Fee Routing

Fee routing sits between a protocol earning money and a token holder seeing any of it. Two variables define it: the share and the destination. The share is easy to misread, because fees and revenue are not the same pool. Fees are the total users pay; revenue is the slice the protocol retains after liquidity providers and other suppliers are paid, and revenue is the pool a buyback can actually draw from. A program described as routing a share of fees, a share of net fees, a share of revenue, or a share of company profits is describing four different bases. Bitget's BGB burn is funded by 20% of exchange profits; Bitfinex commits at least 27% of iFinex consolidated gross revenue to buying and burning LEO; Raydium routes 12% of all trading fees into automatic RAY buybacks and then holds the tokens rather than burning them; Ether.fi funds weekly buybacks with 100% of one fee line, eETH withdrawal-fee revenue. The percentages are not comparable until you know what each is a percentage of.

The destination decides what the flow actually does to supply, and identical headline percentages diverge completely here. A buyback that ends in a burn retires supply permanently. A buyback that holds removes tokens from the float but leaves them outstanding and recoverable. A buyback that redistributes hands the purchased tokens straight back to stakers, so circulation is unchanged. Aave bought AAVE with real protocol revenue and sent it to the ecosystem reserve for redistribution rather than to a burn address, though the programme has been paused since May 2026, after the budget was cut from about $50M to about $30M in March 2026, and Aavenomics 3.0 leaves the destination unspecified. Pendle takes the opposite route on purpose: the roughly 80% share to lockers dates from September 2025, when vePENDLE lockers earned it as a direct fee distribution, and what changed on 2026-01-20 was the mechanism rather than the share, with sPENDLE routing that revenue through a buyback-then-distribute step under a new lock structure. It still never burns. Value reaches holders as cash-flow yield instead of scarcity. Neither design is wrong, but only one of them changes the supply number.

The share is a decision, which means the holder-facing bid can move for reasons that have nothing to do with business performance. Pump.fun originally routed 100% of net protocol fees into PUMP buybacks. In late April 2026 it burned everything it had bought and cut the policy to 50% of net revenue, keeping the remainder for growth, hiring, marketing and acquisitions. The routing change was not a revenue event: on calendar halves buyback spending fell about 16.5%, from $166.49M to $139.03M, while measured revenue fell about 18.0%, so what now caps the bid is the halved policy share rather than a collapse in earnings. The compression also began months before the policy was formalised. The routed share itself is disclosed, at 50% of net fees across Bonding Curve, PumpSwap and Terminal, hard-coded in an immutable contract since 2026-04-28 and committed for twelve months. What is not disclosed is how net is itemised, so what sits between measured revenue and the routed figure is still not visible from outside.

Reading routing well means separating three questions that headlines merge. First, how much: the share and its base. Second, where: whether the destination retires, parks, or recirculates the tokens. Third, whether it matters: routing sizes the bid, it does not decide net supply. Hyperliquid has routed about 99% of trading fees into continuous open-market HYPE buybacks through its Assistance Fund since 2025-08-30, when HLP's retained cut fell from 3% to 1%, and in December 2025 a governance vote closed the loop by formally recognising the roughly 37.51M HYPE in the fund's keyless address as burned. The burn is now documented protocol policy rather than an inference. That is close to the maximum routing a fee-funded program can achieve, and HYPE was still net inflationary because scheduled unlocks outran the burn. Uniswap turned its fee switch on with the UNIfication proposal executed December 28, 2025, but it never buys UNI: fees accumulate in TokenJar and are released only against UNI that third parties burn in a separate contract called Firepit. The burn stays small relative to total fees, about $31M against about $417M, roughly 7.5% capture, because most fees still go to liquidity providers. A routing policy is a claim about direction, not about magnitude.