How Crypto Launchpads Work: Bonding Curves, Graduation and the 2026 Landscape

Introduction

Between 1.5% and 3% of tokens launched on a bonding-curve platform ever reach an open exchange. The rest stall on the curve and stop trading. That base rate sits underneath a category where a single platform has cleared $1B in lifetime revenue, and it rarely appears in the guides explaining how these platforms operate.

Vocabulary is part of the problem. The term crypto launchpad covers two products with opposite risk profiles, and most explainers pick one sense without telling the reader which. Staleness is the other part. Widely read guides still describe a migration route and a 6 SOL fee that both disappeared in March 2025.

This article covers the bonding-curve model end to end: the pricing mechanism, the graduation event, the fee split, the quote assets that changed in 2026, the risks burned liquidity does not remove, and what a federal court actually decided.

What Is a Crypto Launchpad, and Why Does the Word Mean Two Things?

Two different products share the name. One is a vetted fundraising platform that screens projects, enforces know your customer (KYC) checks and releases tokens on a vesting schedule. The other mints a token in seconds for anyone, with no screening at all. Both are called crypto launchpads.

Two products, one word

The traditional model predates the memecoin era. A project applies, the platform reviews its whitepaper and team ↗, and investors clear KYC before buying an allocation. Access is gated behind staking the platform's own token on many platforms, and tokens unlock on a published schedule. Binance Launchpad, TrustSwap and BSCPad all operate this way (CoinGecko, 2026).

The bonding-curve model removes every one of those steps. A creator supplies a name, a ticker and an image, pays a fee measured in cents, and the token trades in a single transaction. No review, no allocation, no vesting, no identity check. pump.fun launched this design on Solana in January 2024 and became the first Solana application to pass $1B in cumulative revenue (CoinMarketCap, 2026). The gap between the two models is not a matter of degree. They screen nobody versus screening everybody.

Which model this guide covers

This guide covers the bonding-curve model. The distinction carries real consequences, because the two products sit at opposite ends of the risk range under one search term. A reader who arrives expecting vetted allocations and meets a permissionless token factory has been misled by vocabulary rather than by any platform.

Sale formats live inside each model instead of replacing it. An initial coin offering (ICO), an initial decentralized exchange offering (IDO) and an initial exchange offering (IEO) are structures a traditional launchpad runs. A token generation event (TGE) marks the moment tokens become transferable. A bonding curve is a pricing mechanism rather than a sale format, which is why none of those labels fit it. Anyone weighing the four formats side by side should start with the dedicated comparison.

Data current as of October 2026.

The mechanism that makes a sub-minute launch possible is the bonding curve itself.

How Does a Bonding Curve Price a Token With No Liquidity Pool?

A bonding curve replaces the order book and the liquidity pool with a formula. The contract holds both sides of the market and quotes every trade itself. A token is therefore tradable from the first block without anyone funding it. Price follows supply sold, not buyer sentiment.

Price as a function of supply

The contract is the counterparty to every trade. A buyer sends the quote asset and receives tokens at the price the formula calculates ↗. That price depends on how much of the curve supply has already been sold. A seller does the reverse. No second trader has to exist, and the quoted price can be calculated before signing the transaction.

Major launchpads including pump.fun use a constant-product rule, the same invariant that runs Uniswap-style pools (DEXTools, 2026). The two sides of the curve multiply to a fixed number. Adding quote asset to one side forces the other side down, and the tokens released are what the buyer receives. Two consequences follow. The curve never runs out of liquidity, and the deployer cannot withdraw the collected funds, because the contract holds them rather than a wallet.

Virtual reserves and the steepening curve

The curve does not start empty. It is initialised with virtual reserves: placeholder balances written into the contract so the first purchase has a defined price instead of dividing by zero. Only part of the displayed reserve is real capital, which is why early market-cap readings on a fresh token overstate the money committed.

That design produces the shape buyers describe as exponential. The opening stretch is close to flat, and the final stretch before the threshold climbs steeply, because each purchase removes tokens from a shrinking pool. A buyer entering at 10% of the curve and one entering at 90% are running different trades on the same asset. Several widely read explainers describe the mechanism as minting tokens on each buy and burning them on each sell. Production launchpads pre-allocate a fixed supply to the curve instead.

The curve is only the first half of a token's life. The second half begins at graduation.

What Actually Happens When a Token Graduates?

Graduation is one atomic transaction, not a migration process. The curve closes. The collected quote asset and the remaining supply seed a pool on an automated market maker (AMM), and the liquidity provider (LP) tokens are burned. Either all of it lands in the same block or none of it does.

What fires at the threshold

The final buy pushes the curve past its target and flips a completion flag ↗. The contract stops accepting trades against the curve and withdraws the accumulated quote asset and the unsold tokens. It creates an AMM pool from both, then destroys the LP tokens that prove ownership of that pool. Burning them makes the liquidity permanent, which removes one specific failure mode: nobody can pull the pool.

The starting price on the new pool matches the final curve price, so holders see no gap at the handover. Trading continues immediately, and aggregators index the pool within seconds.

Where the liquidity actually lands

Guides published well into 2026 still state that pump.fun tokens graduate to Raydium and that roughly 6 SOL is deducted as a migration fee. That stopped being true on 20 March 2025, when pump.fun launched PumpSwap and made it the graduation destination. That 6 SOL charge was removed and migration became instant, though pump.fun still deducts 0.015 SOL at graduation to pay for the accounts the migration creates (pump.fun fee docs, 2026).

The threshold itself is the second detail most guides get wrong. On pump.fun it is denominated in SOL, at roughly 85 SOL. The dollar market cap it corresponds to therefore moves with the SOL price (as of 15 September 2026) . The widely repeated $69,000 figure is a snapshot of that relationship, not a fixed rule written into the contract. Tokens that graduated before March 2025 remain on Raydium, so both venues hold pump.fun graduates and the split is historical rather than current. Instant migration also removed the brief window during which a graduating token was not tradable at all.

Graduation is the milestone almost every launched token fails to reach.

How Many Launched Tokens Actually Make It?

Most tokens never graduate. Beyond that, the honest answer is a range rather than a number. Published graduation rates for pump.fun span 0.198% to 5%, a spread of roughly 25 times. Sources rarely agree on sample size, period or method.

Why the published rates disagree

The lowest figure in circulation, 0.198%, came from an academic preprint whose own authors later relabelled it as a lower bound rather than a true rate (arXiv 2607.02823, 2026). Their collector sampled each launch for only a few minutes, so tokens that graduated later were counted as failures. A 2025 academic study of 655,770 launches reported 0.63%. An August 2026 industry sample of 1.03 million launches reported 1.76%. A September 2026 estimate drawn from 517 tokens reported 3.09%.

Those numbers describe different populations measured in different market conditions, which is the whole explanation for the spread. Sample sizes ranging from 517 to over a million cannot produce comparable precision. Any article quoting one of them as the graduation rate, with no date and no denominator, is quoting an artefact.

The defensible range

A working range of 1.5% to 3% covers the second half of 2026, and the 1.76% figure carries the largest sample among published estimates, at 1.03 million launches (madeonsol, 2026). Citing the range with its provenance is more accurate than citing any single value to three decimal places.

The rate also varies by platform rather than being a property of bonding curves. A launchpad setting its threshold near $10,000 graduates a far higher share of tokens than one setting it near $69,000. That difference reflects the threshold rather than token quality. Graduation measures one event: the curve filled. It carries no information about what happens to the price afterwards. Comparison across platforms needs the same caution. A rate computed over a bull-market week and one computed over a quiet month describe different conditions rather than different products. Anyone reusing a figure from an article should check its publication date first, because a 2025 rate is no longer a current rate.

The economics behind those launches explain why platforms keep running them.

What Do Launchpads Charge, and Who Gets Paid?

Three separate money flows get conflated into one. Creating a token costs a fixed fee measured in cents. Every trade on the curve pays a percentage fee. The protocol's share of that fee then goes somewhere specific, and where it goes changes without notice.

The curve fee and its split

Creation is close to free by design ↗, because the model depends on volume rather than launch fees. The recurring charge is the per-trade fee on the curve ↗, set at about 1% of each buy and sell, deducted in the quote asset. Platforms then divide that fee between the creator and the protocol in ratios that differ sharply between launchpads.

The split is the real variable. Some platforms route the majority to the creator. Others pay creators nothing at all during the curve phase and share only post-graduation pool fees. Effective cost to a trader also rises above the headline rate wherever a launchpad applies a decaying anti-sniper tax to early purchases. That tax starts near 99% on some platforms and falls over minutes.

What the protocol keeps

pump.fun is the clearest worked example, and its history contradicts the version repeated in older guides. On 28 April 2026 the platform burned roughly $370M of previously repurchased PUMP, about 36% of circulating supply, across two Solana transactions (CoinDesk, 2026). The same announcement cut the buyback allocation from 100% of revenue to 50% of net revenue (CoinDesk, 2026). The previous policy had run nine months, and the new one sits in an irreversible contract for one year.

Reported totals now stand at $453.75M deployed and 165.64B PUMP burned (as of 14 September 2026) . The platform's own dashboard shows daily allocation running between 37% and 51%, so the 50% figure is a target rather than a measured constant.

Data current as of October 2026.

Fees are collected in a quote asset, and that asset stopped being predictable in 2026.

What Are New Tokens Actually Priced In?

For three years the answer was the chain's native token. New launches were quoted in SOL, in BNB or in ETH, and every indexer assumed it. During 2026 several launchpads made the quote asset a creator's choice, and tokenized equities entered the list.

The quote asset became a choice

A launchpad that supports multiple quote assets lets the creator pick what the curve collects. Options now include stablecoins, wrapped bitcoin, tokenized stocks and tokenized exchange-traded funds ↗ (ETFs). A memecoin can be priced against a synthetic tracker for a listed share rather than against a native chain token.

Those trackers are debt-like instruments that follow a share price. They carry no voting rights and no direct claim on the issuer, which places the arrangement closer to securities law than to conventional token design. A reader evaluating a stock-quoted launch is holding exposure to two assets at once, and only one of them behaves like a memecoin.

Why a floating threshold follows

Multi-quote support breaks a fixed dollar threshold. A curve denominated in a tokenized equity fills when it has collected a set amount of that equity. The dollar value of that amount changes with the underlying share price. Two tokens launched an hour apart on the same pair can complete at different dollar market caps.

Any tool that hardcodes a graduation figure is therefore wrong by construction, and so is any article quoting a single number across platforms. The same applies to volume. Raw quote amounts cannot be summed across launches, because each uses a different asset with its own decimals and its own price. Normalising per asset before aggregating is the only correct method. Indexers built on the older assumption fail silently rather than loudly. They return a number, the number is wrong, and that makes the error harder to catch than an outright failure would be. Dashboards quoting one dollar threshold across a multi-quote platform carry the same defect.

Which Launchpads and Chains Matter Right Now?

The category turns over faster than almost any other in crypto. A new mechanism plus fresh distribution takes the lead within weeks, then loses it on the same timetable. The names below reflect September 2026 and carry no assumption of permanence.

Four chains, four eras

Solana held the category ↗ from January 2024 through most of 2025 on the strength of pump.fun. BNB Chain took a large share during late 2025. Base held the lead briefly in early 2026 around creator-coin and agent-launch platforms. From July 2026 activity moved again, toward a new Arbitrum-stack layer 2 built around tokenized equities.

Each handover followed the same shape. The winning platform in one era rarely wins the next, and the tokens attached to the losing platforms fall hardest. CLANKER trades 92.07% below its all-time high at an $11.4M market cap, and VIRTUAL trades 88.18% below its high at $390.7M (CoinPaprika API, 2026-09-15). Chain share is a steadier lens than platform share, because a chain keeps its infrastructure when a single launchpad fades. Base retained its tooling after its flagship platforms shrank, and Solana kept its aggregators through two changes of leading launchpad.

What is live right now

pump.fun remains the largest cumulative revenue business in the category, having passed $1B in lifetime revenue (CoinMarketCap, 2026). PUMP carries a market capitalisation of $2.34B at rank 49 (CoinPaprika API, as of 14 October 2026). Its share of new activity has fallen as stock-quoted competitors arrived. Several of the fastest-growing 2026 launchpads publish no market data through major aggregators at all, which limits what can be verified independently about them. Coverage gaps are themselves a finding. A platform absent from major aggregators cannot be checked against independent market data, so any figure it reports about itself stands unverified. Treating those numbers as provisional is the only defensible approach.

Data current as of October 2026.

Table reflects status as of 15 September 2026 .

Platform turnover is only one of the risks attached to this model.

What Are the Real Risks on Both Sides of a Launch?

Burned liquidity prevents exactly one attack. It stops anyone draining the pool, and it stops nothing else. Creator selling, automated sniping, manufactured volume and industrial-scale token farming all survive the mechanism that marketing describes as rug-proof. The distinction matters on both sides of a launch.

Risks the curve does not remove

A creator who buys their own token at launch can sell into the first wave of demand ↗. Burned LP tokens have no bearing on that. Automated snipers compete to buy in the opening seconds, and again at the moment a pool opens. They pay high priority fees to win that race, so manual buyers fill at worse prices.

Volume is the next trap. A launchpad reporting fees that represent an implausible share of its own trading volume is showing a wash-trading signature rather than organic demand. Name collisions are routine, with a large share of new launches sharing a ticker with another launch minted minutes earlier. Deployer history matters more than any of it: a meaningful share of new tokens comes from wallets that have already minted hundreds.

Risks specific to creators

A creator carries the mirror image of these problems. Fee revenue depends on trading volume that most tokens never generate. Expected earnings should be weighed against a graduation rate between 1.5% and 3%, rather than against a headline creator split.

Platform risk sits above that. Buyback rates, fee splits and creator shares are policy settings rather than contract guarantees. pump.fun's cut from 100% to 50% of revenue shows how quickly they move. Any launchpad whose revenue-sharing promise is published on a web page rather than enforced by an immutable contract can revise it at any point. Distribution is the other constraint. A launchpad with weak reach produces no volume regardless of its fee terms. The most generous creator split on a platform nobody uses pays nothing. Terms and traffic have to be assessed together.

The legal position has moved faster than the technology.

Are Launchpad Tokens Securities? What a Court Actually Decided

A United States federal court addressed the question directly in 2026, and the ruling is narrower than the headlines suggest. Securities claims failed. Racketeering claims survived, resting on wire fraud and unlicensed money transmission as their predicates. The legal exposure moved from what a launchpad sells to what a launchpad says about itself.

What the court actually held

Judge Colleen McMahon ruled on 31 August 2026 in Aguilar v. Baton Corporation Ltd., before the Southern District of New York ↗. The Securities Act claims failed because two specific tokens did not satisfy the common-enterprise prong of the Howey test (Wolf Popper, 2026). That holding covers those tokens on those facts. It is not a ruling that bonding curves are exempt, and counsel quoted in the coverage said so explicitly.

The same order allowed the racketeering claims to proceed against the operating company and three named founders (AInvest, 2026). Unlicensed money transmission is one of the predicates inside that claim rather than a separate count, alongside wire fraud, which rested on marketing language describing the platform as an unruggable fair launch. Solana Labs, the Solana Foundation and five named executives were dismissed from all claims.

What is still open

The claims of the plaintiff named in the case caption were dismissed for failing to plead a domestic injury, while the court-appointed lead plaintiff and another proceed. Two gambling predicates were rejected outright: buying a token is not a wager under New York law, and the offence relied on carries too short a maximum sentence to qualify. The case now moves toward discovery rather than a final answer.

Regulators have been more permissive than the litigation suggests. SEC staff stated in February 2025 that a meme coin is not itself a security. No equivalent European enforcement action against bonding-curve launchpads has surfaced. Compliance exposure for platform operators therefore concentrates in two places: registration as a money transmitter, and the accuracy of promotional claims about fairness.

Taken together, the mechanics, the base rates and the legal position point to a short list of checks.

How Should You Evaluate a Launchpad Before Using It?

Every item below is verifiable before committing funds, and none of it requires trusting a platform's own description. The checks split by role: a buyer is assessing one token, while a creator is assessing the platform's durability and terms. Each check takes minutes and needs no special tooling.

Checks before buying

Confirm the migration event on-chain rather than reading a progress bar, because a curve can sit at 99% for hours without completing. Check the deployer wallet's history for prior mints. Compare reported fees against reported volume, and treat an implausible ratio as a wash-trading signal. Read the holder distribution before entering, because concentration in a handful of wallets sets the size of the first sell wave. Confirm that the LP tokens were burned rather than merely locked, since a lock carries an expiry date and a burn does not.

Checks before launching

Read the fee split and the buyback policy in the contract rather than on the marketing page, and establish whether either is immutable. Confirm which quote asset the curve collects and what the threshold means in that asset. Compare the platform's distribution against its terms, because reach determines whether any fee split produces income. Check how long the platform has operated and whether the team is identifiable. An anonymous operator with two months of history offers no recourse if terms change.

Data current as of October 2026.

A launchpad organises access and pricing. It makes no claim about whether the token is worth holding, and the base rates in this guide suggest most are not.

FAQ

?Does burning the liquidity tokens mean a token cannot be rugged?

Burning liquidity provider tokens prevents exactly one attack: nobody can withdraw the pool. Every other extraction route stays open. A creator who bought their own token at launch can still sell into the first wave of demand, and automated snipers can still take the best fills in the opening seconds. Permanent liquidity is a guarantee about the pool, not about the price.

?Why do different sites report different graduation rates for pump.fun?

They measure different populations over different periods, and most omit both. Published figures range from 0.198% to 5%. The lowest came from an academic preprint whose authors relabelled it as a lower bound after their collector sampled each launch for only a few minutes. The largest verified sample, 1.03 million launches, produced 1.76%. A range of 1.5% to 3% covers the second half of 2026.

?Is the pump.fun graduation threshold really $69,000?

No. The threshold is denominated in SOL, at roughly 85 SOL, so the dollar market cap it corresponds to moves with the SOL price. The $69,000 figure is a snapshot of that relationship from an earlier period, not a value written into the contract. On platforms supporting multiple quote assets, two tokens launched an hour apart can complete at different dollar valuations.

?Where do pump.fun tokens go after they graduate?

To PumpSwap, pump.fun's own automated market maker, since 20 March 2025. Before that date graduated tokens migrated to Raydium and roughly 6 SOL was deducted as a migration fee. That charge is gone and migration became instant; pump.fun now deducts 0.015 SOL at graduation. Tokens that graduated before the change remain on Raydium, so both venues hold pump.fun graduates.

?Did a court rule that memecoins are not securities?

Not in general terms. In August 2026 a federal judge dismissed Securities Act claims against pump.fun because two specific tokens failed the common-enterprise prong of the Howey test. Counsel quoted in the coverage stressed that the reasoning does not exempt all memecoins. Racketeering claims against the operating company and three founders were allowed to proceed, on wire-fraud and unlicensed-money-transmission predicates. Gambling predicates were rejected.

?Can a token that never graduates still be sold?

Yes. A token sitting on an incomplete curve remains tradable against that curve, and holders can sell back to the contract at the current curve price minus fees. What disappears is attention rather than access. Most non-graduating tokens lose their buyers within hours, so selling is possible while finding a price worth taking is not.

?What separates a launchpad from a decentralized exchange?

A launchpad creates the market; an exchange hosts one that already exists. On a launchpad the contract itself is the counterparty and the price comes from a formula. On a decentralized exchange, liquidity providers deposit both sides of a pair and traders swap against that pool. Graduation is the handover point between the two.

?Do traditional fundraising launchpads still operate?

Yes, and they serve a different purpose. Platforms running initial exchange offerings and initial decentralized exchange offerings still vet projects, enforce know your customer (KYC) checks, gate access behind staking tiers, and release tokens on vesting schedules. Those launches target capital raising for projects with teams and roadmaps, rather than instant permissionless token creation.

References / Sources