, not a feature flag. The asset list grew far faster than any internal listings process could have managed, because the bottleneck moved from a team's roadmap to a market's willingness to bond capital.
• Price the revenue share before you need it. A 50% split looks generous when builder markets are 2% of volume and expensive when they're 52%. If your platform take rate is lower than your own, model the crossover point at the start.
• Shared risk infrastructure is the moat, not the order book. Deployers aren't renting matching. They're renting margining, liquidation and a trader base. Those are the hard parts, and they're what makes the inherited column in the diagram above worth 50% of fees.
• Concentration risk in a marketplace is real. One deployer holding over 90% of open interest means the platform's numbers are that deployer's numbers. Diversifying tenants is a roadmap item, not an outcome.
• When you go non-crypto, the oracle becomes your product. Every claim about 24/7 trading is a claim about a price feed that has to behave when its underlying market is closed.

The Bottom Line

The interesting thing about Hyperliquid in 2026 isn't the token, and it isn't a market share number that changes depending on who's dividing by what. It's that a derivatives exchange stopped choosing what it lists and became a place other people build markets — and within nine months, most of its flow was in assets its own team never listed.

The volume numbers are real. So is the 43% revenue decline that came with them, and so is the fact that perp DEX volume overall contracted through the first half of the year. Both belong in the same paragraph. A venue that opens its listing process trades margin for reach, and the honest way to evaluate it is to ask whether the reach was worth the margin. That answer isn't in yet.

Frequently Asked Questions