Top DeFi Protocols by Revenue: Who Keeps the Money?

A verified snapshot of protocol revenue shows Pump leading on-chain earners, but Hyperliquid's trailing week is running at nearly double its monthly pace.

Tether and Circle generate more revenue than any on-chain protocol, but the income comes from Treasury-bill yield on stablecoin reserves rather than from DeFi usage. Among protocols monetizing on-chain activity, Pump leads at $47.2M over the trailing 30 days, ahead of Hyperliquid at $39.7M.

Overall Crypto Revenue Ranking

Stablecoin issuers and burn-based chains appear here because DeFiLlama's revenue definition tracks anything an entity retains, regardless of business model. Their position at the top is accurate, but it measures the crypto economy broadly rather than DeFi usage specifically.

Margin equals 30-day protocol revenue divided by 30-day fees. Entries marked at approximately 100% (Tether, Canton, Tron, Grayscale) carry definitions where all tracked fees count as revenue, so the ratio is structural rather than a measure of operating efficiency.

Top On-Chain Protocols by Revenue

Removing issuers, burn-based chains and fund-style entities isolates protocols earning from on-chain user activity finds top by revenue include:

What Changed, and What Cannot Be Measured Yet

The table's sharpest reading sits in the top row:

• Hyperliquid's week ran hot: it earned 41% of its monthly fee total in seven days, a window covering 23% of the month.

• The growth is in trading, not price: those seven days carried 32% of perpetuals volume ($64.1B of $199.5B), so the fee base expanded. HYPE hit an all-time high of $82.43 on August 22, 2026, but a rising token price earns the protocol nothing on its own.

• The stablecoin issuers drifted the other way: their last seven days ran just under the monthly pace, which is what happens when the money comes from interest on reserves rather than from people trading.

Why Tether and Circle Dominate

Both issuers earn the same way. Reserves back the stablecoins in circulation, those reserves sit in Treasury bills and similar short-dated instruments, and the interest goes to the issuer. The tracked Tether definition names Treasury bills alongside repos, commercial paper, money market funds and secured loans. 

Circle's figure adds CCTP transfer fees on top of reserve interest.

Three consequences follow for anyone reading the ranking:

• The revenue is recurring but rate-dependent: revenue persists through quiet markets, and revenue compresses when policy rates fall, regardless of on-chain conditions.

• Float size is the second lever: revenue scales with outstanding supply, so redemptions reduce it even at constant rates.

• Token holders get paid nothing: neither company has issued a token, so the money goes to company owners.

Why Pump Currently Leads On-Chain Protocols, and Why the Lead Is Softer Than It Looks

Pump collects fees at several points: bonding-curve trades and graduation fees on pump.fun, a 0.05% protocol fee on PumpSwap, and terminal fees kept after cashback and referral payouts. Users paid $115.4M over thirty days. Pump kept $47.2M of it.

Where the other $68.2M went explains the 40.9% margin, the lowest of any leader in the on-chain table:

• Creator fees take the largest share: roughly $60.6M of quarter-to-date gross revenue routes to the people launching tokens.

• Liquidity providers take the next: a further $25.7M, per the quarterly income statement on the protocol page.

• The remainder is cyclical by design: the revenue depends on token launches and speculative trading, which historically move in waves rather than a straight line.

Hyperliquid is built the other way around. Users paid $54.3M, and the protocol kept $39.7M: a 73.2% margin. Less money at the top, more of it retained.

Readers who have seen the 99% figure in tokenomics coverage will notice the gap. Both numbers are correct, and they measure different steps:

• The 73.2% measures what becomes protocol money: builder-code fees, maker rebates and HIP-3 deployer fees are paid out first, and the Hyperliquidity Provider vault (HLP) generates fees belonging to its depositors rather than to the protocol.

• The 99% measures what happens next: of the fees that qualify, 99% routes to the Assistance Fund to buy HYPE.

• The contrast lives in the second step: Hyperliquid passes almost everything through, while Pump keeps under half and then spends $22.4M of that on PUMP buybacks.

Pump earns the larger gross figure. Hyperliquid returns more of what it earns.

Fees Versus Revenue

Fees are the gross amount users pay. Revenue is the portion retained after paying liquidity providers, creators, referrers, market makers or other counterparties.

Uniswap shows the gap at its widest. Users paid $80.8M in fees over thirty days, the third-largest fee total in the dataset, while protocol revenue reads $6.56M: a margin of 8.1%.

• The split is structural, not weak usage: most swap fees still go to liquidity providers, exactly as the design intends.

• Only the fee-switch portion reaches the protocol: it was enabled progressively across chains between December 2025 and July 2026.

• The ranking column decides the story: ranking by fees places Uniswap fourth overall, while ranking by revenue drops Uniswap far below protocols with a fraction of the trading volume.

Does Protocol Revenue Reach Token Holders?

Protocol revenue and token-holder revenue are separate measurements:

* Canton and Tron are chain-grain entities: the figure measures fees burned at network level. Every other row is protocol-grain. The two are not directly comparable despite sharing a column heading.

A buyback is indirect value capture. It removes supply and directs cash flow toward the token, but it confers no dividend, no enforceable claim and no guarantee of continuation.

Risks and Limitations

• The same number can be read differently minutes apart: pages refresh on separate schedules, so Pump showed $46.53M in one place and $47.20M in another. Both readings are in the CSV, and neither was averaged.

• Uniswap's momentum reading is the shakiest number here: a later check returned a higher seven-day figure, which would push it well above the +7.8% shown. It was left as captured rather than patched, so the table still reconciles against the data file.

• The rules for counting fees changed mid-period: Uniswap's fee switch rolled out across chains between December 2025 and July 2026, so comparisons crossing those dates measure different things on each side.

• A thirty-day window is slow to react: a fast move takes weeks to show up fully, which is why the Hyperliquid seven-day figure tells a different story than the monthly one.

• Three entries do not disclose where the money goes: GMGN, Axiom Pro and Polymarket record revenue with no tracked payout, so the tables say "not independently verified" rather than guessing at a treasury.

• These are not the same kind of business: interest on reserves, fund management fees and on-chain trading fees sit under one column heading without being comparable.

Conclusion

One revenue table held four different businesses. Separating them changed the answer twice over: Tether and Circle earn from interest on reserves rather than from anyone using DeFi, and among the protocols that do earn from usage, Pump collects the most while Hyperliquid keeps and returns the most.

Three numbers will decide whether that order still holds in Coinjuice’s next Top DeFi Protocols by Revenue market analysis:

• Whether Hyperliquid's hot week was a blip: its last seven days earned at nearly double its monthly pace, and that alone is why the top of the on-chain table looks unsettled.

• Whether Pump keeps buying back less each quarter: $83.0M, then $56.0M, then $33.0M so far this quarter. Either the slide continues or it turns.

• Whether Uniswap's fee switch keeps growing: it has the widest gap between what users pay and what the protocol keeps, and that gap moves when the rules change rather than when trading picks up.

A ranking answers who earned the most. The better question is who kept it, and who passed it on. Readers working through a trading protocol on their own will find the framework laid out in the Coinjuice ebook.