Who Owns
the Trading
Stack?

1. The Control-Point Question

Different markets reward different control points. The pattern repeats across every information-intensive industry of the last forty years:

Each of these companies did something other firms could have done, but they did it from a control point that turned every transaction in the market into a toll. Two of them were infrastructure plays; two were distribution plays; one was a vertical bet on owning both. None of those bets was obviously the right one at the time. They became right by the structure of the underlying market and the contestability of the layer below.

Hyperliquid and Jupiter are running this same playbook against the crypto trading stack. Hyperliquid is building infrastructure from the consensus layer up. A custom BFT consensus, a custom EVM execution layer, unified margin across spot, perpetuals, and HIP-4 outcome markets, plus a permissionless HIP-3 deployer surface that now extends into pre-IPO valuations via Ventuals. Jupiter is building distribution from the user layer down. A dominant Solana DEX aggregator share, a perpetual exchange, a lending product that hit $1.5B TVL inside eight days, a stablecoin (JupUSD) that distributes through the aggregator's installed base.

2. What Each One Owns

Stated plainly so the asymmetry is unambiguous.

Hyperliquid: the stack

Hyperliquid owns its own consensus (HyperBFT), its own EVM execution layer, its own native stablecoin (USDH), and a unified margin system that spans spot, perpetuals, and the HIP-4 outcome markets that shipped to mainnet on 2026-05-03. The platform's own products run on this stack, plus a growing roster of HIP-3 deployers shipping markets the platform itself does not curate (equity perps, FX, commodities, and most recently pre-IPO valuations). Every dollar of volume across that surface settles through Hyperliquid and routes fees through its protocol-level economics.

Jupiter: the route

Jupiter owns the user journey on Solana. Roughly 85 to 95 percent of Solana DEX aggregator volume routes through it depending on period and source. From that anchor, Jupiter has expanded outward into a perpetual exchange, a lending product, a stablecoin (JupUSD), and a handful of additional surfaces (launchpad, DCA, limit orders, prediction markets). Jupiter does not control the underlying L1 it runs on. It controls what the typical Solana trader sees when they open a wallet, and that control point is what Jupiter is monetizing.

One way to read the two architectures: Hyperliquid bets the most valuable layer is the substrate. Jupiter bets the most valuable layer is the user relationship. Both can be true at once. The market does not reward them equally.

3. The Numbers, Sourced

The revenue picture, with both the trailing-twelve-month window and the current 30-day annualized run rate so the reader can pick. Same source for both sides (DefiLlama on the revenue lines, CoinGecko on market caps; pulled 2026-06-09).

*Jupiter total = Aggregator + Perpetual Exchange + Lend per DefiLlama. Other Jupiter products (JupUSD reserves yield, Studio launchpad fees, DCA/Limit, prediction markets) are not consolidated by DefiLlama and are excluded; including them likely pushes the figure up but not the order of magnitude.

The revenue ratio (top-line size) is 7.6-13.6x depending on window. The valuation multiple premium (price per dollar of revenue) is 2.0-3.5x. The conversation should not be "is HYPE expensive" but "what does each multiple imply about the durability of the underlying control point."

4. How Each Token Captures Value

The mechanisms are not equivalent and a sloppy comparison treats them as if they were. They are not.

HYPE: buyback-funded value accrual

Hyperliquid's value-accrual architecture routes approximately 99 percent of protocol fees (raised from 97 percent in September 2025) into the on-chain Assistance Fund. The Assistance Fund automatically converts those fees into HYPE buybacks at the L1 block-execution level. The HYPE accumulated at the Assistance Fund address (0xfefefefefefefefefefefefefefefefefefefefe) was formally recognized as burned via a stake-weighted validator vote (85% for, 7% against, 8% abstain). The address has no private key and is mathematically inaccessible without a hard fork; future fees routed to the AF are likewise considered burned at execution. Per the protocol's 2025 annual financials, the Assistance Fund had absorbed $800M+ across 37.8M HYPE cumulatively at the time of the vote. At current run rates the buyback-and-burn absorbs roughly 13 percent of circulating supply per year. This is buyback-driven value accrual through permanent supply reduction, not direct fee distribution.

JUP: active staking plus product-level economics

Jupiter distributes value to active stakers through a per-product fee-sharing program. DefiLlama's holders-revenue line on Jupiter Aggregator shows ~$13M annualized at recent rates; other product streams (Perp Exchange, Lend, JupUSD reserve economics) are distributed across separate mechanisms. The aggregate is harder to pin down than HYPE's because the architecture is spread across products rather than funneled into one buyback contract.

5. The Real Hyperliquid Risk: Concentration

The auto-generated article framed Hyperliquid's risk as "single new L1." That framing is too blunt. Hyperliquid's risk is concentration, and the concentration runs deeper than the chain itself.

The same architecture that produces extraordinary throughput (unified margin, sub-second finality, single fee-capture surface) concentrates every variable onto one system:

• Execution concentration. All HIP-3 markets settle through the same matching engine and the same liquidation logic. A bug at this layer touches every product at once.
• Liquidity concentration. HLP and the broader market-maker set provide depth across many books from one shared inventory. Stress on one venue propagates to the others through shared capital.
• Governance concentration. Validator set composition, oracle operations, and protocol-parameter changes route through a smaller decision surface than a typical multi-app L1 with independent dapps.
• Economic concentration. 99 percent of fees route through one contract (the Assistance Fund) and back into one token. The mechanism is clean but also the entire monetary policy of the platform.

None is fatal in isolation, and none has produced a material loss event to date. But the combination is the actual risk profile. The strength and the weakness are the same thing: concentration.

6. The Real Jupiter Risk: Dependency Plus Replaceability

Jupiter's risk is the inverse. The platform does not control any layer beneath the user interface; it depends on every layer beneath it being healthy.

Solana dependency. Jupiter's revenue today correlates closely with Solana DEX activity. If Solana's onchain trading volume contracts (regime shift, network outage, regulatory pressure, faster competitor), Jupiter's run rate compresses regardless of execution quality. The dependency is mitigable in principle (multi-chain expansion is a path Jupiter has publicly discussed) but for now Jupiter's growth ceiling is highly correlated with Solana's growth trajectory.

Aggregator replaceability. Solana DEX aggregators are not a moat in the same way an L1 is. The 85-to-95 percent share is sustainable only as long as no Solana-native competitor produces a meaningfully better route or as long as Jupiter's distribution stays sticky enough that route quality is secondary. Both could erode.

JupUSD precision. The auto-generated article said JupUSD was "backed by BlackRock's BUIDL." That elides a layer. JupUSD reserves are 90 percent USDtb (a GENIUS-compliant stablecoin issued by Ethena, collateralized by BlackRock's BUIDL Fund) plus 10 percent USDC. The distinction matters for risk attribution: JupUSD has Ethena-issuance risk on top of BUIDL-asset risk, not just BUIDL exposure directly. Jupiter does not hold BUIDL on its own balance sheet.

None of this is fatal in isolation either. It is the inverse profile of Hyperliquid's concentration: distributed dependencies rather than one concentrated control surface, with the corresponding distributed failure modes.

7. Which Control Point Wins

The historical pattern that decides between infrastructure plays and distribution plays is usually the contestability of the layer below.

Microsoft won the operating-system control point because the hardware layer beneath it was a commodity, contested by many competing makers (IBM, Compaq, Dell). Google won the distribution control point because the substrate beneath it (the open web on a contested set of browsers) was even more commoditized. Visa won the settlement control point because the bank layer beneath it could not standardize on its own. Bloomberg won the workflow control point because no single exchange controlled all the data the desk needed. In each case, the layer that won was the layer where competition was thinnest at the moment of contestability.

Crypto's trading stack is still at the contested-substrate stage. Multiple credible L1s compete for the trading flow itself: Solana, the Ethereum L2s, Base, BNB Chain, and now Hyperliquid as a derivatives-native L1 in its own right. While the L1 layer is contested, distribution control points like Jupiter benefit because users want a unified route across whichever L1 wins. Once the L1 layer consolidates, the value migrates to whichever player owns the winning substrate.

There are intermediate outcomes. Hyperliquid wins as a derivatives-native L1 while Solana wins as a spot-and-meme-trading L1, in which case both control points hold and the relative multiples normalize closer to revenue ratios. Or a third party (a future Base derivatives stack, a Coinbase-aligned venue, a TradFi entrant) eats both, in which case the contested-substrate phase persists indefinitely and distribution control points like Jupiter compound.

Falsifiers. Three conditions would tell you the framing here is wrong: (1) HYPE monthly fees fall below ~$30M for two consecutive months while non-Hyperliquid L1s grow derivatives share, weakening the infrastructure-wins case; (2) a Solana-native competitor takes more than 25% of Solana DEX aggregator routing within twelve months, weakening the distribution-wins case; (3) a clear winner emerges at the trading-L1 layer (regulatory clarity, ETF concentration, or a merger event), collapsing the contested-substrate phase early.

8. The Question That Matters

Investors often ask whether HYPE is expensive relative to JUP. The more useful question is whether infrastructure deserves a premium to distribution while the trading stack remains unsettled.

History suggests control points matter more than products. Microsoft did not win because Windows was the best operating system; it won because the OS was the right control point in a market with a commodity layer below it. Google did not win because Search was the best ranking algorithm; it won because distribution was the right control point in a market with a commoditized substrate below it. The same logic applies, in reverse, to every era when the substrate stopped being commodity.

HYPE and JUP are not competing tokens with different revenue numbers. They are competing bets on which control point crypto will reward. The disagreement is simply which one.