Hyperliquid valuation: HYPE isn’t cheap at $80

What the strongest bullish writing gets right

Artemis’s July 2025 valuation priced the exchange and HyperEVM independently. CoinShares later built a seven-stream framework, with a $147 base-case outcome in 2031 and 707 million tokens. Multicoin’s June 2026 report projected roughly $8 billion of 2028 earnings and a $319 token price using approximately 502 million tokens. Their future dates, revenue frameworks, and token denominators make them different valuation cases.

The best bullish insight is that applications can build around existing demand, execution and collateral. The next application need not recruit an entire financial market from scratch. The investment question is where that reuse lowers costs or creates pricing power. Market growth, on-chain adoption, durable fee policy and favorable supply each need evidence. An attractive story about one doesn’t make all of them likely together.

Multicoin holds trading monetization at 3.28 basis points and grows USDC balances approximately with turnover. It discloses a substantial HYPE investment. CoinShares explicitly avoids a second supply reduction for future buybacks, preventing double counting. My principal disagreements concern fee durability, capital efficiency and economic claims.

Damodaran’s Nvidia analysis after DeepSeek provides the better discipline: translate a changed story into changed market, economic and risk assumptions. Admiration does not settle price. The story tested here is an expanding execution network whose marginal dollar of turnover may earn less than its earlier dollar. ^S15

A real fee engine, with an accounting trap

DeFiLlama’s perpetuals page displayed $208.687 billion of trailing-30-day turnover, $85.39 million of fees and $69.72 million of holder revenue. The adapter behind that page also adds AQAv2 stablecoin reserve yield to fees and holder revenue. The product label says perpetuals; the numerator contains more than trading.

Dividing $69.72 million by $208.687 billion gives 3.34 basis points. That is a mixed-product receipts ratio, not an identified trading take rate. Extrapolating it into trading income and then adding stablecoin yield would count some economics twice. The forecast instead rebuilds each stream independently; its trading rates are assumptions.

Builder payments, deployer economics, maker rebates and the HLP allocation also divide user fees among claimants. Applying the headline 99% allocation again to an already-net series is another mistake. Payroll, taxes, and maintenance investment sit outside both series. On-chain receipts are not an audited company’s free cash flow. ^S3

The parent dashboard also fails a reconciliation: Q3 gross protocol revenue of $224.23 million less $45.42 million of costs exceeds the displayed $173.97 million gross profit by $4.84 million. That difference has no identified expense line. Even current price fields differ between the page header and FAQ. I therefore use $80 as a rounded reference. ^S1

Scroll horizontally to inspect the figure →

Today’s buybacks do not underwrite the price

The parent dashboard’s $71.65 million of 30-day holder receipts annualizes to approximately $872 million. Against $76.4 billion of fully diluted value, that is 1.14% of value per year. Against circulating value, it is 4.90%. Both ratios describe the same receipts. Neither is a dividend yield paid into a holder’s account. ^S1

A flat-perpetuity thought experiment makes the growth premium visible. Hold that $872 million stream constant indefinitely, allocate it across 955.31 million claims and require 18%: supported value is approximately $5.07 per token. This is not an alternative forecast or a liquidation floor. It strips away growth to show that most of the central valuation already pays for an expansion that has not happened.

The fee documentation states that Assistance Fund purchases convert fees into HYPE and that its HYPE is burned, permanently reducing circulating and total supply. That is stronger than a discretionary treasury repurchase. It still transfers one economic dollar only once: sellers receive cash, while continuing holders have fewer competing claims. A higher purchase price removes fewer tokens for the same budget. I have not independently replayed the on-chain burn record.

A burn is a meaningful economic mechanism without being a contractual distribution. Its value rests on the future stream and the policy routing that stream to the fund. Mature-exchange equity supplies a different set of residual and legal rights. I would use the comparison to understand fee generation, then value HYPE’s policy-dependent claim.

For scale, an illustrative $800 million repurchase at $80 buys ten million tokens: about 1.05% of 955.31 million modeled claims, versus about 4.50% of the retrieved 222.45 million circulating tokens. At $40, it buys twenty million. These examples hold price fixed and omit price impact. For the Assistance Fund, the documentation specifies permanent burn; the unresolved questions are the future budget, purchase prices and competing allocations.

The dollar budget, resulting claim reduction and new allocations must travel together. Capitalizing future token-directed receipts while also shrinking the denominator by the purchases those same receipts finance can count the benefit twice. The fixed-denominator model avoids that second credit. A dynamic burn-and-issuance model is another valid approach if purchases, price and participation are reconciled explicitly.

The next user may never visit Hyperliquid

Builder codes and HIP-3 perform different jobs. A builder can attach a user-approved charge to fills; a HIP-3 deployer defines and operates markets and has its own fee economics. ^{S18 S5} The additional builder charge does not mechanically reduce the protocol fee. The customer’s total trading cost still includes it, and deployer shares must be distinguished from the builder’s own receipt.

I expect those roles to matter more than loyalty to a trading homepage. I would measure adoption through persistent fee-paying order flow regardless of whether customers consciously choose a blockchain. A broker can make the infrastructure disappear from view without making it economically unimportant.

The price of execution is nevertheless larger than the posted exchange fee. It includes spread, slippage, funding, collateral requirements and the risk of an interrupted position. Deeper liquidity can make a higher-fee venue cheaper for a sufficiently large order. That is a credible source of pricing power. Matching software alone is not.

Automation strengthens the distribution opportunity while making routing more price-sensitive. An agent does not care whose interface is familiar. It can compare execution costs across venues. More automated activity can therefore raise turnover without creating proportionate fee income or independent new customers.

A distributor can route elsewhere when another venue offers a genuinely better complete execution. That bargaining position is real, but its income may come from concessions, its own charge, or both. I would compare protocol contribution and execution by route. More distribution could deepen demand and preserve the protocol’s price, or it could create a concentrated buyer able to negotiate it down.

The strongest bull case is that applications compete away their own margins around a scarce execution service. The scope of that scarcity matters. HIP-3 documentation describes independent order books and margining for each deployed perp DEX, with cross-margin features subject to additional conditions. Reusing an execution stack is not the same as automatically pooling every order and dollar of collateral. I would look for durable depth in the markets earning the forecast fees before awarding the whole network one universal liquidity moat.

The Kalshi distribution analysis examines how a route to the customer can become a competing exchange and change the fee economics.

More markets need not mean the same margin

S&P Dow Jones Indices has licensed the S&P 500 to Trade[XYZ] for perpetual contracts on Hyperliquid, describing access for eligible investors. Traditional data providers can participate in this network. An index license is not stock ownership or permission for every customer in every jurisdiction to trade. ^S8

The fee rules expose the economic catch. HIP-3 growth mode applies a 0.1 multiplier to protocol fee rates, with deployer settings affecting the all-in customer charge. New markets can produce large notional volume at much lower monetization. Builder and deployer income do not automatically belong to HYPE.

The central case puts $12 trillion of crypto turnover through a 2.5-basis-point net fee for $3 billion, while $8 trillion of non-crypto turnover earns 0.7 basis points for $560 million, so non-crypto activity supplies 40% of perpetual turnover but only 16% of its fee stream. Market mix changes what growth earns. The forecast assumes neither permanent universal subsidies nor effortless repricing.

Repricing needs its own demand case. Ending a 90% fee reduction won’t guarantee ten times the receipts because customers can trade less or switch. At one-tenth the volume, a tenfold fee leaves revenue unchanged before other responses. Higher yield is valuable when enough demand survives at the new complete trading cost. The forecast’s new-market contribution has to earn persistent demand after the promotional period.

HIP-3 also requires 500,000 HYPE of deployer stake, a requirement the documentation says may fall. At $80, that is $40 million tied to entry. It supports token utility but is neither annual revenue nor a permanent burn. A higher token price can make adding markets harder. ^S5

Scroll horizontally to inspect the figure →

Collateral is a business. Turnover is not a balance

Coinbase announced in May 2026 that it would become the official USDC treasury deployer on Hyperliquid. Under AQAv2, approximately 90% of cost-adjusted reserve yield associated with eligible on-network supply goes to the protocol. The protocol receives income from that defined balance; the principal remains the customers’ property.

The central case assumes $20 billion of eligible average balance, a 2.5% cost-adjusted yield and a 90% protocol share: $450 million in 2031. The yield is after reserve costs but before the protocol split. Total value locked is not substituted for eligible average stablecoin supply.

Better collateral use can make the trading product stronger while weakening a simple reserve-income extrapolation. More turnover per dollar of eligible balance produces less reserve income per dollar traded. Segmented margin accounts can impose costs in the other direction. The correct forecast follows average eligible supply and trading capital requirements; it does not assume one fixed relationship between volume and deposits.

A one-percentage-point reserve-yield decline costs $180 million annually at the modeled balance. Trading may offset part of that loss, but no offset is assumed. AQAv2 accrues over 30-day intervals and pays eight days after each interval; accrued yield and cash received are different observations. The dashboard already incorporates this stream. ^S6

The supply denominator is part of the thesis

At $80, the provider’s 222.45 million circulating tokens imply $17.8 billion of circulating value; 955.31 million total tokens imply $76.4 billion. Neither is equity enterprise value. Choosing the denominator can change the apparent bargain before a single business assumption changes. ^S1

The central allocation recognizes all 955.31 million potential claims. It does not predict that every token circulates by 2031. This is conservative, and a legitimate source of disagreement: allocations might be distributed later, receive no economic participation, or never be issued. A smaller denominator requires evidence for those outcomes.

Documented burns remove claims. Locks and delayed transfers can leave the economic interest alive, including vested allocations that benefit before free trading. I use all 955.31 million modeled claims because a precise annual participation curve is unavailable. A smaller denominator needs evidence about issuance and participation.

On unchanged operating assumptions, the model yields approximately $30 at 955 million claims, $40 at 707 million and $57 at 502 million. Assigning everything to the current float gives $129, but that ignores competing allocations. This sensitivity changes which claims share the flows while holding the operating forecast fixed. Future free float is outside the test.

Staking rewards come from the future emissions reserve; they redistribute tokens without creating external earnings. Unstaking is permitted. Similarly, I capitalize future HYPE-directed flows without also subtracting projected buybacks from supply. A repurchase model can be valid, but issuance, purchase prices and ownership must all reconcile. Customer collateral and HLP assets are not added as treasury cash. ^S10

Scroll horizontally to inspect the figure →

A model that lets the business keep growing

The first five annual periods end in 2027–2031. Each forecasts crypto perpetuals, non-crypto perpetuals, spot, reserve income, other net fees and hypothetical outcome fees. For unit checking, $1 trillion at one basis point produces $100 million. Modeled trading rates are already net to HYPE. Allocation policy controls the modeled flow to HYPE; corporate-dividend rights are outside the model.

The $12 trillion crypto endpoint is $200 trillion of assumed annual crypto-derivatives turnover, multiplied by 20% on-chain penetration and a 30% Hyperliquid share. For scale, CoinGecko measured $86.2 trillion at its top ten centralized perpetual exchanges and $6.7 trillion at its top ten decentralized venues in 2025. Their combined $92.9 trillion covers the two top-ten samples. ^S19

Decentralized volume is 7.2% of that combined sample; the 7.8% DEX-to-CEX ratio uses CEX volume alone in the denominator. The central case’s 20% assumption therefore demands considerably greater on-chain participation. Its $200 trillion market is a little over twice that historical sample; the report’s incentive-driven growth doesn’t verify it. The $8 trillion non-crypto opportunity is not added into the crypto market denominator.

Spot adds $180 million, reserves $450 million and other fees $50 million to the $3.56 billion perpetual stream, reaching $4.24 billion in 2031. This is already strong execution: annual perpetual turnover approaches eight times the retrieved annualized activity.

Growth does not abruptly stop there. The central final-year growth rate is 36.7%; a five-year fade to 3% produces $8.94 billion in 2036. All three operating cases use an assumed 18% required return, and the same supply allocation. The dominant case reaches $15.03 billion of annual flow in 2031 and $43.30 billion in 2036. The erosion case is not a liquidation floor.

Scroll horizontally to inspect the figure →

Three-quarters of the value arrives after 2031

The central value is $28.63 billion. Only $6.96 billion comes from the first five driver-based years. The next five contribute $9.94 billion, and the terminal stream contributes $11.72 billion. Saying that terminal value is 41% of the model understates its dependence on distant assumptions: 75.7% arrives after 2031.

Moving straight to 3% growth after 2031 reduces value to $20.61. Postponing the entire original stream by two years gives $21.52 at the same return requirement. The two-year delay is a timing sensitivity holding later cash flows and the required return fixed. Operating activity during that interval is outside the timing test.

The operating stresses also distinguish scale from economics. Reducing crypto retained fees by one basis point in every explicit year lowers value to $21.59. A one-percentage-point reserve-yield cut produces $28.68. In both tests, the original post-2031 growth path stays fixed; otherwise changing one year’s receipts could spuriously change every later year’s growth.

A new 20% claim on future modeled receipts reduces value to $23.97. It represents security, development, or distribution funding beyond recipients already netted out. A finite emissions reserve can’t fund an infinite operating life. I leave the eventual funding mix unresolved.

Each stress holds the business’s other responses fixed to show the cost of one assumption. In practice, fee cuts may stimulate volume, delays may preserve development resources, and better funding may lower the required return. These are independent sensitivities, with no probabilities; they are not a combined downside forecast or a confidence interval. The 18% return hurdle is not an annual failure probability. Adding a collapse probability would require an explicit risk model.

Scroll horizontally to inspect the figure →

What $80 asks the network to earn

At $80 and 955.31 million claims, reference value is $76.4 billion. Starting with $872 million of annualized parent holder receipts, the reverse model requires approximately 60% annual flow growth for five years, followed by the same fade method. Flow reaches $9.1 billion in 2031 and $27.6 billion in 2036. That starting benchmark already contains reserve income. ^S1

This reverse check solves for aggregate cash-flow growth while leaving the product mix unassigned. Earning $9.1 billion entirely at 2.5 basis points would require approximately $36.5 trillion of annual turnover. Reserves reduce that trading requirement; a lower-fee market mix increases it.

There is a less spectacular route to $80: keep the central operating forecast, use a 12% required return and allocate value across 707 million claims. The result is $79.62. A lower return hurdle alone gives $58.92; the smaller denominator alone gives $40.49. Neither individually explains the reference price.

That is the strongest counterargument to the central valuation. The market might be anticipating a safer claim and fewer participating tokens, instead of vastly more trading. I would need evidence for both concessions before accepting them together. The $30 estimate prices the conservative claim allocation used here.

An additional reverse check leaves every central operating and supply assumption intact and solves for the return requirement instead. The $80 reference clears at approximately 10.04%, versus the 18% hurdle used here. That is the return hurdle consistent with this conditional stream. A buyer comfortable with that return requirement can rationally reach a different conclusion without changing the volume forecast.

The model gives me no catalyst for convergence to $30. Price may reflect utility, other token-participation assumptions or a different view of growth. My objection is that $80 offers too little compensation for uncertainty under the claims I model.

HYPE also has utility through staking, deployment and fee discounts. Those benefits are not assumed worthless. They are simply not an unmeasured plug for the valuation gap. A discount saves a trader money; it is not another dollar of exchange profit. Utility, buybacks and economic flows must not be capitalized repeatedly.

The failure modes that matter

The largest operating risk is not running out of assets to list. It is losing retained economics as professional distributors and market makers gain bargaining power. On $12 trillion of annual crypto turnover, one basis point is $1.2 billion of annual receipts. A new listing matters less than the price of persistent execution.

Solvency rules allocate the costs of losses and disruption. The auto-deleveraging description closes selected profitable counterparties against an underwater account. HIP-3’s slashing rules say burned stake is not paid to affected users and distinguish protocol irregularities from subjective disputes. Those choices can protect the system while leaving a trader with a closed hedge or disputed market outcome. Trust in execution, market design and collateral can therefore weaken together; reserve income is not wholly independent diversification.

The incumbents can respond. CME reported $1.7 billion of Q2 2026 revenue, $1.1 billion of operating income and $238 million of market-data revenue, alongside initiatives including 24/7 one-ounce gold contracts. Those are not HYPE valuation multiples. They demonstrate diversified monetization and an ability to update its product. ^S16

Outcome markets deserve similar restraint. The main fee documentation describes outcome charges, while the HIP-4 proposal page still discusses zero-fee testing. DeFiLlama displays small outcome receipts. The central case leaves that small reported flow out as a conservative choice. The bull case includes hypothetical future income.

A better business is not a margin of safety

I expect the execution network to serve more workflows as applications and specialist markets build around it. Whether turnover outpaces receipts depends on which service remains scarce and how its fee survives the customer’s all-in cost comparison. Additional builder fees and permanent burns make the positive case more concrete. Independent books, operating risks and policy-dependent claims prevent it from becoming an automatic valuation conclusion.

The forecasts imply observable milestones. By 2028, the central crypto case requires $4.2 trillion of annual turnover at 3.0 net basis points, or $1.26 billion of annual receipts before modeled reserves. That corresponds to roughly $105 million per month of crypto-only trading receipts. A headline protocol number that includes reserve income or non-crypto activity can’t verify that milestone.

For non-crypto markets, the 2031 case requires about $21.9 billion of average daily turnover at the assumed fee mix. I would look for sustained trading after incentives change. Extra listings still need to produce that volume. In collateral, I would measure average eligible stablecoin supply across the period; a quarter-end balance and broad TVL use different denominators. If the reported data cannot isolate those measures, the forecast becomes harder to test.

There are three reasons to revise value, and they require different evidence. Persistent protocol receipts change the operating forecast. Burns or documented changes in participating allocations change the claim count. Durable security, access and fee policy can justify a different required return. An exceptional month of activity can support the first while leaving the other two unresolved. I would keep the valuation bridge as explicit as the fee routing.

I would upgrade for sustained fee-paying execution, executable depth after incentives and protocol receipts that remain attractive as distributors expand. Permanent burns belong in the supply accounting under the documented treatment; future allocations need evidence. I would accept a lower return requirement only for a more secure claim. Those tests give the infrastructure bull case a way to win without relaxing every assumption at once.

I anchor near $30 for the conservative claim count and 18% hurdle. At $80, a buyer needs better economics, fewer competing claims, or a lower required return. Hyperliquid can earn a stronger answer through sustained execution and durable fee policy. I’d pay for that particular scarce service and claim.

Assumptions

• Five driver-based annual forecasts end in 2027–2031. Five further years fade growth to 3% in 2036; terminal flows begin in 2037.
• All operating scenarios use an 18% required return and 955.31 million economic claims. Supply and return assumptions are tested separately.
• Trading rates are already net to HYPE. Reserve yield is modeled once. Future buybacks do not receive a second benefit through a shrinking denominator.
• Isolated fee and reserve shocks use the unshocked scenario’s growth-fade anchor. A diversion is an additional hypothetical future policy change, not a second deduction of existing recipient shares.
• 955.31 million is a conservative total-supply allocation, not a prediction of 2031 circulation. Unstaking, vesting, transfer and sale are different events.