A liquidity bootstrapping pool (LBP) is a token-sale mechanism, first built on Balancer, that lets a project distribute a brand-new token and discover its price fairly while committing very little starting capital. Unlike a fixed-price sale – where whales and bots race to buy the entire allocation at launch – an LBP starts the token priced high and lets that price drift down over the sale unless buyers step in, so participants who wait are rewarded and no single actor can corner the supply cheaply. Because it produces an open, permissionless primary market rather than a negotiated round, it sits alongside bonding curves and curation markets in the family of on-chain capital-formation mechanisms that DAOs use to launch a governance token.
An LBP is a variant of a Balancer weighted pool holding two assets: the project token being sold and a reserve asset (typically a stablecoin or ETH). What makes it special is that the pool owner sets a weight schedule: per the Balancer documentation, the owner specifies "the starting and ending weights for both tokens" and "these weights change linearly over time." A typical sale starts around 90/10 (project token / reserve) and ends near 20/80.
Because a weighted-pool's spot price is a function of the balances and the weights, steadily lowering the project token's weight creates what Balancer calls "controlled downward price pressure." If nobody buys, the price keeps falling; if demand appears, buys push it back up. The net effect is a "continuous Dutch-auction-style mechanism" that stretches price discovery across hours or days instead of resolving it in a single opening block.
LBPs came after several years of fixed-price and capped token sales that failed in recurring ways. Vitalik Buterin catalogued the failures in Analyzing Token Sale Models in June 2017. Capped sales, which sell a fixed number of tokens at a fixed price, drew so much demand that buyers raced to transact first: Brave's Basic Attention Token sale that May raised $35m within 30 seconds, with 185 purchases succeeding, more than 10,000 failing, and over 70 ETH spent on transaction fees. Uncapped sales, including Ethereum's own, left buyers unsure what valuation they were paying. Gnosis tried a reverse Dutch auction, in which the share of supply sold shrank the sooner the sale closed; it reached its $12.5m cap within hours having sold about 5% of its tokens, and the team kept the rest, promising to hold 90% of them for a year.
From these cases Buterin listed what a sale should provide: certainty over the valuation a buyer pays; certainty of being able to buy; a cap on the amount raised; no large retained stake that lets the issuer manage its token's price like a central bank; and no wasted value. He argued that the first two cannot both hold, because a sale that fixes the valuation must turn buyers away once demand exceeds it, and that the last three cannot all hold, because selling most of the supply at a capped price below what buyers will pay produces a shortage, rationed by queueing or by resale. He also objected to single-round sales on timing: a team receives all its capital at the start, when there is least information to tell good teams from bad ones, and once the sale closes it has less reason to keep working.
Fixed-price and first-come sales reward whoever transacts first with the most gas – the classic gas-war and sniping dynamic. An LBP inverts that incentive. Since the schedule pushes price down over time, buying early is buying at the top: a bot that sweeps the pool at launch simply overpays and then watches the weight schedule mark its position down. Rational buyers instead wait for a price they consider fair, which spreads the distribution across many participants and dampens the launch-day volatility that plagues fixed-price listings. The mechanism does not prevent large buys, but it makes buying everything at once economically self-defeating.
The pool owner holds significant, and asymmetric, power over an LBP. Per Balancer's specification, the owner is "the only address that can add liquidity to the pool, which must be done prior to the start of the sale," and "the proceeds can only be removed after the end time." The owner configures the start and end times, the weight schedule, and swap fees, and can pause swaps outright. A dedicated blockProjectTokenSwapsIn setting can prevent the project token from being sold into the pool during the launch, blocking a common manipulation. This concentration of control is the mechanism's central trust assumption: buyers must trust that the owner will not pause the sale, alter the schedule, or pull reserves adversarially.
LBPs moved from a Balancer primitive to a mainstream launch tool largely through Copper Launch, a front-end that made configuring one a point-and-click exercise. Copper rebranded to Fjord Foundry in 2022 and later signed an exclusive LBP partnership with Balancer (BIP-98), extending the format to NFTs. That arrangement is history rather than description: Fjord now runs its own platform and token, and the Balancer factory underneath it has been switched off (see below). Projects including GuildFi, Merit Circle, and Avocado DAO have distributed tokens this way, and Balancer has facilitated nine-figure LBP raises such as HydraDX. For a DAO, an LBP is often the moment a governance token first reaches a broad holder base – the event that turns a core team's project into a token-holder community.
Everything above is sourced from Balancer's v2 documentation, and that stack is no longer the one a project would reach for. Both v2 LBP deployments now sit under v2/deprecated/ in Balancer's own deployment registry, and the factory Copper and Fjord actually built on – NoProtocolFeeLiquidityBootstrappingPoolFactory – has had its on-chain kill switch pulled. isDisabled() returns true on Ethereum mainnet, Base and Arbitrum alike. On mainnet the FactoryDisabled event fires in block 25,070,461 at 2026-05-11T07:46:35Z, inside a Safe execTransaction. The factory is still there and still readable; it just cannot mint another pool.
The successor is Balancer v3's LBPoolFactory, now in its fourth generation (20260501-v3-liquidity-bootstrapping-pool-v4), sitting beside a FixedPriceLBPoolFactory shipped in December 2025 – a sale format that keeps the LBP plumbing and removes the falling price the mechanism is named after.
How much either has been used is a single call. getPoolCount() on every v3 LBP factory ever deployed to those three chains, returned 36 pools between them on 19 August 2026, and 37 when the same call was re-run on 9 September. Each deployment creates one MockLBPool or MockFixedPriceLBPool for its own verification, and the registry names those addresses, so they can be subtracted exactly. Twenty-two pools remain, one more than three weeks earlier. The table below is the September reading; only the mainnet row moved.
| Chain | v3 LBP factories deployed | Pools created | Non-mock | Distinct symbols |
|---|---|---|---|---|
| Ethereum mainnet | 5 | 7 | 2 | 2 |
| Base | 5 | 23 | 18 | 5 |
| Arbitrum | 5 | 7 | 2 | 2 |
Base carries what activity there is, and even there the eighteen pools resolve to five ticker symbols: SINC-USDC-LBP seven times, LBPT-USDC-LBP four, LBP-Long three, AMPS-USDC-LBP three, LBPT-WETH-LBP once – and the two LBPT tickers, five pools between them, read as one team's test series. Arbitrum's two include one literally named TLBP-undefined-LBP. Ethereum mainnet, the chain where the format was invented and where the nine-figure raises happened, carried exactly one non-mock v3 pool for seventeen months: LBPREEE, created 12 March 2025, five days after the first v3 factory shipped.
Re-running the same call on 9 September 2026 found the counter had moved by exactly one in three weeks, and what moved it is the more interesting half of the reading. Mainnet’s second non-mock pool is $JPT-USDC-LBP, which reports itself as Jackpotter Fixed Price Liquidity Bootstrapping Pool, created in block 25,882,547 on 1 September 2026. It came off the FixedPriceLBPoolFactory rather than an LBPoolFactory – the December 2025 sibling that keeps the LBP plumbing and drops the falling price. The first new mainnet pool in eighteen months is therefore not an LBP in the sense the mechanism is named for. Base and Arbitrum returned identical counts across the two readings.
Counting pools is still too generous, and the same registry makes the stricter count one more call. A Balancer v3 pool exists from the moment its factory deploys it, but it holds nothing until someone seeds it, and the v3 Vault records that separately: isPoolInitialized(pool) is false until the first liquidity arrives and true from then on, whatever is withdrawn afterwards. Run over all twenty-two non-mock pools on 9 September 2026 it returns true for ten, and the pool-token supplies agree with it exactly – every uninitialised pool reads zero.
Twelve pools were deployed and never funded. One of them is LBPREEE, which was Ethereum mainnet's entire v3 LBP history for seventeen months and turns out to have held nothing for all of it. Of the ten that were funded, seven hold exactly 1,000,000 wei of pool token – the smallest amount a v3 initialisation can post, which is a smoke test and not a sale – and an eighth holds about a tenth of one. Two carry real size: $JPT-USDC-LBP at 450,000 pool tokens, and Arbitrum's TLBP-undefined-LBP at roughly 555,940, and that second one names itself a test in its own ticker. The count that matters is therefore not 37, nor 22, but two.
Two caveats keep this honest. This is a census of Balancer's contracts on three chains, not of the format everywhere: Balancer deploys to more networks than these, and the mechanism has been reimplemented off Balancer entirely. And Fjord Foundry is still trading – its token-sales app answered 200 on 19 August 2026, and it now advertises Fjord's own $FJO token, staking, and a self-serve sale builder rather than a Balancer front-end. What the numbers do settle is narrower and harder to argue with: the reference implementation this page describes has been switched off where it started, and its replacement has been used a couple of dozen times in seventeen months.
For a DAO the practical reading is not "LBPs are dead". It is that a launch mechanism is a piece of someone else's infrastructure with a governance-controlled off switch, and that the gap between a well-documented mechanism and a used one can be three orders of magnitude. Before planning a raise around a format, count its pools – then ask the vault how many of them anyone ever put money into.
Owner trust. As above, the owner can pause the sale and controls the reserves and schedule. Participants are trusting an EOA or multisig, not just a formula.
Buyer education. The downward-drifting price is counter-intuitive; inexperienced buyers who treat a falling price as a dump can panic-sell or over-pay early. A successful LBP depends on a market that understands the mechanism.
Not manipulation-proof. Sophisticated actors can still accumulate patiently near the floor, and the project can seed wash volume to fake demand. The mechanism improves distribution; it does not guarantee a "fair" outcome.
One-shot. An LBP is a bounded event with a start and an end – it discovers a launch price but does not provide the token a standing, always-open market afterward.
Uniswap released a different launch mechanism on 13 November 2025, the continuous clearing auction (CCA), built with Aztec (Uniswap). Instead of a price that drifts down until buyers step in, supply is released block by block, and at the end of each block the auction sets one clearing price, the highest at which all of that block's tokens sell; everyone filled in the block pays it. A bidder names a maximum price and a total budget, and the bid is spread across all remaining blocks at the same maximum, so an early bid competes in every later block and a late one in fewer. When the auction ends, the proceeds seed a Uniswap v4 pool at the discovered price, which answers the one-shot objection above. The protocol is permissionless, and an optional ZK Passport module lets bidders prove eligibility privately.
The first sale ran from 2 to 6 December 2025. Aztec sold 14.95% of its token supply for 19,388 ETH, about $59m, from 16,741 participants, at a fully diluted valuation of $557m (Unchained); Uniswap reported that 96% of bidders put in less than $10,000 (Uniswap). Tally chose CCA for the token-sale platform it launched in February 2026. Its reasoning was that with a formula-priced bonding curve, a Dutch auction or a batch auction whose bids are visible on-chain, the rational move is to wait and bid late, while CCA rewards bidding early at a true maximum (Tally). Tally shut down before running its own sale.
An LBP and a bonding curve both aim at fair, permissionless price discovery, but they are different shapes. An LBP is a time-boxed event: an owner schedules falling weights, the sale runs, and it ends – after which liquidity has to be arranged separately. Caper takes the always-on route. Every caper's token trades against its own bonding curve where price is a function of how much supply has sold, so raising funds is not a one-off auction but a standing market that opens the moment a caper is created and never needs a counterparty or a close.
There is no owner-set schedule to trust and no window that expires: the curve is the same rule for the first buyer and the ten-thousandth. The trade-off is the mirror image of the LBP's – a bonding curve gives up the LBP's deliberate anti-whale price decay in exchange for a market that is continuous, formulaic, and permanent.
status for every LBP factory.FactoryDisabled, Ethereum block 25,070,461 – the transaction that switched off the v2 LBP factory, 11 May 2026.FixedPriceLBPoolFactory – the December 2025 fixed-price sibling of the LBP.