Bundles: The Question Isn't Whether They Bundled, It's Whether They Can Get Out

Eighty-five percent of the memecoin launches we scanned had a coordinated group buying at the open. The ones that ran and the ones that died look the same at launch. The difference shows up in whether that group can already sell.

By JackedBasedDev. Measured October 9, 2026.

---

You open a new token, the chart is twenty minutes old, and the checker says "bundled: 62%." Most people read that as a verdict. It is not. In our scans of 18,211 Solana launches over the last 90 days, 85 percent had a coordinated cluster buying in the first slot, and 62 percent had a cluster holding more than half the supply. If "bundled" meant "walk away," there would be nothing left to trade.

What a bundle is, and why nearly every launch has one

On Solana a bundle is a Jito bundle: up to five transactions that land in the same slot, in order, all or nothing, for a tip to a validator that is usually a fraction of a cent. A creator puts the token creation and a set of buys from their own wallets into one bundle, so the first buys happen before anyone can see the token exists. That is a dev bundle. A sniper bundle is the same trick run by a stranger whose bot out-tips the creator to land right behind them. Pine Analytics found that more than half of pump.fun tokens are now bought in the block they are created in, and that snipers funded by the deployer exit fast: 55 percent fully out within a minute, 85 percent within five.

It is everywhere because it is cheap. A pump.fun curve starts at 30 virtual SOL against 1.073 billion virtual tokens, and the price of a slice is a fixed formula: SOL in = 30 × T ÷ (1.073B − T), where T is the tokens you take.

Ten percent of the supply costs 3.1 SOL, about $340. Twenty percent, 6.9 SOL, $757. Forty percent, 18 SOL, $1,965. Sixty percent, 38.4 SOL, $4,194. The whole curve, the 79.3 percent that is for sale, costs 85.9 SOL, about $9,369. Across 14,830 bundled scans the median cluster spent 10.1 SOL, about $1,100. A "$40,000 bundle" cannot exist on the curve; there is nothing left to buy after $9,400.

The exit arithmetic: paper value is not money

When the curve fills, the token graduates and its liquidity moves into a pool: about 85 SOL on one side, 206.9 million tokens on the other, a market cap of roughly $45,000 at today's SOL price (the "$69K graduation" people still quote dates from when SOL was $168). A pool can only ever pay out the SOL sitting in it, whatever the market cap says.

What a sale returns is proceeds = √(k·p) × S ÷ (√(k/p) + S), less the fee, where S is the tokens sold, p the price and k the pool constant. The part to remember is the square root: four times the market cap is roughly twice the money out.

A cluster that bought 40 percent of the supply for 18 SOL and dumps all 400 million tokens into a fresh graduation pool gets back 56 SOL and takes the price down 88 percent. Let the token run to a $1 million cap instead:

A 40 percent bag is "worth" $400,000 at a $1 million cap and returns $38,985 if sold. A 60 percent bag is "worth" $600,000 and returns $40,311. The pool has about 85 SOL plus whatever buyers added, and that is the ceiling.

So, the question everyone asks: you spent $40,000 on a bundle, can you sell at a $100,000 market cap? No, and the reasons stack. You cannot spend $40,000 on the curve. Spent in the pool at graduation, $40,000 buys 16.8 percent of the supply and lifts the cap to $1.24 million by itself, so "a $100,000 cap with $40,000 in" only exists after the price has collapsed. Selling there returns about $7,488, nineteen cents on the dollar, because the pool holds about 127 SOL at that cap. Getting the $40,000 back needs the cap to hold at about $1.29 million on other people's money; doubling it needs about $5 million. At $10 million the bag sells for $126,335; at $100 million, $421,514. Selling in pieces only helps if other people buy between the pieces. The receipts exist: WAP launched with 60 percent of supply in a bundle, and Bubblemaps put the realized take at about $250,000 against a "$310,000" pool.

Bundles that ran, bundles that died, side by side

From our scans, the biggest bundled runners and the clearest collapses, all tokens with thousands of holders:

STONK, 15.5 percent coordinated at launch, peaked at $225.6 million and trades near $120 million with 107,039 holders. ANSEM, launched by a bundler group we have tracked across 285 launches, had 85.8 percent of supply in 64 coordinated wallets for 86.1 SOL; it peaked at $216.5 million and trades near $142 million with 150,201 holders. apeonfone, 63.6 percent, peaked at $34.2 million. CLAW, 78.7 percent, 58 SOL, $15.3 million.

On the other side: KET, 79.6 percent, peaked at $17.4 million and is down 85 percent. The Toad Pepe, 86.3 percent, 67.7 SOL in, peaked at $17.2 million and was 92 percent lower within about a day. GrokBot, 100 percent of the early supply in one cluster, $10.4 million peak, $9,614 now. creator capital, 97 percent, $10.1 million peak, $193,154 now.

The supply shares do not separate the two groups. Nobody could tell ANSEM from GrokBot at the open. What separates them is what arrived afterwards. In every runner the cluster sold into the run: ANSEM's dumped 100 percent, apeonfone's 95.9 percent, CLAW's 91.7 percent. They were not holding the price up; demand from everyone else was, and the cluster was exiting into it. The collapses are the same mechanic with nobody underneath.

The aggregate agrees. Among clusters holding 10 percent or more, those that dumped early reached 5x 5.7 percent of the time, those that dumped later 4.8 percent, and those that never sold 2.7 percent, with a median peak of $3,900, because a cluster that never sells is usually sitting in a token nobody bought. The bundle's behaviour follows demand; it does not create it. We looked for a documented case of a team bundling to block snipers, then vesting, with a good outcome, and found none. The 198 recurring wallet groups we track launched 1,144 tokens: 44 percent rugged, 1.7 percent ran.

The trapped read: when a bundle is a reason to look closer

Traders have passed around a thesis for years: a bundle found below a certain market cap is worth a look, because its owners need the token to run before selling returns what the bundle cost. We tested it on 30,549 Solana scans from March to October, using only information available at the moment of the scan and only price action after it.

Call the cluster's breakeven the market cap at which selling its whole bag returns what it paid, slippage included. The bundle is "trapped" if the cap at the scan is below that number, and "in profit" if the cluster could already sell at a profit.

Where it holds: tokens still on the bonding curve at a $5,000 to $10,000 cap. Trapped bundles there reached 5x 55 percent of the time (125 tokens), against 34 percent for same-cap tokens whose bundle was already in profit (1,052 tokens) and 37 percent for every token in that band regardless of bundle. The sharper version holds too: the bigger the run the cluster needed, the more often it got one. Needing under 1.5x, 44 percent hit 5x; needing 2x to 3x, 52 percent; needing 3x to 5x, 91 percent, though that cell is only 11 tokens. It still shows at $10,000 to $30,000 on the curve, 26.5 percent against 12 percent.

Where it fades: after graduation at $30,000 to $100,000 the two groups are the same, 9.5 against 8.1 percent.

Where it inverts: at $100,000 and up, a bundle still under water is a bad sign. Those tokens reached 5x 0.7 percent of the time, every one is dead, and the bundles in profit beside them did better at 4.3 percent. A cluster under water at a $300,000 cap with hundreds of SOL in it does not pump the token back; it walks away or sells into whatever bid exists.

Three caveats belong in the same breath. The bundle's cost does not matter on its own; across cost bands on the curve the 5x rate sits between 18 and 25 percent with no trend. Trapped bundles are rare where anyone is looking, because bundles are cheap: the median cluster cost 8.2 SOL against a median breakeven cap of $4,010, so 93 percent of trapped bundles were trapped for seconds, below a $5,000 cap. And nothing here is safe: between 96 and 100 percent of the tokens in every one of these groups eventually died. The edge is that the trapped ones tend to run first.

The honest version: a bundle is not the signal; whether the bundle can already get out is. Below its breakeven, the people who control the supply still need the token to run. Above it, the same bag is overhead that can leave at any time.

Three checks before you buy

1. How much supply did one group buy at the open? Bubblemaps, RugCheck and GMGN show a bundled or clustered share; the PumpPill scan shows it with the funding links between the wallets. Under 15 percent is ordinary. Over 50 percent means one group decides what happens next: those tokens reached 5x 5.4 percent of the time in our scans, against 11.9 percent when the cluster was under 10.

2. Has the cluster sold? The same tools show how much of its bag is gone. Fine: the token is running and the cluster is selling into it, which is what every big runner looked like. Walk away: the cluster is selling and nobody else is buying.

3. Where is the cluster's breakeven against the current cap? Most tools do not show this. Work it out from the cost and the pool, or read it off the PumpPill scan, which lists the cluster's breakeven market cap next to the current one. Below breakeven and still on the curve: the group needs a run, and in our data it got one more often than not. Above breakeven: assume the bag can leave at any moment and size for it. Under water at $100,000 or more: a bag being abandoned, not defended.

Methodology

• Bundle detection. An early buyer is coordinated if its funder funded two or more early buyers, it landed in the same slot as others, it sits inside a confirmed Jito bundle, the creator funded it, it shares an address lookup table with other buyers, or it received identical amounts from one funder. Coordinated supply is the share those wallets bought.
• Outcomes. Latest scan per contract; peak cap from DexScreener polling that starts after the scan; "dead" is the scanner's status. Multiples run from the cap at the first alert, so launch-to-peak multiples are understated.
• Exit math. pump.fun constants: 1 billion supply, 793.1 million on the curve, virtual reserves 1.073 billion tokens and 30 SOL, 1 percent fee; graduation at 85 real SOL; pool 84.985 SOL and 206.9 million tokens, 1.25 percent fee. SOL at $109.11 on 9 October 2026. Sales modelled as one swap with no other buyers.
• Trapped test. 30,549 scans, 4 March to 9 October 2026, each within ten minutes of the alert whose cap it uses, Solana only. Whether the cluster had sold before the scan is only recorded from July 2026, and the trapped-and-unsold cell at $5,000 to $10,000 holds eight tokens, so the headline compares trapped against in profit without the unsold condition.
• Limits. Wallet counts are addresses, not people. Holder counts checked on GMGN on 9 October; caps and holders move.

Sources

• PumpPill bundle scans and the fact pack for this article, with every number and the query behind it.
• Jito documentation on bundles and the tip floor; pump.fun fee documentation.
• Pine Analytics, "Exit Liquidity Machines" (April 2025).
• Bubblemaps on WAP and on bundle versus cluster; MadeOnSol, "Solana Bundling Explained."
• DL News on the JESSE launch on Base; Decrypt on GWOOD.
• GMGN token pages for holder counts and current market caps, 9 October 2026.