A pump.fun Bonding Curve Never Gets Deleted. Here Is What $CATE’s Still Says.

A bonding curve is a pricing formula that quotes a buy and a sell price from nothing but its own reserves. On pump.fun it is also an account, and accounts on Solana do not get deleted when they stop being useful. So every coin that ever graduated left one behind.

That account is the most honest document a launchpad coin has. It records what the coin was worth on its first second, how much of the supply was ever for sale, and the exact moment trading was taken away from it. It is also readable by anyone, for free, in a single call.

What follows reads one of them. All chain figures were pulled on 11 August 2026 and every platform claim is checked against pump.fun’s own documentation rather than against the explainers currently ranking for it, because three of the most repeated numbers in those explainers are out of date.

What the curve actually does

Pump.fun’s own documentation describes it plainly: the bonding curve is “a constant-product AMM, similar to the math used by Uniswap and other automated market makers”, where “two virtual reserves (SOL and the coin’s supply) are multiplied together to form an invariant”.

Multiply the two reserves together, and the result must stay the same after every trade. Buy tokens and the token reserve falls, so the SOL reserve has to rise to keep the product constant, which is another way of saying the price went up. There is no orderbook and nobody quoting you a spread. The price is arithmetic.

The word doing the work is virtual. The curve does not begin with a real deposit of thirty SOL from anybody. It begins with a number written into an account that behaves as though thirty SOL were there. This is what solves the problem the docs call cold start: a brand new coin has no liquidity and no price, and pretending it has some is what makes it tradable from the first block instead of after a presale.

The consequence people underestimate is how back-loaded the shape is. Because price moves with the square of the shrinking token reserve, the early buys barely move it and the late ones move it enormously. Sell half of everything the curve will ever sell and the price has only reached about a sixth of where it finishes.

Every curve starts with the same four numbers

These are not folklore. They can be read off any coin that has just been created and not yet traded, and several such coins were checked on 11 August 2026 to confirm the values below are still current rather than inherited from an older version of the platform.

A fresh curve opens with a virtual reserve of 30 SOL and 1,073,000,000 virtual tokens. Their product, 32,190,000,000, is the invariant that every subsequent trade has to preserve. Divide one by the other and you get the opening price of any pump.fun coin, which is 0.00000002796 SOL per token, identical for all of them.

The fourth number is the one almost nobody quotes. Of the billion tokens minted, only 793,100,000 are ever available on the curve. The remaining 206,900,000, a fifth of the supply, is held back and becomes the token side of the liquidity pool at graduation. So the curve is not selling the coin. It is selling 79.31% of the coin and reserving the rest to seed the market that replaces it.

It is worth sitting with what that implies for the first person through the door. They are not buying from a seller, because there is no seller. They are not buying from the creator, because the creator holds nothing at that point. They are buying from a formula that will mint the tokens into existence at a price nobody negotiated, and the SOL they pay accumulates in the curve rather than going to anyone. That is the sense in which a pump.fun launch is genuinely fair at the start, and it is also why the fairness argument stops being interesting the moment the coin graduates and ordinary market structure takes over.

Graduation is priced in SOL, and the dollar figure moves

The threshold is usually quoted as a market cap of ninety thousand dollars. That figure is a snapshot, not a rule, and the distinction matters more than it sounds.

The curve completes when the last of those 793,100,000 tokens is sold. At that instant the virtual token reserve has fallen to 279,900,000, and the invariant forces the virtual SOL reserve to 115.0054 SOL. It started at 30. So completing a pump.fun bonding curve takes 85.0054 SOL, always, for every coin, regardless of what SOL is worth that day.

That derivation is not theoretical. A coin that graduated and was checked on 11 August 2026 showed a final virtual SOL reserve of 115,005,359,451 lamports. The arithmetic above predicts 115,005,359,057, so the two agree to within four ten-millionths of a single SOL.

The dollar headline follows from it. Graduation price works out to a market capitalisation of 410.88 SOL. At the SOL price on 11 August 2026, around $76 according to both CoinGecko and GeckoTerminal, that is roughly $31,000. The familiar $90,000 figure corresponds to SOL trading near $219, which is where it was when that number entered circulation. The threshold never changed. The currency it was quoted in did.

The practical consequence is that graduation gets cheaper or dearer in dollar terms without anybody deciding anything. When SOL falls, a coin needs less real money to complete its curve, so more coins graduate on less demand, and the graduation badge means slightly less than it did the month before. When SOL rises, the reverse. Any comparison of graduation rates across a long period is quietly measuring the SOL price alongside whatever it thinks it is measuring, which is a good reason to be careful with the statistics that circulate about what percentage of coins make it.

$CATE cleared the curve in eight minutes

Pump.fun records a creation timestamp for every coin, and DexScreener and GeckoTerminal independently record when the pool that replaced the curve came into existence. For this token the three line up to the second.

The coin was created at 16:24:38 UTC on 26 July 2026. Its liquidity pool was created at 16:32:43 UTC the same day. GeckoTerminal reports the identical pool creation time. That is a gap of 485 seconds, or eight minutes and five seconds, from a curve nobody had touched to a graduated coin trading on an AMM.

Eighty-five SOL in eight minutes is not remarkable in itself, and plenty of coins do it faster. The reason to state it precisely is that the whole of the curve phase, the part every explainer devotes most of its words to, lasted about as long as it takes to read this article. For a coin that catches, the bonding curve is not a period of price discovery. It is a formality that happens before most people have heard the name.

It also means something specific about the price history. Nothing on the curve is visible on a normal chart, because the chart belongs to the pool and the pool did not exist yet. The first candle anybody can see already starts at 14.7 times the opening price, and the run that got it there is only reconstructable from transaction logs.

The liquidity went to PumpSwap, not Raydium

Here is where the reference material and the chain disagree, and the chain is checkable in one call.

Pump.fun’s documentation now carries a section headed “Graduation to PumpSwap”, stating that once the threshold is hit “the curve is closed and the entire liquidity pool is migrated atomically to PumpSwap”. It adds that “Pump.fun does not seed or remove this liquidity after graduation. The pool is owned by the protocol.”

The account confirms it. This token’s pool is owned by program pAMMBay6oceH9fJKBRHGP5D4bD4sWpmSwMn52FMfXEA, which is PumpSwap. Raydium’s AMM program is 675kPX9MHTjS2zt1qfr1NYHuzeLXfQM9H24wFSUt1Mp8, and it owns the pools of older graduates but not this one.

The Wikipedia entry that ranks on the first page for how pump.fun works does not mention PumpSwap at all. It describes coins being “listed on DeFi platforms such as Raydium”, and states that “Pump earns a 1% swap fee on all tokens traded and 1.5 Solana tokens when a coin created on the site is listed, or graduates, after hitting a market cap of $90,000”. Every one of those figures was accurate once. The current documented numbers are a 1.25% curve fee and a graduation fee of 0.015 SOL, which is a hundredth of the figure still in circulation.

This is worth understanding as a pattern rather than as a complaint about one page. Launchpad mechanics change faster than the pages describing them, the descriptions get cited, and the citations outlive the facts. Reading the pool’s owner program takes one call and settles it for any coin.

The curve is still there, and it still says no

Every bonding curve lives at an address derived from the coin’s mint, so you can compute it without looking anything up. For this token that address is 3Sg92V4Mre9Apm7dJsM39B6vrAEVVErE1vBsZMKyUbxT, and it is still owned by the pump.fun program more than two weeks after the curve closed.

Read it and you get a set of zeros. Virtual token reserves, virtual SOL reserves, real token reserves, real SOL reserves: all four are 0. One flag, complete, is set to true. That combination is the difference between a live curve and a finished one, and it is the cleanest way to tell whether a coin has graduated without trusting any interface to tell you.

The interesting part is that people are still trading into it. On 11 August 2026 at 05:03 UTC, three transactions in the same three seconds tried to sell into this closed curve and all three were rejected by the program with a named error:

Program log: Instruction: SellV2
AnchorError thrown in programs/pump/src/sell_v2.rs:242.
Error Code: BondingCurveComplete. Error Number: 6005.
Program failed: custom program error: 0x1775

A second cluster failed the same way nineteen minutes earlier. These are almost certainly bots working from stale routing tables rather than people, but the effect is the same and it makes the point better than any diagram: the curve did not disappear at graduation. It is still on chain, still owned by the launchpad, still being addressed, and answering every one of those attempts with the same refusal.

The curve still thinks there are a billion tokens

One field on that account never gets zeroed, and it turns out to explain something this site has run into repeatedly.

The curve records token_total_supply as 1,000,000,000 tokens. That was true on 26 July 2026 and it has not been true for a while. A Solana RPC node queried on 11 August 2026 reports the mint’s actual supply as 964,170,083, because roughly 35.8 million tokens have been burned since launch. That reading was already eight tokens lower than one taken a few hours earlier the same morning, which is the point: burns continue, so the real figure keeps falling while the curve’s copy of it never moves. The curve overstates supply by about 3.72% today and by more tomorrow.

That would be a curiosity if it stopped there. It does not. Pump.fun’s own fee schedule states that for its canonical pools “the market cap is calculated as the current price of the token in SOL or USDC multiplied by 1 billion tokens”. The billion is hardcoded into the fee maths for every coin the platform has ever launched, whether or not a single one of them still has a billion tokens.

So when a third-party tracker publishes a supply that disagrees with the chain, the explanation is usually not carelessness. The launchpad itself treats one billion as the number, in an account that never updates and in a fee formula that ignores burns entirely. Anything downstream that inherits it inherits the error too. The live figures for this token are on the stats page, read from an RPC node hourly for exactly this reason.

What it costs, according to the platform

The fee page is worth reading directly rather than through summaries, because it is the part that has moved most. Everything in the table below is quoted from pump.fun’s fee documentation as it stood on 11 August 2026.

Creating a coin costs nothing. Trading on the curve costs 1.25%, of which 0.300% goes to the creator and 0.95% to the protocol, with no share to liquidity because on a curve there is no liquidity provider to pay. Graduation costs 0.015 SOL. After graduation the canonical pool’s fee is tiered by market cap, and any non-canonical pool for the same coin charges 0.3%.

Two dates on that page are useful context. Creator fees apply to coins present on the curve or on PumpSwap from 13 May 2025 onward, which is why older coins behave differently. And since 21 May 2026 a creator can pair a new coin against USDC instead of SOL, with a slightly different schedule. The page also states the platform may change any of this at any time without notice, which is the most quotable sentence on it.

The split is more informative than the headline rate. On the curve, none of the 1.25% goes to liquidity, because a virtual reserve has no provider to compensate. On a graduated pool the reverse is true and most of the smaller fee goes to liquidity instead. So the fee structure is not just cheaper after graduation, it is pointed at a different party, and that shift is a reasonable summary of what graduation does: it converts a coin from something the platform operates into something a market operates.

How to check any of this on any coin

None of the above needs a paid API or an account anywhere. The bonding curve address is derived from the mint, so given a mint you can find the curve, and the curve tells you the rest.

Read the curve account and look at two things. If complete is false and the reserves hold real numbers, the coin is still on the curve and the price is pure formula. If complete is true and all four reserves are zero, it graduated, and the account is now a receipt. Either way token_total_supply tells you the launch supply, which you can compare against a live getTokenSupply call to see exactly how much has been burned since.

Then read the pool’s owner program to see which AMM it actually landed on, rather than assuming from an article. Pool ownership is a single field and it does not have opinions.

One caution carried over from earlier work on this site. This mint is a Token-2022 mint rather than an original SPL Token mint, and the two programs return disjoint results, so tooling that queries only the older program will report nothing at all for it. That trap is covered in how wallets actually get drained. The related question of what the burn on the graduated pool does and does not protect is in what LP burned and mint revoked actually protect you from, and why the pool you are shown is rarely the whole market is in how accurate DexScreener is.

The curve is the least discussed of the four and the easiest to read. It cannot be edited, it cannot be taken down, and it is the only record that says what the coin looked like before anybody was watching.

This site is community run and does not speak for the project. Every figure above was read from public sources on 11 August 2026 and is stated with that date attached, because several of them will be different by the time you read this. The ones that will not change are the constants: 30 SOL, 1,073,000,000, 793,100,000, and 85.0054 SOL to the end of the curve.