Invest1 publisher2 min readPublished
Hyperliquid collected a third of its lifetime revenue in the first nine months of 2026
Nine months of fees account for roughly a third of everything Hyperliquid has ever earned. Almost all of that money goes straight back out again into buying and burning its own token. The full-year total depends on which run rate holds.
The Investor · Invest desk
What happened
- Hyperliquid, a decentralized perpetual futures exchange running on its own Layer-1, took about $429 million in revenue between January and September 2026, ranking first among crypto projects on gcko.io's year-to-date income data.
- Its cumulative all-time revenue as of mid-September sat somewhere between $1.26 billion and $1.31 billion, the range the publisher gives for everything the venue has collected since inception.
- Roughly 99% of eligible perpetual futures trading fees are routed into what Hyperliquid calls the Assistance Fund, which spends the proceeds buying back and burning HYPE tokens.
- About 48.7 million HYPE, some 4.9% of total supply, had been removed from circulation as of September 2026.
- Recent weekly revenue ran around $13.5 million with daily peaks approaching $3 million, and the most recent 30-day total came in near $64 million.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint Essentially all fee income is spent on buybacks as it arrives, so the fund meets a downturn on what it earns that week. Crypto Briefing puts the consequence plainly: less cushion for adverse conditions.
- exposure HYPE holders are exposed to this week's volume, because the bid retiring their float is funded out of this week's fees rather than a standing balance.
- contradiction The publisher's standfirst sizes the burn at nearly half of HYPE supply while its body sizes it at 4.9%, a tenfold difference for anyone working out what the buyback has done to float.
- capability In-house market-making keeps outside liquidity providers from taking a share of fees, so more of each dollar of trading revenue reaches the buyback than a venue paying external LPs could manage.
Two ways of extending the year sit about $100 million apart. Nine months at $429 million averages $47.7 million a month [1][1]. The most recent week ran near $13.5 million, which over 52 weeks is $702 million [6][2]. That is where the trajectory above $700 million comes from [8]. Add 13 more weeks of $13.5 million to the $429 million already booked, though, and the year closes at $604.5 million [3], about 14% under the annualized figure [4]. The last 30 days were faster, at roughly $64 million, or $2.13 million a day [7][5]. Hold that through December and the full year lands near $625 million [6]. The weekly snapshot implies $1.93 million a day, about 11% below the 30-day pace, which is most of why the two extensions disagree [11].
The burn is easier to pin down. 48.7 million tokens at 4.9% of supply implies a total supply near 994 million [5][7]. Assume essentially all of the cumulative $1.26 billion to $1.31 billion passed through the Assistance Fund at the stated 99% share, and the fund has paid somewhere between $25.60 and $26.60 for each HYPE it retired [3][4][8]. The figure averages the whole history, and it holds only if older revenue was routed the same way as current revenue.
The comparison in the gcko.io ranking is protocol against protocol. What centralized venues earned over the same nine months sits outside it. Within Hyperliquid's own history the concentration is real: $429 million against $1.26 billion to $1.31 billion all-time is between 32.7% and 34% [9], leaving $831 million to $881 million booked before January [10].
Crypto Briefing wrote that the buyback "works elegantly when fee revenue is growing, but it offers limited buffer if volume drops sharply" [10], and that a platform burning its income has "less cushion for adverse conditions" [11]. Two design choices point the same way. Market-making is in-house, so no external liquidity providers take a cut of the fees, and development is funded separately from the revenue that flows to buybacks [9].
On this evidence I would expect the full year to print between $600 million and $630 million [3][6], and the $700 million number to stand as an artifact of annualizing a single week [8][2]. A strong December would break that view. The record here is a handful of weekly and 30-day snapshots. Peak days approaching $3 million against an average day near $1.93 million mean a month's total can turn on a few sessions [6][12].
What to watch
- Whether the fourth quarter prints nearer $175 million (the weekly rate) or $196 million (the 30-day pace) settles whether the $700 million annualization was ever reachable.
- A sharp drawdown in perps volume would test a fund that spends what it takes in. The weekly revenue figure moves before the burn total does.
- Any change to the 99% fee split, or a decision to retain part of it, changes what the token is a claim on.