The Biggest Token Launchpad Made a Billion Dollars While 99.7% of Its Launches Died
A business that earns a billion dollars while 99.7% of its output fails is not a broken business. It is a business whose customers were never the projects.
Pump.fun crossed a billion dollars in lifetime revenue this year. Over the same stretch, its token graduation rate — the share of launches that accumulate enough buying to leave the bonding curve and reach an open market — collapsed to roughly 0.26% by mid-June, down about 80% in three months. Roughly one launch in four hundred made it out.
That gap is the whole story of token launches in 2026, and it is finally showing up in the revenue line.
The revenue collapse is not a memecoin story, it is a distribution story
Pump.fun's platform revenue averaged around $800,000 a day in June, against roughly $4.8 million a day six months earlier. Activity fell about 80% in a quarter. The knock-on effect reached the chain itself: Solana's daily network fees dropped from a January average near 33,000 SOL to about 5,300 SOL in June, a decline of roughly 84%.
Pump.fun alone had driven something like 36% of Solana's Q1 application revenue, out of $342.2 million across the whole chain. One app, one third of a major network's take.
Meanwhile the underlying category shrank hard. Total memecoin market capitalization sits near $30 billion, down more than $110 billion from the 2024 peak.
The instinctive read is that memecoins died and took their casino with them. That read is too easy. Memecoins have died four or five times now, always temporarily. What actually broke is subtler and more durable: the price the market is willing to pay for undifferentiated distribution.
Frictionless launch was the product, until frictionless launch was worthless
The founding insight of the modern launchpad was correct and genuinely useful. Launching a token used to require a developer, a liquidity budget, and a week. Compressing that to ninety seconds and a few dollars removed a real barrier, and the volume that followed proved the demand was there.
But distribution is a positional good. The value of being easy to launch depends entirely on launching being hard for someone else. Once everyone can do it in ninety seconds, the ninety seconds is worth nothing, and the scarce resource moves one layer up — to attention, then to credibility, then to whatever proves a team will still exist in six months.
Launchpads that never built anything above the ninety seconds got repriced accordingly. SunPump, TRON's equivalent, generated $33,220 in gross protocol revenue in Q2 2026 against $5.30 million in Q3 2024. That is not a downturn. That is a business evaporating.
The clearest signal came from the launchpad that quit
The most instructive event of the year happened in July, and it barely registered. Noxa, the launch platform that powered most of Robinhood Chain's memecoin activity, stopped accepting launches on July 11 and gave away its entire revenue — roughly $12 million in cumulative fees — before going dark two days later.
Operators do not walk away from a functioning $12 million business. They walk away from one they have concluded is finished.
Read that alongside Pump.fun's capital behavior. In the first half of 2026 the platform spent $72.2 million on buybacks, a 67% cut from the second half of 2025, while protocol revenue over the same period fell only 18%. Revenue held up four times better than the buyback held up. That is a team building a cash cushion, not a team that believes the next six months look like the last six.
The BOOST recovery proves the point rather than refuting it
Pump.fun did respond. Its BOOST feature pushed the graduation rate to 6.7% on one recent Friday — roughly eight times the June average — and the company says it reinjected more than $7 million into the market.
Which is a fix in the sense that adding subsidy to a market fixes the market. Buying graduations works while you are paying for them. The interesting question is what the graduation rate settles at once the subsidy normalizes, and whether 6.7% represents demand that was always there or demand that was rented.
I would not bet against Pump.fun as an operator. The company has been faster and more clear-eyed about its own decay than most of its critics have been. But the fix is aimed at a symptom. The disease is that a launch fee model earns identically whether a project ships or vanishes, and the market has now watched enough launches vanish to price that indifference in.
Incentive alignment is the only feature left that costs anything
Consider what a launchpad actually sells when creation is free. It sells the belief that this particular launch is different from the four hundred that failed last week. Everything that supports that belief is expensive by design — diligence on the team, vesting that lasts past the first liquidity event, liquidity that cannot be pulled on Tuesday, a supply schedule someone can verify without trusting a Discord announcement.
Each of those is friction. Each of them reduces launch volume. Which is precisely why the free-launch model could never offer them: its revenue was a direct function of the number it needed to suppress.
The infrastructure to do this properly is not exotic, and it has existed the entire time. Locking liquidity and team allocations through a service like Team Finance turns a promise into a public record with a timestamp. A vetted raise through a launchpad with structured vesting produces a token whose float is knowable at launch rather than discovered by the market three weeks later, usually on the way down.
None of that guarantees anything about price, and anyone selling it as such is selling something else. What it does is make the failure modes legible. A buyer who can see when supply unlocks and how much liquidity is committed is making a different decision than one who cannot — even if both decisions turn out badly.
The counterargument is real: most projects should fail
The strongest case against all of this is that a 0.26% graduation rate is not a bug. Venture capital funds a hundred companies expecting three to matter. If token launches are early-stage risk capital with worse disclosure and better liquidity, then a brutal failure rate is the honest baseline, and demanding otherwise imports an expectation the asset class never earned.
There is something to that. A launch venue is not obligated to underwrite outcomes it cannot control, and a lot of gatekeeping in this industry has been credentialism wearing a safety vest.
But venture capital does not charge the startup a fee at incorporation and then earn nothing further from whether it survives. The asymmetry in the launchpad model is not that most projects fail. It is that failure and success pay the platform the same, and the volume that maximizes revenue is the volume that maximizes noise. Any market where the intermediary is indifferent to outcomes eventually gets what it pays for.
The retail buyers absorbed that indifference for two years. Then they left, and the revenue charts show exactly when.
What survives is smaller, slower, and worth more per launch
The next version of this market will not look like a casino floor with a ticket booth. It will look more like underwriting — fewer launches, real preparation, terms that bind the team past the first month, and fees that scale with something other than raw creation count.
That is a much less exciting business. It is also the only one with a defensible reason to exist once anybody can mint a token from a phone in ninety seconds.
Pump.fun will probably be fine; a billion dollars of accumulated revenue buys a lot of time to figure out what comes next, and its team has earned the benefit of the doubt on execution. The launchpads that will not be fine are the dozens that copied the mechanic without the volume, and are currently discovering that a business built on charging for something free was never a business at all.
The number to watch is not daily revenue or memecoin market cap. It is what share of launches are still trading, with committed liquidity, a year after the raise. Nobody publishes that figure — which is a reasonable clue about how bad it is, and a reasonable bet on which platform decides to publish it first.