Half the Tape? What HIP-3 Builder Markets Really Take and Give

Revised Sep 14, 2026: this report now runs on a market-by-market census pulled directly from Hyperliquid's perpDexs endpoints (all ten builder DEXs, asset-by-asset), replacing earlier third-party share estimates. Every table below is re-derivable from the API.

The question in the title is deliberately pointed, because the popular answer is wrong. Builder-deployed markets under HIP-3 are frequently described as taking over half of Hyperliquid's tape. We pulled every builder DEX, market by market, and measured it: HIP-3 open interest is $4.12B inside a $14.44B platform — 28.5% of the book, and roughly 30% of recent volume. Large, growing, strategically central — but not half. That measured share is the honest starting point for the real questions: what deployers take (a fee split that can reach 50% on their markets), what Hyperliquid gets (SP500, gold, SK Hynix and Brent oil perps a core team would never list), and what happens when a deployer exits — which, as the HyENA shutdown shows, they now do.

1. Background — what HIP-3 actually is

HIP-3 is Hyperliquid's builder-markets standard: it lets approved deployers launch and operate their own perpetual markets on Hyperliquid's infrastructure — their own symbols, oracle configurations, fee schedules and market parameters — while sharing Hyperliquid's order book engine, margin system and liquidity. A deployer is not a frontend; it is a market operator. The economic bargain is simple: the deployer brings markets Hyperliquid's core team would never build one by one — tokenized equities, commodities, single-name stock perps, niche crypto indices — and in exchange keeps a share of the fees those markets generate, up to half on their own markets. Everything runs on the same execution and margin engine, so the liquidity, the risk engine and the brand are shared; only the market ownership and the fee split are divided.

2. The census, in data

We queried the perpDexs registry and then each deployer's full asset context, on Sep 14. The structure that emerged:

Three findings deserve emphasis. First, the share is 28.5%, not 50%+ — and the historical shape is not a monotonic march: reconstructed snapshots put HIP-3 at ~18% of OI in March, rising through the summer to roughly a third in August, and settling back to 28.5% by mid-September as August's crypto rally re-inflated the core book faster than the builder book. The rotation is real; so is its ceiling so far. Second, "builder markets" is effectively "Trade.xyz": one operator runs 98% of builder open interest across 104 markets. Third, the deployer set is not permanent — HyENA's shutdown (zero live markets at measurement) is the first complete exit, and it happened quietly.

3. What deployers take — the fee economics, quantified

HIP-3's split rule: on core markets, essentially all net fees flow to the protocol, of which ~97% funds buybacks. On builder markets, the deployer can keep up to 50% of the fees its markets generate. The measurable aggregate effect shows up in the weekly reconciliation — here is the most recent verified week (Sep 7–13):

The structural point survives any single week's noise: roughly 18 cents of every fee dollar now routes to the builder ecosystem rather than the Assistance Fund — on the recent ~$13–16M weekly gross base, that is ~$2.3–2.9M a week, on the order of $120–150M annualized, redirected from token buybacks to market operators. That is the take. We label the ~18% as an estimate reconciled from weekly prints; if the transparency proposal at final vote passes, per-deployer splits become disclosed data instead of our arithmetic.

4. What Hyperliquid gets — the case for the bargain

The census makes the get concrete, because you can name the markets. The largest builder markets by open interest, all measured Sep 14:

• Market breadth no core team could build. Equities, memory-chip clusters, gold, Brent and WTI crude — none of this would exist on Hyperliquid without deployers chasing it. It is also the segment where venue volume decouples from crypto volatility: the SK Hynix complex trades on semiconductor catalysts, not BTC beta.
• Headroom by design. Deployers set explicit OI caps — $1B on CL, $750M on BRENTOIL, $200M on AAPL — meaning the largest RWA markets are running at a fraction of their configured ceilings. Growth in this tape does not require new listing decisions; it requires raising caps.
• A fee base that trades when crypto doesn't. Roughly 30% of the tape now runs on non-crypto or index catalysts. In a crypto risk-off week, that is the part of the book most likely to keep paying fees — which makes protocol revenue less cyclical than it looks, even after the split.
• The platform OI context. Total platform OI of $14.44B sits about 3% below the pre-peak of ~$14.7B (Oct 10, 2025) — the level that preceded a violent deleveraging. The builder book diversifying that exposure is a risk argument, not just a growth one.

5. The concentration picture — one operator, and one exit

Two facts reframe the concentration debate. First, the scale of it: Trade.xyz runs 98% of builder open interest (104 of the meaningful active markets). "28.5% of the tape" is really "28.5% of the tape, one operator" — a deployer-level oracle or contract failure there would hit more than a quarter of the entire platform book at once. That is a larger concentrated exposure than the old "half the tape" narrative implied in absolute dollars ($4.05B) while being smaller in relative share. Both matter.

Second, the HyENA shutdown — the first complete deployer exit. When we queried HyENA's registry it returned zero active markets, corroborated by dated reports. Nothing about HIP-3 locks a deployer in: a builder can wind down its book and leave. The implications cut both ways. Negative: open interest and fee flow attributed to a deployer can go to zero without protocol approval, and traders on those markets bear the transition. Positive: the mechanism is self-cleaning — weak deployers exit rather than zombie on, which is how a permissionless market layer stays credible. Watch whether HyENA's markets migrate to other deployers or simply disappear from the tape.

6. Scenarios

We weight C higher than most commentary does, because the census changes its denominator: a Trade.xyz incident no longer threatens a niche corner — it threatens $4.05B of open interest in one operational domain. The single variable that moves probability mass from C to A is the transparency proposal: disclosed per-market fee passes would let deployers compete on split, let the market price concentration, and let reports like this one measure net-new volume instead of estimating it.

7. Risks & what to watch

• Transparency proposal final vote. The highest-leverage governance event for this thesis. Adoption = disclosed fee passes = priceable concentration. Failure = the estimate regime continues.
• Trade.xyz operational health. Oracle configurations, margin parameters, incident history across 104 markets. Not disclosed today; watch unusual funding dislocations on its markets specifically.
• Further deployer exits. HyENA proved the exit door works. Each exit removes OI and fee flow; a cluster of exits would undercut the breadth argument this whole bargain rests on.
• Share trajectory vs. core. Watch whether the August peak (~one third of OI) or the current 28.5% is the trend. A sustained decline favors scenario B; cap-raises on SP500/GOLD/CL favor A.
• Competitive splits. Any rival chain publicly offering deployers a better take forces Hyperliquid's pass-through to mature faster — good for deployers, immediately negative for capture.

8. Conclusion

So — half the tape? No. 28.5% of open interest, roughly 30% of volume, $4.12B of a $14.44B platform, measured market by market rather than estimated from aggregator dashboards. That is the honest denominator, and it changes the framing in both directions: the take is smaller than the "half the tape" narrative implies, but the concentration is larger — one operator, 98% of the builder book, $4.05B — and the exits have started. The take remains real and measurable: roughly a fifth of every fee dollar, on the order of $120–150M annualized, routes to operators instead of the buyback. The get remains real too: SP500, SK Hynix, gold and oil perps with configured headroom to grow, and a fee base that trades when crypto doesn't. For HYPE holders, the operational summary is one line: track net capture, not gross share — and now that the census is public and repeatable, the share, the concentration and the exits are all trackable in the same place: the API itself.

FAQ

Related reading

• Hyperliquid's Flywheel, Measured: Fees, Buybacks and the Second Engine
• Hyperliquid Weekly Data Report — Week of August 31–September 6, 2026
• Hyperliquid Weekly News Review — Week of August 31–September 6, 2026
• HYPE Market Intelligence — Week of August 31–September 6, 2026