How Long Do Meme Coins Last? We Measured 216,247 of Them. The Median Lasted Two Minutes.

The median meme coin launched on pump.fun in the first week of February 2026 was finished two minutes after it was created. Not down two minutes later. Finished, in the sense that it never traded again. Of the 216,247 tokens created that week, 867 were still being traded six months later, which is one in 249.

Those are measurements rather than estimates, and this article shows where every one of them came from. The “99% of meme coins go to zero” line gets repeated constantly and sourced almost never, which is a bad combination for a number people make decisions around. So instead of repeating it, we went and counted.

Nothing here predicts anything. It is a description of what happened to a large group of tokens that already lived and mostly died, and the useful part is the shape of the distribution rather than any single headline figure.

What we measured, and why a single week

Every token created on pump.fun leaves a permanent record of its own creation. So does every trade. That means the question “how long do meme coins last” has an actual answer sitting in public data, and it does not require anyone’s opinion.

We took every token created in the seven days from 2 to 8 February 2026. One full week, no filtering, no cherry picking the interesting ones. The week was chosen because it sits roughly six months before we ran the numbers on 20 August 2026, which gives the cohort enough time to have finished happening. Measuring last week’s launches would tell you almost nothing, because most of them have not had time to fail yet.

That week produced 216,247 tokens. It is worth pausing on that figure before going further. Not 216,247 in a year. In a week, on one launchpad. Roughly 31,000 new tokens a day, or about 21 a minute, every minute, for seven days.

Then we followed that exact cohort forward and asked three questions of each token. Did it ever trade at all. How long was it from creation to the last time anybody traded it. And was it still being traded in the week of 13 to 19 August 2026, six months on.

Of the 216,247, some 4,087 never recorded a single trade. They were created and then nothing happened, which is its own kind of answer.

The median token lasted two minutes

For the tokens that never made it off the launchpad, and that is almost all of them, the last trade is the end of the story. There is no second act somewhere else. So the gap between creation and final trade is the token’s whole life, and we can put a number on it.

Across the 210,213 tokens that traded at least once and never graduated, the median life was two minutes. Half of them were done faster than that.

The rest of the distribution is worth more than the median. Three quarters were finished inside 159 minutes, which is about two and a half hours. Nine in ten were finished inside 5.1 days. Even at the 99th percentile, the ones that hung on far longer than nearly all their peers, the figure is 26.2 days. That is the top one percent of survivors within the group, and it is still under a month.

Grouped a different way: 151,611 of them, or 72.1%, had their final trade within one hour of being created. 176,468, or 83.9%, were finished within a day. 192,944, or 91.8%, were finished within a week.

The thing that surprised us was not how many died. Everyone expects that part. It was the timescale. The popular mental image of a meme coin failing involves a slow bleed over weeks, a fading community, a chart drifting down. That does happen, and it is what most writing on the subject describes. It is also, numerically, a rounding error. The overwhelmingly typical outcome is that a token is created, trades for a couple of minutes, and is never touched again by anyone.

The first day does nearly all of the killing

Plotting the same cohort as a survival curve makes the shape obvious in a way the percentiles do not. Take the tokens that traded, and for each day after creation, count how many were still trading at that age or later.

On day zero, by definition, all of them. One day later, 16.1%. That single step removes almost 84% of the population, and everything after it is a long thin tail.

By day seven, 8.2% remain. By day fourteen, 3.5%. By day thirty, 0.52%, which is 1,087 tokens out of the 210,213 that ever traded.

The curve does not have a knee or a cliff further out. It has one enormous drop at the very start and then a slow grind. If you were looking for a moment when a token becomes safe, this data does not offer one. What it offers instead is a much less comfortable observation: nearly every meme coin that will ever fail has already failed before most people would have finished reading about it.

This is also why timeframes matter so much when anybody quotes a survival figure at you. “Most meme coins fail” is true at every horizon, but the number behind it swings wildly depending on whether you are asking about day one or month six, and people quoting it rarely say which.

Graduation is the line that actually divides them

Everything above treats the cohort as one population, and that turns out to be the wrong way to look at it. There is a hard structural line running through the middle of these tokens, and once you split on it the picture changes completely.

A token launched this way starts on what the launchpad calls a bonding curve, which is a self contained mechanism where the price rises automatically as people buy. It is not a normal market with buyers matched to sellers. It is a formula. If enough money goes in, the token “graduates”, and its accumulated liquidity gets moved out into an ordinary trading pool where it trades like anything else. We explained that machinery in detail in what a pump.fun bonding curve actually is.

Of the 216,247 tokens in the cohort, 1,947 graduated. That is 0.900%, or one in 111. The other 99.1% never got there.

Now compare survival across that line, six months on. Of the 1,947 graduates, 509 were still trading in the week of 13 to 19 August 2026, which is 26.1%. Of the 214,300 that never graduated, 358 were still trading, which is 0.167%.

Graduating was associated with roughly 156 times better odds of still being traded half a year later. That is not a subtle effect or a statistical artifact you have to squint at. It is the difference between a one in four chance and a one in six hundred chance.

Be careful about what that does and does not mean, because it is very easy to over read. It does not mean graduation causes survival, and it certainly does not mean a graduated token is safe. Nearly three quarters of the graduates were also gone. What it means is that the enormous headline failure rate is mostly made of tokens that never cleared the first hurdle at all, and lumping those together with tokens that did produces a number that describes neither group well.

Six months later, in one table

Putting the whole cohort in one place makes the scale legible in a way that prose does not. Every figure below describes the same 216,247 tokens, created in one week and checked again six months on.

One thing to watch as you read it. The middle rows are expressed as a share of the entire cohort, including the 4,087 tokens that never traded at all. Earlier in this article the lifespan percentages used a different denominator, the 210,213 tokens that traded and never graduated, because a token that never traded has no lifespan to measure and including it would quietly flatter the numbers. Both denominators are legitimate and they answer slightly different questions. That is exactly why survival statistics are so easy to quote misleadingly, and why the first thing worth asking anybody who tells you what share of meme coins fail is what they counted as the starting population.

The second thing worth watching is that the final two rows are not shares of the cohort at all. They split the survivors according to whether the token graduated, so their denominators are 1,947 and 214,300 rather than 216,247. That split is the actual finding of this study, and collapsing it into a single cohort-wide percentage is precisely how the interesting part normally gets lost.

Outcome	Tokens	Share of cohort
Created 2 to 8 February 2026	216,247	100%
Never traded even once	4,087	1.9%
Traded, then finished within one hour	151,611	70.1%
Traded, then finished within one day	176,468	81.6%
Graduated off the bonding curve	1,947	0.900%
Still trading anywhere, 13 to 19 August 2026	867	0.401%
Graduates still trading	509 of 1,947	26.1%
Non graduates still trading	358 of 214,300	0.167%

The two numbers most worth carrying away are in the last three rows. Across everything launched that week, one in 249 was still alive as a traded asset six months later. Restrict it to tokens that cleared the launchpad and the figure is better than one in four.

What this measurement cannot tell you

Four honest limits, and the third one is a mistake we made and had to fix.

First, the observation window for lifespans runs to 10 March 2026, so any token that kept trading past about thirty days is recorded as “still going” rather than given a final age. The day thirty figure of 0.52% is a floor, not a ceiling.

Second, “still trading” means at least one trade during the seven days from 13 to 19 August 2026. A token that trades once a month, or that has holders but no recent activity, reads as gone by this definition. It measures active trading, not existence. Every one of these tokens still exists, and always will.

Third, and this is the one that nearly produced a wrong article. Our first pass measured the last trade on the launchpad’s own records and called that the token’s lifespan. That is wrong for graduates, because a graduate leaves the launchpad and continues trading somewhere else entirely. The check that caught it was deliberately chosen: we ran the query against this site’s own token, which launched on 26 July 2026 and graduated eight minutes later, and the launchpad’s records showed its last trade on 26 July, the day it launched, despite it having traded every single day since. Had we published that first number, every successful token in the cohort would have been scored as dead within minutes of birth. The published figures use the launchpad records for tokens that never graduated, where the last trade genuinely is the end, and follow graduates across all trading venues instead.

Fourth, this is one week. February 2026 is not guaranteed to represent every week, and this counts tokens rather than money, so it says nothing about how much anyone gained or lost.

Where the token this site follows sits in all of this

This site is run by holders of one particular cat coin, so it would be dishonest to publish a survival study and quietly leave that token out of it.

$CATE was created on 26 July 2026 at 16:24:38 UTC and its trading pool was created at 16:32:43 UTC the same day, eight minutes and five seconds later. Both timestamps are read directly from the chain and from two independent trackers that agree to the second. So it graduated, which puts it in the 0.900% rather than the 99.1%, and it did so faster than almost anything does.

That is a fact about the past and it is worth exactly what a fact about the past is worth. It is not a forecast, it is not a reason to buy anything, and it does not exempt this token from the second half of the finding, which is that 73.9% of graduates in our cohort were gone within six months anyway. Clearing the first hurdle moved the odds a great deal and it did not remove them.

There is a related trap we have written about before. In why meme coins fall 80% and keep going we looked at this token’s own hourly record and concluded that surviving so far tells you nothing about surviving further. This study is the same lesson from the other end of the telescope. That article measured one token’s volatility over time. This one measures the base rate across a whole population. They agree, which is mildly reassuring, and neither of them lets anybody predict anything.

If you want to know what the token is doing right now rather than what it did in July, the live figures are on the stats page and they refresh hourly.

So how long do meme coins last

The honest answer has three parts, and only the first one is short.

Most of them last minutes. The median token in a full week of 216,247 launches was finished two minutes after it was created, and 72.1% of the ones that traded were finished within the hour. That is the real distribution, and it is far more brutal and far faster than the folklore version.

A very small number clear the launchpad. About one in 111, and that single structural step separated the population more cleanly than anything else we could measure.

And of those, roughly one in four was still trading six months later. Not most. Not none. Somewhere distinctly in between, which is the least quotable and most accurate thing this data says.

The reason to publish the whole distribution rather than the headline is that the headline is what people already believe. “Almost all of them fail” is not new information and it does not help anybody. Knowing that the failure is measured in minutes rather than months, that it happens overwhelmingly before graduation, and that the odds change by a factor of 156 at that one specific line, is a different and more useful shape of knowledge. All of the underlying queries and results are published in this site’s research folder so anybody can re-run them and disagree with us.