In October 2026 the question is no longer whether a DEX token sees protocol revenue but which mechanism, what share, and whether anyone can check. Hyperliquid converts fees to HYPE automatically and treats the result as burned. Aster buys ASTER with 99% of daily fees and hands it to lockers. Uniswap switched its protocol fee on in December 2025 after four years of delays, dYdX lifted its buyback to 75% of net fees, and PancakeSwap voted on 24 October 2026 to give liquidity providers 88% of v3 fees while keeping the CAKE burn. Curve has paid fees to lockers since 2020, and Aero, the merged Aerodrome and Velodrome, re-based that model on stakers on 22 October 2026.
This guide records what each tracked venue's own documentation or governance record says as of 28 October 2026; where a venue states no number we say so rather than repeat a tracker's estimate. The second half covers what the announcements leave out: discretion, dependence on volume, the securities question and why a buyback is not a dividend. Some links are affiliate links; they do not affect which venues appear or how they are described.
- Four mechanisms: buy-and-burn (Hyperliquid, Uniswap, edgeX, Lighter, PancakeSwap), buy-and-hold (Jupiter, Raydium, dYdX), buy-and-distribute (Aster, GMX) and direct fee sharing to lockers or stakers (Curve, Aero). Only the last two put anything in a holder's wallet.
- Shares range from 12% of trading fees (Raydium) to 99% of daily fees (Aster). Hyperliquid's docs do not state the split; Lighter, Paradex, edgeX and GRVT publish no percentage.
- Buybacks track volume: DefiLlama's holder-revenue series puts Hyperliquid at $289.8M in Q3 2025 and $148.6M in Q2 2026, edgeX at $45.1M in Q2 2026 and $2.0M in Q3.
- Most programmes are discretionary. GRVT's buybacks are "entirely at the issuer's discretion", Paradex uses net revenue for buybacks "and/or" ecosystem growth, GMX gated its staker share behind a $90 token price.
- For a trader the buyback changes nothing about execution. What matters is whether staking the token lowers your fees, whether a protocol fee taken from LPs thins liquidity, and whether you want token exposure at all.
The four mechanisms, and what each actually delivers
Buy-and-burn. Fees buy the token on the open market and the tokens are destroyed or sent to an address nobody controls. Hyperliquid's documentation now states that "HYPE in the assistance fund is burned", after a December 2025 validator vote; Uniswap's TokenJar and Firepit contracts, edgeX, Lighter since mid-2026 and PancakeSwap's weekly CAKE burn work the same way. The holder receives nothing; the claim is that a smaller supply makes each remaining token worth more.
Buy-and-hold. Jupiter locks purchased JUP for three years in the Litterbox Trust, Raydium holds bought-back RAY at a public address with no release schedule, dYdX stakes its purchases to validators. A future vote can release them.
Buy-and-distribute. Aster adds every day's buyback to the veASTER Loyalty Rewards pool; GMX's Solana proposal sends purchased GMX to the staking rewards pool. This is the closest to a dividend, paid in the protocol's own token.
Direct fee sharing by vote. Curve pays 50% of pool fees to veCRV lockers weekly in crvUSD. Aero, the merger of Aerodrome and Velodrome that went live on 22 October 2026, pays sAERO stakers a proportional share of the revenue of the pools they allocate to; before the merger, veAERO and veVELO voters received 100% of the previous epoch's fees from the pools they voted for (our report). Nothing is bought back; the fee goes to the locker instead of the LP or the treasury (see the governance tokens lesson).
Venue by venue: mechanism, share, start date, published figures
Built from each venue's documentation and governance record as of 28 October 2026. "Published?" asks whether the venue itself exposes an address or dashboard that can be checked, not whether a tracker estimates a figure.
| Venue | Token | Mechanism | Share of fees | Since | Published? |
|---|---|---|---|---|---|
| Hyperliquid | HYPE | Assistance Fund converts fees to HYPE automatically in L1 execution; holdings treated as burned | Not stated in docs; trackers quote 93–99% of perp fees | Late 2024; burn status Dec 2025 | Yes: system address 0xfefe…fefe |
| Aster | ASTER | Daily TWAP buyback paid to veASTER lockers; matched burn from reserve every two weeks until 3B supply | 99% of daily platform fees, plus 50,000 USDT per spot listing | 17 June 2026 | Yes: buyback wallet 0xa0ed…1b60 |
| Lighter | LIT | Daily 24-hour TWAP buybacks from trading fee revenue; bought LIT burned since June 2026 | Not stated | January 2026 | Partly: treasury account in the explorer, no share |
| Paradex | DIME | Periodic (e.g. daily TWAP) buybacks from net revenue "and/or" ecosystem growth; held in the DIME Fund | Not stated; split is discretionary | March 2026 | Partly: DIME Fund page in the app |
| edgeX | EDGE | Buyback and burn from revenue | Not stated ("accumulative revenue") | April 2026 | Yes: tokenomics page shows 5.05% of supply burned |
| GRVT | GRVT | Monthly open-market buy-and-burn by the Foundation's token issuer, at its discretion | Not tied to fees; funding source not stated | 20 October 2026 (1.1M GRVT) | Dashboard announced; we could not open it |
| Jupiter | JUP | Hourly buybacks locked three years in the Litterbox Trust | 50% of protocol revenue (70% proposed, not voted) | 17 February 2025 | Yes: trust wallet on-chain |
| Raydium | RAY | Automatic buyback once fees reach $10; held at a public address, no release schedule | 12% of trading fees (84/12/4 on CLMM and CPMM, 88/12 on AMM v4) | Long-standing | Yes: addresses in docs |
| PancakeSwap | CAKE | Non-stablecoin protocol fees converted to CAKE and burned weekly; stablecoin fees kept in treasury since 2 March 2026 | Varies by pool; after The Flip 12% of v3 and Infinity fees to protocol, 88% to LPs | Burns since 2021; Flip rollout from November 2026 | Yes: weekly burn posts |
| Uniswap | UNI | Protocol fee to TokenJar, withdrawable only by burning UNI in Firepit; Unichain sequencer and aggregator-hook fees feed the same burn | v2: 0.05 of the 0.30% fee; v3: one quarter of LP fees on 0.01% and 0.05% pools, one sixth on 0.30% and 1% pools | December 2025 (vote 25 Dec; 100M UNI retroactive burn) | Yes: on-chain contracts |
| GMX | GMX | Arbitrum/Avalanche: LPs get 63%; staker share redirected to treasury buybacks since March 2026, payout gated behind a $90 price (secondary reports). Solana: revenue left after GT buybacks buys GMX for the staking pool | 27% staker share (not in current docs); Solana initially the 1% LP share | March 2026; Solana from 28 October 2026 | Partly: forum posts, no vote record for GMLabs-003 |
| dYdX | DYDX | Open-market buybacks, tokens staked to validators | 75% of net protocol fees (12.5% → 25% → 75%); 5% Treasury SubDAO, 5% MegaVault | 23 April 2025; 75% since 13 November 2025 | Yes: 8.46M DYDX bought by 1 January 2026 (annual report) |
| Velocity (ex-Drift) | DRIFT | No active programme. DIP-4 was a one-off ($1.1M, 2.55M DRIFT); DIP-8/9 drafts proposed routing 80% of fees to a buyback module | None | — | n/a |
| Aero (ex-Aerodrome and Velodrome) | AERO | sAERO stakers allocate continuously to pools and receive a proportional share of those pools' revenue (47-hour cooldown); replaces the weekly veNFT vote, under which voters got 100% of the fees of pools they voted for | "100% of the revenue it earns goes back to the people who stake AERO" (Dromos CEO to The Block); only stakers, not all holders | 22 October 2026 (ve-vote since 2022/2023) | Yes: on-chain per pool |
| Curve | CRV | Admin fee converted to crvUSD and distributed weekly to veCRV; plus a share of crvUSD interest | 50% of pool fees | September 2020 | Yes: FeeDistributor contract |
Compiled 28 October 2026 from venue documentation and governance records listed under Sources. "Not stated" means the venue publishes no percentage.
Two details are easy to misread. Hyperliquid's fees page says fees go "entirely" to the community (HLP, the Assistance Fund and asset deployers) but never gives the split; the 93%, 97% and 99% figures in circulation are tracker estimates. PancakeSwap's 88/12 applies to v3 and Infinity pools only; v2 fees are hard-coded and the CAKE burn there continues unchanged, as our report on the vote explains.
Why volume decides everything
A buyback is a percentage of fees, and fees are a percentage of volume. DefiLlama's holder-revenue series, which books fees routed to token holders, makes the dependence visible. Hyperliquid's Assistance Fund received $289.8M in Q3 2025, $148.6M in Q2 2026 and $174.0M in Q3; $1.37B in total by 27 October 2026. edgeX booked $45.1M in Q2 2026, the quarter its programme started, and $2.0M in Q3. dYdX fell from $4.9M in Q3 2025 to $0.8M in Q3 2026 despite tripling its share, because net fees shrank faster than the percentage grew.
The table ranks perp venues by volume. A venue near the top with a 99% share returns real money; one near the bottom with the same percentage returns a rounding error. Volume is also the number most easily inflated by incentives, so a buyback funded by points-driven churn can stop with the points (how to read volume).
| # | Exchange | 24h volume | Open interest | Taker fee | Max lev | Score |
|---|---|---|---|---|---|---|
| 1 | Hyperliquid Hyperliquid L1 | $2.08B | $10.59B | 0.045% | 40× | 90 |
| 2 | Lighter zkLighter (Ethereum L2) · Robinhood Chain | $1.27B | $1.10B | 0% | 50× | 88 |
| 3 | Variational Omni Arbitrum | $559.0M | $914.9M | 0% | 50× | 65 |
| 4 | GRVT GRVT Chain (ZKsync validium) | $520.2M | $411.4M | 0.045% | 50× | 83 |
| 5 | Aster Aster Chain · BNB Chain | $467.7M | $1.41B | 0.04% | 1001× | 78 |
| 6 | trade.xyz Hyperliquid L1 | $337.4M | $4.08B | 0.09% | 50× | 56 |
| 7 | Pacifica Solana | $333.9M | $110.3M | 0.04% | 50× | 75 |
| 8 | ApeX Protocol Arbitrum · BNB Chain | $314.9M | $71.5M | 0.05% | 50× | 80 |
| 9 | SoDEX ValueChain | $214.2M | $153.7M | 0.04% | 40× | 72 |
| 10 | edgeX EDGE Chain · Arc | $199.4M | $134.0M | 0.038% | 100× | 82 |
Snapshot Nov 1, 2026. n/a = no verifiable public source. Score = dexwatch score (methodology).
Spot AMMs face a second drag. A protocol fee carved out of the LP share, as at Uniswap, means LPs earn less per dollar of volume; if liquidity leaves, volume follows and the fee base shrinks. The Flip is PancakeSwap's bet that the opposite also holds: pay LPs 88% and the pools deepen.
Discretion: what the small print says
Almost none of these programmes is a contractual obligation. The wording ranges from firm to vague:
- Fixed by code or governance: Raydium's 12% is a parameter in every pool, Curve's 50% admin fee is on-chain, Uniswap's protocol fee can only leave TokenJar through a burn. Changing them takes a vote.
- Fixed by policy, changeable by the team: Aster's 99% and Jupiter's 50% are documented but not enforced by contract; Jupiter's forum has carried a 70% proposal since May 2026 without a vote. Hyperliquid's conversion is automatic, but the split between HLP and the fund is set by the protocol and has moved before.
- Explicitly discretionary: Paradex uses net revenue for buybacks "and/or" ecosystem growth, with no stated ratio and only $0.28M of holder revenue booked by DefiLlama since launch. Lighter's buybacks run "depending on market conditions" with no documented share. GRVT's October release says each purchase is "entirely at the issuer's discretion, with no predetermined purchase amount or guaranteed schedule". GMX's March 2026 plan redirected the staker share to the treasury and releases it only if GMX trades above $90; the October Solana proposal was published for comment with no vote.
- Not happening: Velocity, the former Drift, has drafts on its forum and no programme; since the April 2026 exploit its protocol revenue is a few thousand dollars a month.
A documented percentage is intent; the on-chain record is what was done. Where the two cannot be compared, as at Lighter, Paradex and GRVT, treat the buyback as a marketing statement until the numbers appear.
Buyback is not dividend, and the securities question
A dividend puts cash in your account. A burn reduces supply and leaves you holding the same token, whose price still depends on what buyers will pay. A buy-and-hold programme does not even reduce supply permanently: Jupiter's Litterbox unlocks after three years and Raydium's holding address has no release rule, so both are reserves a future vote can spend. Only Aster, GMX's Solana pool, Curve and Aero hand tokens or fees to a holder, and in every case to a locker or staker. The unstaked holder gets the burn argument and nothing else.
That distinction is why the regulatory angle still matters. Uniswap's fee switch stalled for four years, with the Foundation citing a stakeholder issue in May 2024 and coverage pointing to the concern that paying revenue to token holders looks like a dividend from a security. The climate changed in 2025: the SEC chairman's November 2025 taxonomy speech placed "network tokens" outside securities law while stating that tokens which "behave like digital shares or bonds will continue to be regulated as securities". The speech is not a rule and no court has tested where a 99% fee-to-locker programme falls; the venues that pay most directly (Aster, Curve, Aero) look most like a profit share, and burn-only designs such as Hyperliquid's and Uniswap's are built, in part, to avoid that comparison (see the regulation lesson).
Tax follows the same line: a fee distribution in crvUSD or pool revenue is income on receipt in most jurisdictions, a burn is not a taxable event for the holder (taxes lesson).
What it means for traders
Your fee is unchanged. No buyback lowers the taker fee; Uniswap's protocol fee comes out of the LP's share and PancakeSwap's swap tiers stay the same after The Flip. Compare venues on the fee guide and the fee calculator, not on tokenomics.
Staking discounts are the one direct benefit. Hyperliquid cuts fees by 5% to 40% by staked HYPE; Paradex gives up to 30% for staking plus a flat 40% for paying fees in DIME; Lighter's Premium accounts get discounts for staking LIT; Aster's lockers receive the buyback itself. For a high-volume trader this can outweigh any price effect, but it ties capital to the venue's token.
Liquidity may shift. Protocol fees taken from LPs can thin pools; higher LP shares can deepen them. Watch price impact on the pairs you trade after each change, with the quote comparison for spot and open interest on the ranking for perps.
Separate the two decisions. Trading on a venue and holding its token are different bets. A trader on Hyperliquid benefits from the fee tiers and liquidity whether or not the Assistance Fund buys another HYPE. A holder is exposed to volume, to the team's discretion and to an unlock schedule no buyback offsets; the token due-diligence lesson lists what to check before taking that second bet.
Checklist. Read the venue's own page; find the address or dashboard; compare the documented share with DefiLlama's holder revenue; note who gets paid; ask what happens when volume halves.
Frequently asked questions
Which DEXes buy back their token in October 2026?
What share of fees goes to the token?
Do token holders actually receive money?
Are buybacks guaranteed?
Does a fee switch or buyback change my trading fees?
Rankings on this page are computed from the dataset behind the whole site: volume and open interest from the venues' own market APIs and public trackers, TVL from DefiLlama, fees and leverage from official documentation. Affiliate relationships never change an order. Anything we could not verify is shown as n/a rather than estimated. Read the full methodology.
Sources
- Hyperliquid documentation: Fees (fee destinations, Assistance Fund)
- Cointelegraph: Hyperliquid validators vote on Assistance Fund supply status (17 December 2025)
- Aster documentation: Tokenomics (Protocol Revenue Buyback Initiative, 17 June 2026)
- Lighter documentation: LIT Utility (Buybacks)
- The Block: Lighter's LIT rises as protocol indicates active buyback (5 January 2026)
- Paradex documentation: DIME Utility
- edgeX: EDGE Tokenomics and Buyback Program
- Chainwire via The Block: GRVT token issuer burns 1.1 million GRVT in first buyback (20 October 2026)
- SolanaFloor: JUP buybacks begin 17 February (2025)
- Jupiter governance forum: Increase the Litterbox buyback allocation from 50% to 70% (May 2026)
- Raydium documentation: RAY Buybacks
- PancakeSwap forum: CAKE Tokenomics Proposal 3.0 (April 2025)
- Crypto Briefing: PancakeSwap keeps stablecoin fees in treasury (implemented 2 March 2026)
- Uniswap blog: UNIfication (10 November 2025)
- The Block: Uniswap governance passes UNIfication proposal (25 December 2025)
- GMX governance: GMLabs-003 Solana Revenue Buybacks and 180-Day Unstaking Timelock (21 October 2026)
- dYdX: Annual Report 2025 (buyback programme)
- Drift governance forum: Draft, Programmatic Protocol Fee Re-Routing (October 2025)
- Blockworks Token Transparency: Velodrome Finance filing (pre-merger fee distribution to veVELO voters)
- The Block: Aerodrome and Velodrome merge into Aero (21 October 2026)
- Curve documentation: Fee Collection, Burning, and Distribution
- DL News: Uniswap token drops amid delay on fee switch vote (31 May 2024)
- Reed Smith: The SEC's approach to digital assets, inside Project Crypto (November 2025)
- DefiLlama API: Hyperliquid holders revenue series (quarterly figures in this guide)
Written and reviewed by the dexwatch editorial team. Some outbound "Trade" links are affiliate links; dexwatch may earn a commission at no extra cost to you. Educational content, not financial advice. Spotted an error? Tell us.