Perp DEX Tokens in 2026: Does the Buyback Flywheel Work?

The breakout asset of this market cycle was not a new Layer 1 or a meme coin. It was a perpetual-futures exchange. Hyperliquid’s HYPE token trades around $90, for a market value north of $20 billion that ranks it among the ten or eleven largest crypto assets, according to CoinGecko. It got there with no venture round, no centralized matching engine, and, until this year, no legal way for a US resident to touch it.

How a perp DEX actually works (how orders match, who takes the other side of your trade, what happens when you are liquidated) is now well-covered ground. The quieter question is the one that decides whether any of this is a business: what does the token do? Governance votes are cheap. The harder claim is that a HYPE, a GMX or a JUP captures some slice of the fees its exchange prints every day. In 2026 almost every major venue converged on the same answer, and it has a name: the buyback.

This is a guide to perp DEX tokenomics: the fee machine underneath these exchanges, the buyback flywheel they all adopted, how HYPE, GMX, dYdX, Jupiter and Aster differ, and the increasingly loud argument over whether any of it builds durable value or simply rents a higher price for a while. Prices are in US dollars, and the regulatory frame is the American one, where the perpetual contract is the CFTC’s problem and the token is where the SEC’s questions live.

What a perp DEX token is actually for

A share of stock is a legal claim on a company’s earnings and assets. A crypto token, as a rule, is not. That gap is the whole problem of tokenomics. Historically a perp DEX token did one or more of four jobs: it let you vote on governance (fee levels, new listings, treasury spending); it earned you fee discounts if you held or staked it; it paid for gas and secured the validator set, on venues that run their own chain; and, in the version everyone actually wants, it gave you some claim on the fees the exchange earns.

The first three jobs are real but weak price drivers. The fourth is the prize, and it is legally awkward. A token that simply pays you a cut of protocol cash flow starts to look a lot like a security, which is the SEC’s domain and a place most teams would rather not be (see our guide to SEC crypto enforcement in 2026). As the trading desk Keyrock put it in a 2026 study, tokens do not guarantee dividends, confer legal rights, or offer the clarity of earnings metrics. So teams reached for an indirect route to push revenue into the token without calling it a dividend. They buy the token back.

The fee machine: where the money comes from

Before the token, the fees. A perp DEX is a fee machine, and understanding tokenomics means understanding where that money originates. Most of it is trading fees, charged in basis points on every taker and maker order. On top of that sit liquidation penalties when a position is force-closed, borrow fees for holding leverage over time, and, increasingly, market-deployment fees paid by third parties who launch their own markets on the venue. The design of those fees, who pays, who collects, and who the protocol shares with, is the hidden plumbing that every token sits on top of.

One line item is widely misread: funding. The funding rate on a perpetual is a payment between longs and shorts that keeps the contract tethered to spot; it is not, by itself, protocol revenue, though some venues skim a share. The table below sorts the main sources.

On a high-volume venue these numbers are large. Hyperliquid’s gross protocol revenue peaked near $357 million in a single quarter in 2025, according to CoinDesk. That is real cash, earned in stablecoins, and the token is the pipe it is supposed to flow through.

Those fees are not flat. Like their centralized cousins, perp DEXs use maker and taker tiers, charging less to the market makers who post resting orders and more to the traders who take liquidity, and many cut those rates further for users who hold or stake the native token. That is the first and oldest link between trading and the token: activity feeds the fees, and the token buys a discount on the activity. The buyback simply adds a second, larger link on top, turning the same fee stream into steady demand for the token itself.

The buyback flywheel, explained

The mechanism is simple to state. The protocol earns fees, mostly in stablecoins. It uses some share of those fees to buy its own token on the open market. The purchased tokens are then burned, locked, or parked in a fund, which shrinks the circulating supply or at least adds steady demand. A higher token price makes the treasury, the team’s holdings and the venue’s incentive budget more valuable, which funds growth, which brings more traders, which prints more fees. Round and round: a reflexive loop that feeds on itself.

GMX ran a version of this as far back as 2021. In 2025 and 2026 the rest of the field piled in: Jupiter launched its buyback in February 2025, dYdX the same year, Aster overhauled its token around the idea in mid-2026, and Hyperliquid had been doing it since launch. Crypto protocols spent well over a billion dollars repurchasing their own tokens in 2025, The Block reported, and perp DEXs are among the heaviest spenders. The reason they buy rather than pay a dividend is partly regulatory caution (a cash distribution looks more like a securities payment) and partly flexibility, since a fund of tokens can be redeployed later. Whether the tokens are destroyed or merely set aside turns out to matter a great deal, as we will see.

It is worth being precise about what a buyback is and is not. It is not a dividend: holders receive no cash, only the hope that steady buying and a smaller supply lift the price. It is not a promise either, because a team can slow or stop the program the moment revenue dips. What it is, at best, is a credible signal that the people running the venue would rather spend their income supporting the token than sit on a pile of stablecoins. Whether that signal is worth anything depends entirely on the numbers behind it.

The shift is bigger than any single venue. For years the standard way to reward a token was emissions: print new units and hand them to liquidity providers and traders. That bought activity but diluted holders, and it trained users to farm the rewards and leave. Buybacks are the mirror image, spending earned revenue to take units off the market rather than printing more onto it. By 2025 the model had become an industry habit; the trading firm Keyrock estimated that revenue returned to holders through buybacks and direct distributions had grown more than fivefold since 2024, reaching close to $800 million in a single quarter, with perp DEXs near the center of the shift, per Keyrock.

HYPE and the Assistance Fund: the most aggressive flywheel in crypto

No venue runs the loop harder than Hyperliquid. Its Assistance Fund routes almost all of the exchange’s trading fees, reported at between 97% and 99%, into open-market purchases of HYPE, which are then removed from circulation; the fund has burned tens of millions of tokens. By the middle of 2026 it had spent more than $1.3 billion buying HYPE, an annual pace equal to roughly 7% of the token’s market capitalization, several times more aggressive than comparable protocols, according to research from AMINA Bank.

The fund was not originally built as a buyback engine. It began as a reserve meant to backstop the system in emergencies, then grew into the most visible value-accrual machine in the sector as the team pointed its fee income at HYPE and let the purchases run continuously, without the quarterly votes or discretionary pauses that slow other programs. That relentlessness is the point: where GMX and Jupiter openly debate whether to keep buying, Hyperliquid’s purchases are close to automatic, which is why the fund has become shorthand for the token itself.

In the autumn of 2026 the protocol added a second fuel line. A mechanism known as AQAv2 began directing the yield earned on the exchange’s large USDC reserves into the same buyback, on top of trading fees; CryptoBriefing reported roughly $15 million of USDC yield earmarked for HYPE purchases in early October. The effect is that almost every dollar the venue earns, from fees or from float, ends up chasing the token. That is why HYPE behaves less like a governance stub and more like a concentrated claim on the exchange itself, and why its price has tracked protocol performance more tightly than most tokens track anything.

The scale has started to unnerve incumbents. Jeffrey Sprecher, founder and chief executive of Intercontinental Exchange, the owner of the New York Stock Exchange, told a Bernstein conference that Hyperliquid is “bigger than NASDAQ,” adding, “It’s 11 people,” as reported by CoinDesk. A tiny team is running fee flows large enough to worry the people who own the Big Board, and nearly all of those flows are pointed at one token.

Why HYPE is more than a fee token

HYPE has an advantage the others lack: it is the native asset of its own Layer 1, not a token riding on someone else’s chain. So its demand does not rest on the buyback alone. HYPE pays gas on HyperEVM, the general-purpose smart-contract layer Hyperliquid launched in early 2025. It secures the network’s delegated proof-of-stake validator set, where staking HYPE earns a native yield of roughly 2.4% plus tiered fee discounts, per AMINA Bank. And it backs HIP-3, the framework that lets outsiders launch new markets, each of which requires a large, slashable HYPE bond.

That staking yield is modest next to the double-digit returns some networks still advertise, and it reflects a maturing market for on-chain security (the same squeeze we traced in validator economics). But the ecosystem underneath it is real. HyperEVM pushed past $2 billion in total value locked across more than 175 teams during 2026, with protocols such as Pendle, Felix and Kinetiq each topping $1 billion, according to The Defiant. Stack it up and HYPE’s demand comes from four places at once: gas, staking, bonding, and the buyback. Most perp DEX tokens have only the last.

There is a circularity worth naming. The buyback is funded by fees, fees depend on trading volume, and volume on a young venue is partly bought with token incentives and the promise of future airdrops. So the token helps generate the very revenue that is later used to support the token. On the way up that circle is a virtuous one. The danger is that it can spin the other way: if the incentives stop and traders drift off, the fees and the buyback shrink together, and the four pillars holding HYPE up begin to lean on one another instead of standing apart.

GMX: the original real yield

GMX, the oracle-pool exchange that runs on Arbitrum and Avalanche, wrote the first draft of this playbook. Its 2021 pitch was “real yield”: stakers earned a share of actual trading fees, paid in ETH or AVAX, rather than in freshly printed tokens. The split sent about 70% of fees to the liquidity providers who act as the house, and roughly 30% to GMX stakers; more than $134 million has reached stakers since launch, according to Coin Bureau.

In 2026 GMX changed shape. Rather than pay stakers directly, the protocol now routes about 27% of fees into buying GMX back, and it paused the distribution of those rewards, accumulating the repurchased tokens in the treasury until the price reclaims the $90 mark. The move may reduce sell pressure, but it trades the clean “stake for cash flow” story that made GMX distinctive for something closer to price engineering. It is the oldest token in this group and the one wrestling most openly with what a buyback is really for.

dYdX: three-quarters of revenue, bought and staked

dYdX, which abandoned Ethereum for its own Cosmos-based chain, started modestly. Its first buyback, introduced in 2025, committed 25% of net protocol fees to repurchasing DYDX, as The Block reported. The community then voted, through proposal #313, to triple that to 75% of net fees, with small slices sent to the treasury and the protocol’s MegaVault.

The twist is what dYdX does with the tokens. It does not burn them; it buys and stakes them, because DYDX secures the chain’s validators, so the buyback doubles as a security budget, per the dYdX Foundation. By the start of 2026 the program had bought and staked roughly 8.46 million DYDX, and the last of the token’s early-investor unlocks cleared by June 2026, lifting a long-standing supply overhang. The result binds token value to both fees and the cost of keeping the network safe.

Jupiter and the Litterbox: when the flywheel stalls

Jupiter, the Solana trading hub that runs both an aggregator and a perps venue, built a buyback with a memorable name. Its Litterbox Trust, launched in February 2025, takes 50% of protocol fees to buy JUP and lock it away for three years rather than burn it, as The Block reported; a 2026 governance proposal pushed to raise the share to 70%.

Jupiter is also the cautionary tale, told by its own founder. In early January 2026 co-founder Siong publicly questioned the program, noting that it had spent more than $70 million buying JUP with little effect on the price, and floated redirecting the money to user growth instead, as covered by crypto.news. The reason was brutal arithmetic: the buyback was absorbing only about 6% of the JUP being unlocked each month, while circulating supply ballooned and the token sat down roughly 89% from its peak. The lesson generalizes. A buyback is only as strong as the supply it is fighting; against heavy unlocks, even eight figures of buying can vanish without a trace.

Aster: incentives first, burn later

Aster, the BNB Chain venue backed by Binance founder CZ and YZi Labs, launched its ASTER token in September 2025 with a very large emission schedule and leaned hard on trading incentives to win volume. In June 2026 it overhauled the model to resemble Hyperliquid’s. The protocol now directs 99% of daily platform fees into automatic ASTER buybacks for staked holders, with matching burns that retire the team’s allocation first, aiming to shrink supply from roughly 8 billion tokens toward 3 billion, according to KuCoin. It also cut monthly emissions by about 97%, from 78.4 million tokens to 1.8 million, and froze insider unlocks until late in the year.

On paper that is as aggressive as HYPE. The caveat is the one this sector keeps running into: Aster’s headline volumes have been questioned as partly incentive-driven and self-reported, and DeFiLlama removed its perpetual volumes from the rankings in late 2025 over suspected wash trading. A policy of routing 99% of fees to the token means little if the fees themselves are inflated. Aster’s tokenomics are a bet that the activity turns organic before the incentives run out.

Five models, one scoreboard

Lined up side by side, the five tokens rhyme more than they differ. Each takes a slice of fees and points it at the token. What separates them is the detail that decides whether the flywheel actually turns: how big the slice is, what happens to the bought tokens, and how much utility the token carries beyond the buyback itself.

Burn, hold, lock, or stake: the choice is not cosmetic. Burning permanently cuts supply; locking only defers it; holding in a treasury leaves tokens that can later be sold back into the market; staking ties the buyback to network security. Read the row carefully before you read the price.

If you want one question to rank them, ask which model best survives a long drought. A burn paired with broad utility, HYPE’s combination, keeps working even when fees fall, because the supply cut is permanent and the token still has jobs to do. A three-year lock, like Jupiter’s, only delays the day those tokens come back to the market. A treasury that merely holds repurchased tokens, like GMX’s paused program, can reverse itself the moment the team decides to sell. The strongest tokenomics are the ones that still make sense in a bear market, not only in a bull one.

Does the buyback flywheel actually create value?

Here is where 2026 got honest. A buyback is not a dividend and not a legal claim on anything; it is discretionary demand that a team can start, stop, or shrink. Keyrock’s study noted a structural flaw: programmatic buybacks tend to spend more when the token is expensive and less when it is cheap, the opposite of how a disciplined buyer would behave. Jupiter’s experience put hard numbers on the skeptics’ case, and the skeptics were loud. Lex Sokolin of Generative Ventures has argued that token buybacks create far less demand than the selling pressure coming from vesting and emissions, a view reported by The Block.

It helps to compare this with the stock market, where a buyback rests on a foundation that crypto usually lacks. When a listed company repurchases shares, each remaining share is a legal claim on real earnings, so shrinking the count mechanically lifts earnings per share. A token buyback shrinks supply too, but the token underneath carries no enforceable claim on anything, so the effect is sentiment and scarcity rather than arithmetic. That is not worthless, but it is softer, and it leans entirely on traders continuing to believe the fees will keep coming.

Zoom out and the picture is sobering. Even among the minority of protocols that share real revenue, Keyrock counted only about a dozen sending a serious slice back to holders, on the order of two-thirds of their revenue, while the vast majority of tokens pass through almost nothing and work as governance badges, per Keyrock. Perp DEXs are the exception that proves the rule: they earn enough, and route enough, for the buyback to be more than decoration. For most of the market, a buyback headline is marketing wrapped around a revenue line too small to move the price at all.

There is a deeper problem too. Buyback fuel is cyclical, because it is a bet on trading volume, which is itself a bet on volatility and leverage appetite. In a quiet market the fees dry up exactly when the token most needs support, and the reflexive loop that lifted the price on the way up runs in reverse on the way down. It is the same uncomfortable lesson from other corners of the market, where a token’s advertised job did not translate into real demand, as with the restaking security that nobody turned out to be renting.

The fair counterpoint is that buybacks can work when the conditions are right: durable revenue, small unlocks, and a consistent, rules-based program paired with genuine product investment. On that test HYPE is the strongest case in the group, with the largest fee base and a shrinking float. The honest read is that a buyback is a multiplier on real revenue, not a substitute for it. Where the revenue is real and growing, the flywheel compounds. Where it is thin or faked, the buyback just burns runway that could have funded the product instead.

The cracks: thinning revenue, unlocks, and float

Even the best case shows strain. Hyperliquid’s gross protocol revenue has fallen for most of the past year, from the roughly $357 million peak in the third quarter of 2025 to about $202 million in the second quarter of 2026, before stabilizing in the third, according to CoinDesk. The cause is structural and, ironically, a product of success: HIP-3. Builder-deployed markets keep 50% of the fees they generate, and HIP-3 has grown from nothing to more than a tenth of gross fees in under a year. More volume, but a smaller share reaching the Assistance Fund that backs HYPE. Deployers also pay a steeply discounted fee while the segment is in growth mode, and the team has signaled it wants to narrow that discount over time, which would claw some of the revenue back toward the token. For now, though, the pipe that feeds the buyback is carrying less than the headline volume implies.

Then there is supply. HYPE’s circulating supply is only about 222 million of a one-billion maximum, so its roughly $20 billion market cap sits against a fully diluted value near $87 billion (CoinGecko). Most of the token has not yet reached the market. On 7 October 2026 that abstraction turned concrete: Hyperliquid Labs unstaked 3.75 million team-allocated HYPE, worth roughly $330 million and about 1.7% of circulating supply, for an over-the-counter sale to a single undisclosed institution, as reported by KuCoin. Selling off-market avoids direct pressure on exchanges, but it creates a large concentrated holder and is a reminder that insider supply is moving even as the buyback shrinks the float. Every token in this group faces some version of that squeeze: the fuel can thin while the overhang stays heavy.

Put the two pressures together and you get the real stress test for the whole group. On one side, the fee stream that funds the buyback is cyclical and, for the leaders, is being quietly shared with outside builders. On the other, large tranches of supply are still scheduled to unlock over the next several years. A flywheel that looks unstoppable at the top of a volume cycle can stall when fees halve and a vesting cliff lands in the same quarter. None of this is unique to Hyperliquid; it is the shape of the entire category, and the leader simply shows it first.

Who regulates the token: the CFTC owns the perp, the SEC eyes the token

US oversight splits cleanly, and tokenomics lands on the sharper edge of it. The perpetual contract is a derivative, which places the trading product under the Commodity Futures Trading Commission, not the SEC; that is the lane through which regulated perpetuals have reached American venues. The token is a different object. A token deliberately engineered to funnel protocol cash flow to its holders is close to the textbook description of what the SEC’s securities test is built to catch, which is precisely why these teams run buybacks instead of paying dividends: a repurchase is far easier to argue is not a distribution.

At the same time, the spot token has walked in through the SEC’s own front door. By late 2026 US issuers including Bitwise and Grayscale had brought spot HYPE products to market, with Grayscale’s structured as a staking vehicle that passes network yield through to holders, per its SEC prospectus; it is part of the broader shift in how crypto ETFs now get approved. The irony is sharp. Because the exchange itself is non-custodial and permissionless, a US resident still cannot legally trade on it directly, the same wallet-level reality we covered in account abstraction; the compliant route to the asset is a wrapper that holds the token and never touches the perp.

What it means if you hold a perp DEX token

The buyback flywheel turned perp DEX tokens from governance stubs into the closest thing DeFi has to an equity-like claim. That is genuine progress, but it is not a guarantee, and the difference lives in a handful of numbers that, for once, are mostly on-chain and checkable. Before you treat one of these tokens as a bet on its exchange, work the checklist.

• Is the revenue real? Read protocol fees and open interest, not headline trading volume, which is the easiest metric to fake.
• What share of fees actually reaches the token, and is it burned, locked, staked, or merely parked in a treasury that can sell later?
• How much supply is still unlocking? A buyback smaller than the monthly unlock is swimming upstream, as Jupiter learned.
• Does the token do anything beyond the buyback, such as paying gas, securing a chain, or bonding new markets?
• Is the fee base cyclical, and could it survive a long, quiet, low-volatility market?
• How wide is the gap between circulating market cap and fully diluted value?

None of these checks require insider access. Protocol fees, open interest, buyback spending, unlock schedules and circulating supply are all published on-chain or by independent trackers, a genuine break from the opacity of early crypto. The tokens that look cheap in hindsight tend to be the ones where the revenue was real, the buyback was disciplined, and the supply overhang was finally behind them. The ones that look expensive are those propped up by a buyback quietly smaller than the next wave of unlocks.

Answer those honestly and the token stops being a slogan and starts being a spreadsheet. In a corner of crypto built on leverage and narrative, that is not the worst place to stand.

Frequently Asked Questions

What is a perp DEX token actually used for?

Governance, fee discounts, and, on venues that run their own chain like Hyperliquid and dYdX, paying for gas and securing the network. The headline use in 2026 is value capture: the exchange uses its trading fees to buy the token back, giving holders indirect exposure to protocol revenue.

How does the Hyperliquid HYPE buyback work?

The Assistance Fund routes almost all of Hyperliquid’s trading fees, reported at between 97% and 99%, into open-market HYPE purchases that are removed from circulation. It had spent more than $1.3 billion by mid-2026, and a 2026 mechanism called AQAv2 adds the yield earned on the exchange’s USDC reserves on top of fees.

Do token buybacks actually raise the price?

Not reliably. They help when revenue is durable and unlocks are small, but Jupiter spent more than $70 million buying JUP in 2025 with little price effect because monthly unlocks dwarfed the buying. A buyback is best understood as a multiplier on real revenue, not a substitute for it.

Is HYPE a security, and who regulates perp DEXs in the US?

The perpetual contracts are derivatives under the CFTC. The token is a separate question, and one engineered to pass protocol cash flow to holders is the kind of instrument the SEC scrutinizes, which is one reason venues buy back rather than pay dividends. US residents cannot trade on the DEX directly but can now buy spot HYPE ETFs.

Which perp DEX has the best tokenomics?

It depends on what you value. HYPE has the most aggressive buyback and the broadest utility but falling protocol revenue; dYdX buys and stakes its token for network security; GMX carries real-yield heritage but has paused rewards; Jupiter locks rather than burns; and Aster looks aggressive on paper but is dogged by questions about how real its volume is.

Liam Brennan covers markets and on-chain derivatives for HOGE Wire.