Pump.fun: The Machine That Turned Attention Into a Billion Dollars

A token-launch machine made more than a billion dollars in fees in under two years, while most of the people using it lost money. Both of those things are true, and the gap between them is the whole story.

Key facts

• Pump.fun is a Solana memecoin launchpad that lets anyone create a tradable token in seconds. It launched on 19 January 2024.
• It has earned roughly 1.1 billion dollars in cumulative fees and over 1 billion in revenue (DefiLlama, mid-2026), peaking near 138 million dollars of revenue in a single month in January 2025.
• By mid-2025 it was behind more than 80 percent of all new tokens on Solana.
• About 98.6 percent of its tokens collapse to near-zero, and for most of 2024 and 2025 the majority of traders lost money (Solidus Labs, CoinGecko).
• The related $TRUMP coin (18 January 2025) peaked near an 11 billion dollar market cap, with 80 percent of supply held by a Trump-affiliated entity.

In January 2024 a small team of English founders, operating through a company called Baton Corporation, put a website live that did one thing. It let anyone create a tradable crypto token in seconds, for a few cents, with no code. They called it pump.fun. By mid 2026 that website had taken in roughly 1.1 billion dollars in fees and over a billion in revenue, according to the data tracker DefiLlama. For a stretch of 2025 it was, by revenue, one of the most profitable consumer products in crypto history.

It is also the clearest example yet of a machine that manufactures attention, monetizes it, and leaves most of its users poorer. This is how it worked, why it worked, and the math that decides who paid for it.

What pump.fun actually is

Before pump.fun, launching a token took effort: writing a contract, seeding liquidity, paying for a listing. pump.fun deleted all of that. You type a name, upload an image, click once, and a token exists, trading immediately.

Each new token trades on something called a bonding curve. In plain terms, the price is set by a formula, not by a traditional order book: the more people buy, the higher the price climbs, automatically. If enough buying arrives, the token "graduates" to a public exchange and trades freely. If it does not, and the overwhelming majority do not, it simply dies on the curve. pump.fun took a cut of the trading on the way, reportedly around one percent of trades in its early period, and that small cut, multiplied across millions of tokens, became the billion.

The numbers that make people stare

Two honest notes on those figures. The fee and revenue numbers are live dashboard readings that drift every day, so treat them as mid 2026 snapshots, not fixed facts. And the business is past its peak: revenue fell roughly 80 percent from the January 2025 high as the mania cooled. A machine this good at capturing a frenzy is, by definition, exposed when the frenzy fades.

Why it worked, as a business

Strip away the noise and pump.fun is a near perfect digital business. The marginal cost of creating one more token is close to zero. It holds no inventory, takes no market risk on the tokens themselves, and earns a toll on every trade regardless of whether the token goes up or down. The house does not bet. The house charges admission.

The real product was never the tokens. It was friction removal plus attention. By making creation instant and free, pump.fun turned token launching from a project into a reflex, and the sheer volume of launches became its own marketing. At its height the platform was behind more than 80 percent of all new tokens on Solana, the blockchain it runs on. That is a network effect: people launched there because everyone traded there, and everyone traded there because everyone launched there.

The psychology: a casino dressed as a market

To understand why millions of people fed this machine, you have to look at behaviour, not charts. The pattern is old, and the literature on financial manias describes it precisely.

Charles Kindleberger, in Manias, Panics and Crashes, mapped the anatomy of every speculative episode: a displacement (here, a new technology that makes a new kind of bet possible), then euphoria (everyone is winning, or claims to be), then revulsion (the money leaves). pump.fun did not invent this cycle. It industrialized it, and ran thousands of tiny versions of it per day.

The fuel is a specific bias. Daniel Kahneman showed in Thinking Fast and Slow that humans systematically overweight tiny chances of enormous payoffs. It is why people buy lottery tickets with terrible odds, and it is exactly the shape of a memecoin: a few dollars in, a one in ten thousand chance of life changing returns. We also remember the winners and forget the dead tokens, the survivorship bias that makes the game look far more winnable than it is.

Howard Marks, in Mastering the Market Cycle, calls sentiment a pendulum that swings between greed and fear and spends almost no time at the sensible middle. And George Soros, in The Alchemy of Finance, named the engine underneath: reflexivity. A memecoin goes up because people believe it will, and their buying makes the belief true, until the belief breaks and the same loop runs in reverse. A pump.fun token is reflexivity with no fundamentals left to anchor it. It is pure belief, priced.

Who actually wins

This is where the romance ends and the arithmetic begins. A 2025 report from the compliance firm Solidus Labs found that 98.6 percent of pump.fun tokens collapsed into worthlessness shortly after launch, and that of more than seven million tokens with any real trading, only about 97,000, roughly 1.4 percent, ever held even a thousand dollars of liquidity. Read carefully: that number measures tokens that died, not fraud proven in court, and pump.fun disputes the "scam" framing, fairly noting that most memecoins dying is just normal. But the shape is not in dispute. Almost everything launched here goes to zero.

And the traders? CoinGecko's data shows that from early 2024 through late 2025, in most months the majority of active traders lost money, bottoming at just 30.1 percent profitable in June 2025. The share of winners did rebound to 73.28 percent by April 2026, but CoinGecko itself attributes much of that to survivorship, the unprofitable crowd simply leaving, and even among the winners most made between one and five hundred dollars. The study also flatters the picture, because it counts only realized profit and ignores the bagholders still sitting on tokens that went to zero. The honest summary: this was, for years, a game most players lost.

And these are not abstractions. Behind that 98.6 percent are real people who put real savings, sometimes everything they had, into the promise of a number going up. The wallets are a dataset. The lives behind them are not.

The macro backdrop

None of this happened in a vacuum. pump.fun launched into a wave of cheap liquidity and high risk appetite, and it rode a genuine mania. By CoinGecko's count, the total memecoin market hit an all time high of 150.6 billion dollars in December 2024, propelled by Trump's re election, dwarfing the 2021 peak of 88 billion. Then it deflated to about 47 billion by late 2025. The deeper point is that in a market with no cash flows to value, the scarce resource is not capital. It is attention. pump.fun was the most efficient attention to money converter anyone had yet built, which is also why its revenue tracked the mania down once attention moved on.

When the President launched a coin

The mania reached its logical conclusion on 18 January 2025, when the $TRUMP token went live on Solana. Within hours it touched roughly 75 dollars and a peak market value near 11 billion dollars. The structure is the lesson: per TRM Labs, only 20 percent of the supply was sold to the public, while 80 percent was held by a Trump affiliated entity on a multi year unlock schedule. CoinDesk later reported the Trump family had earned around 320 million dollars in fees even as the token fell 87 percent from its first day, and reporting found a tiny number of insider wallets made millions while hundreds of thousands of buyers lost money. ($TRUMP did not launch on pump.fun itself, but it is the genre's defining case.) It made the wealth transfer that is implicit in every memecoin completely explicit: the people who control the supply win, and they win from the people who provide the attention.

The critique, and the risks

The strongest case against pump.fun is simple: it is a negative sum game with a toll booth. Value is not created, it is transferred, from many late and hopeful buyers to a few early and informed sellers, with a one percent cut skimmed off the top each way. That has drawn the obvious legal heat, including a proposed class action alleging securities violations and a separate multi billion dollar suit branding the platform an "unlicensed casino." Regulatory classification remains unsettled, and the platform already excludes some jurisdictions.

The business risk is just as real. A product whose revenue is a function of mania is structurally fragile when mania fades, which is exactly what the 80 percent revenue decline shows. The company has since moved to prop up its own token with buybacks and supply burns, the behaviour of a business defending a peak, not extending one.

The Goldzweig take

The thesis. pump.fun is, at once, a genuinely brilliant business and a wealth extraction machine, and the discomfort most people feel comes from refusing to hold both ideas at the same time. It solved a real problem (friction) and built a real toll (the fee), and it did so by industrializing a casino. As an operating model it is close to flawless. As a place to put money, the published data says you were, on average, the product.

The counterargument. Defenders are right that nobody is forced to play, that most memecoins dying is normal rather than fraudulent, and that the same "it is all gambling" charge was once thrown at every new market. Entertainment has value, and some users treat this as exactly that.

The risk, and what we are watching. The open question is not whether pump.fun was a good business. It was. It is whether the model survives contact with a cooler market and a warmer regulator. We are watching three things: whether revenue stabilizes or keeps bleeding from the peak, whether the class actions force a securities classification, and whether attention, the only thing this machine truly runs on, comes back. Read the data, not the hype. On this one, the data is unusually loud.

Frequently asked questions

What is pump.fun? Pump.fun is a Solana based launchpad that lets anyone create and trade a memecoin in seconds for a small fee, using an automated bonding-curve price. It launched in January 2024.

How much money has pump.fun made? Roughly 1.1 billion dollars in cumulative fees and just over 1 billion in revenue as of mid-2026 (DefiLlama), though revenue is down about 80 percent from its January 2025 peak.

What percentage of pump.fun tokens fail? About 98.6 percent collapse to near-worthlessness shortly after launch, and of more than 7 million tokens traded, only around 97,000, about 1.4 percent, ever held more than 1,000 dollars of liquidity (Solidus Labs).

Do most people make money on pump.fun? No. For most months from 2024 through late 2025 the majority of active traders lost money, bottoming at 30.1 percent profitable in June 2025 (CoinGecko). A later rebound to 73.28 percent profitable is partly explained by unprofitable traders leaving.

Who profited from the Trump coin? A Trump affiliated entity held 80 percent of the $TRUMP supply, and the family reportedly earned around 320 million dollars in fees even as the token fell 87 percent from its first day, while a small number of insider wallets made millions and hundreds of thousands of buyers lost money (TRM Labs, CoinDesk, NBC News).

Sources. DefiLlama (fees, revenue, volume). CoinDesk and Fortune (revenue history, PUMP ICO, Trump family fees). Wikipedia (founding, milestones). CoinGecko Research (memecoin market cap, trader profitability). Solidus Labs (token survivorship). TRM Labs (TRUMP supply and structure). NBC News and The Block (TRUMP and Melania holder outcomes). Figures marked as snapshots are live and date-stamped to mid 2026. This is research, not financial advice.