AMM Design in 2026: The Liquidity Provider’s Playbook

Every time someone swaps one token for another on a decentralized exchange, a group of strangers on the other side of that trade either gets paid or gets picked off. Those strangers are liquidity providers, and there is no trading desk behind them, no quoting engine, and no human deciding when to step away from a bad market. There is only a formula. Providing liquidity to an automated market maker gets marketed as passive income, but it is really a market-making business run by a piece of code that never flinches, never widens its spread, and never turns down a losing trade. Whether that business pays you or quietly drains you comes down to one thing: the design of the pool you picked.

That question carries more weight in 2026 than it ever has. Decentralized exchanges cleared a record share of spot crypto trading in July, with the DEX-to-CEX volume ratio hitting 24%, the highest since The Block began tracking it in 2019, as centralized-exchange spot activity slumped and capital moved on-chain. More money sits in AMM pools than at any earlier point in DeFi. And the uncomfortable finding, established years ago and still relevant, is that a large share of those providers would have been better off holding their tokens and doing nothing.

This piece reads AMM design from the liquidity provider’s seat. We toured the machinery in a companion article on how one formula branched into bins, hooks, and batch auctions; here the question is narrower and more personal. Given how these pools are built, which design choices pay a liquidity provider, which ones tax one, and how do you tell the two apart before your money is in the contract?

What You Actually Become When You Provide Liquidity

You deposit two tokens in the ratio the pool requires, and you receive something back that represents your share: a fungible LP token in older designs like Uniswap v2 and Curve, or a position NFT in concentrated-liquidity designs like Uniswap v3 and v4. From that moment, you are the counterparty to every swap in the pool. A trader who wants ETH takes it out of your inventory and leaves USDC behind; a trader who wants USDC takes yours and leaves ETH. You collect a small fee on each swap, and in exchange you hold whatever mix of the two assets the market decides to hand you.

The framing that trips newcomers up is treating this as lending or staking. It is neither. It is market making with the discretion removed. A human market maker pulls quotes before a big news release; the AMM keeps quoting straight through it. That obligation to trade at the formula’s price no matter what is exactly what earns the fees, and exactly what creates the losses.

So an LP position has two profit-and-loss lines that pull against each other. One is fee income, the reason to show up. The other is the cost of rebalancing, meaning what the market extracts from you as prices move and traders true up your stale inventory. Everything that follows is about making the first line larger than the second. Academics who modeled the payoff put it more bluntly: a liquidity provider is structurally short volatility, earning a little when prices sit still and losing when they move sharply. That one sentence explains most LP outcomes.

The Formula That Trades Your Inventory For You

The original design, and still the most common, is the constant-product market maker, written as x times y equals k. The pool holds reserves of two assets whose product stays constant, and every trade slides the price along that curve. The elegance is that the pool never needs an outside price feed; supply and demand set the price on their own.

What matters to an LP is what the curve does to your holdings. When ETH rises, arbitrageurs buy ETH out of the pool until its price matches the wider market, which means the pool sells your ETH into the rally and hands you dollars. When ETH falls, the pool buys ETH, loading you up on the asset that is dropping. Mechanically, the constant-product curve is a rule that sells your winners and accumulates your losers. It does the exact opposite of what a trend-follower would do, and that is the root of the losses described below.

This constant drift back toward a fifty-fifty split of value is not a flaw; it is the cost of always being willing to trade. And not every AMM uses the same curve. Constant-sum designs offer zero slippage but drain one side entirely; Curve’s StableSwap blends the two for assets that should trade near parity; weighted pools let you choose a split other than fifty-fifty; concentrated liquidity confines the curve to a chosen price band. Each choice changes how your inventory behaves, and we mapped that whole design space in the companion piece linked above.

Impermanent Loss: The Cost You Sign Up For

Impermanent loss is the gap between what your deposited pair is worth inside the pool and what the same two tokens would have been worth if you had simply held them in your wallet. It shows up whenever the two assets’ prices move apart, and it is symmetric: a token doubling in price costs you the same as it halving.

The word impermanent is doing a lot of work. If prices drift apart and then return to where you entered, the loss evaporates. If you withdraw while they are still apart, the loss is realized and entirely permanent. Since most providers exit a position eventually, it is safer to treat it as a real cost, not a temporary one. Its size follows directly from the formula and depends only on the ratio of the price change, not the direction. Using the standard derivation reproduced by Binance Academy:

One caution about that table: it is the floor of your cost, not the whole of it. It assumes trading is free and instantaneous, so it ignores the fees you earn, which offset it, and the continuous bleed to arbitrage, which adds to it. Necessary to understand, but not sufficient. That second, larger cost has a name.

LVR: The Bleed the Brochure Leaves Out

Impermanent loss compares you to holding, which quietly flatters the AMM, because holding is a do-nothing benchmark. Loss-versus-rebalancing, or LVR, compares the pool instead to a strategy that makes the very same trades at the true market price, and it isolates exactly what arbitrageurs pull out of a pool whose quote is always one step behind.

The mechanism is simple once you see it. An AMM’s price only updates when someone trades against it. So every time the outside market moves, the pool’s price is momentarily stale, and the first arbitrageur to notice trades against that stale quote and pockets the difference. You are perpetually selling a little too cheap and buying a little too dear. That repeated haircut, summed over every price move, is LVR.

The team that formalized it, Jason Milionis, Ciamac Moallemi, Tim Roughgarden, and Anthony Lee Zhang, showed that instantaneous LVR for a constant-product pool equals its volatility squared divided by eight. For an ETH pool running at 5% daily volatility, that works out to roughly 3.125 basis points a day, about 11% a year, and it means the pool needs daily volume near 10.4% of its assets just to cover a 30-basis-point fee, according to a16z crypto.

Their conclusion is the single most important line for anyone weighing an LP position: “An AMM can be successful only if it has happy LPs, which means that fee revenues need to scale with LVR.” In plain terms, if the pool is not earning enough in fees to outrun what arbitrage takes, you are the one funding the difference.

Who sits on the other side of that trade? Arbitrageurs, whose profit is realized as MEV. A large share of the arbitrage MEV extracted on Ethereum is simply LVR being harvested out of AMM pools, a pipeline we traced in our coverage of on-chain value extraction. The practical lesson: your real competition is not other liquidity providers, it is informed trading flow, and the more volatile your pair, the harder that flow hits you.

Fees: The Only Line That Pays You

Fees are the sole source of income for a liquidity provider, so the entire exercise reduces to arranging for fee income to beat impermanent loss plus LVR. Everything else is detail.

Uniswap v2 charged a flat 0.30% on every trade. Uniswap v3 broke that into tiers, commonly 0.01%, 0.05%, 0.30%, and 1.00%, so a pool can price its own risk: a percentage-point fee on an exotic pair, a few basis points on tightly correlated stablecoins, and the familiar 0.30% on blue chips, per the v3 design. Higher volatility should command a higher fee, precisely because it drags in more LVR.

The number that actually predicts LP returns is not the APR on a dashboard. It is the ratio of fee-generating volume to the liquidity sitting in the pool. A pool turning over 100 million dollars a day on 50 million of TVL pays its providers far better, per dollar deposited, than one doing 5 million on the same 50 million, at an identical fee tier. Chase the volume-to-TVL ratio, not the headline yield.

Because a fixed fee cannot track changing volatility, newer pools use dynamic fees, raising the charge when volatility and therefore LVR spike, and lowering it when markets are calm. And there is now a new mouth at the table. The UNIfication upgrade switched on Uniswap’s protocol fee: for a standard v2 pool, the provider’s cut drops from 0.30% to 0.25%, with the remaining 0.05% routed to buy and burn UNI, alongside a substantial one-time token burn, as laid out in Uniswap’s proposal. For LPs the message is blunt. After arbitrage and impermanent loss, the protocol itself now takes a slice, and the switch is live and rolling out chain by chain through 2026.

Concentrated Liquidity: The Skill Game Where Half of LPs Lose

Uniswap v3’s headline idea was to stop wasting capital. In v2, your money is spread thinly across every possible price from zero to infinity, and almost none of it ever gets used. Concentrated liquidity lets you pack your capital into a price band you choose, where it does real work. Uniswap advertised up to 4,000 times the capital efficiency of v2 for a tight range in its v3 announcement, and because everyone picks a different range, your position becomes an NFT rather than a fungible token.

The catch comes in three parts. First, if the price leaves your chosen band, you stop earning fees completely and your position converts entirely into the weaker asset, which is maximum impermanent loss with zero income to show for it. Second, a tighter band earns more fees while the price sits inside it but falls out of range faster, so you are always trading yield against staying in the game. Third, moving your range costs gas and locks in impermanent loss every time you do it.

The receipts are sobering. A widely cited study by Bancor and Topaze Blue found that just under half of Uniswap v3 liquidity providers, 49.5% of them, actually lost money compared with simply holding, because the pools it examined earned 199.3 million dollars in fees while suffering 260.1 million in impermanent loss over the study window, as The Defiant reported. The only cohort that reliably beat holding were just-in-time providers: bots that add liquidity for a single block to skim the fee on one large trade, then withdraw before any price move can hurt them.

The takeaway is not to avoid concentrated liquidity. It is that concentrated liquidity turned market making into a skill game. A passive investor picking a narrow range on a volatile pair is sitting down to poker against professionals with bots. That is precisely why most people who keep providing liquidity either widen their ranges, move to calmer pools, or hand the job to someone whose full-time work is managing it.

Passive or Active: Rebalancing and the Auto-Managers

There are two honest ways to run a concentrated position: babysit it yourself, or outsource it. There is also a dishonest third way, which is to open a tight range and forget about it, and that is how the losses in the study above pile up.

Doing it yourself means widening when volatility rises, tightening when it calms, shifting the band as the price approaches its edge, and harvesting fees along the way. It works, but it is a part-time job, and every rebalance pays gas and realizes a slice of impermanent loss.

Automated liquidity managers do the same work programmatically. Gamma runs non-custodial vaults that rebalance concentrated positions across Uniswap v3, v4, and other concentrated-liquidity AMMs using volatility-based strategies, per its documentation. Arrakis built trustless market-making vaults and grew into one of the largest liquidity providers on Uniswap without ever paying token incentives. Others, from Bunni to Beefy, wrap the whole process into one-click products, some routing idle capital into lending markets for extra yield.

The trade-off is real. A manager charges a fee and adds another layer of smart-contract risk, but for most providers a competent automated manager beats a hand-tuned range that goes stale the moment you look away. Read the strategy before you trust it, though. The label active management covers everything from genuine volatility modeling to a script that rebalances on a fixed timer and slowly bleeds your position in gas.

The Low-Drama Corner: Stablecoin and Correlated Pools

If price divergence is what drives impermanent loss and LVR, the obvious defense is to provide liquidity for assets that are not supposed to diverge. Two dollar-pegged stablecoins. A liquid staking token paired against the asset it tracks. Two wrapped versions of the same coin.

This is where StableSwap earns its keep. Curve’s invariant blends a constant-sum curve, which gives near-zero slippage, with a constant-product curve, which keeps the pool from ever fully emptying, tuned by an amplification coefficient. A USDC/USDT pool therefore quotes close to one-to-one with tiny slippage and tiny divergence loss. Curve still anchors this part of DeFi: it holds around 1.36 billion dollars in total value locked and has stayed in a rough 1.5-to-2-billion range through 2026, far below its 2022 peak but steady through a punishing stretch for the rest of the sector, per DefiLlama. Nearly every major stablecoin and most liquid staking tokens route through it.

Liquid staking pools add a twist worth understanding. A wstETH/ETH pool earns swap fees on top of the staking yield already compounding inside wstETH, with only mild divergence risk as long as the peg holds. The risk is the peg itself: staking tokens can and do trade below their redemption value under stress, and that gap becomes real impermanent loss. We compared the major staking tokens and how their pegs behave in a separate breakdown.

The honest trade here is lower risk for lower reward. Stable and correlated pools pay in single digits, not the eye-watering numbers on volatile farms, and in the current interest-rate environment those yields compete directly with simply holding a dollar stablecoin or a tokenized Treasury. Know what you are actually beating before you lock up capital.

Where the Yield Actually Comes From

A pool can pay you in two very different currencies, and confusing them is one of the most common ways liquidity providers get hurt. The first is fees from real trading. The second is token emissions, freshly printed by the protocol to attract deposits.

Real fee yield is sustainable, because it comes from volume you can verify on-chain. Emissions yield is a subsidy that lasts exactly as long as the protocol keeps printing and the emitted token holds its value. A 40% headline APR that turns out to be 5% fees and 35% newly minted governance token is a bet on that token wearing the costume of market making.

The dominant design in this category is ve(3,3), popularized by Solidly and now best known through Aerodrome, the largest DEX on Base, which has taken in over a billion dollars in deposits and handles the majority of the chain’s DEX volume, according to The Block. Holders lock the token for voting power, then direct emissions to the pools they choose and collect the fees and incentives those pools generate. It aligns incentives better than plain farming, but for a passive LP the test is unchanged: strip out the emissions, and does the pool still pay? If not, you are the exit liquidity for everyone selling the printed token.

The rule of thumb is to separate the two yields before depositing. Ask what the pool pays in fees alone, treat emissions as a volatile bonus rather than income, and remember that mercenary liquidity vanishes the instant the subsidy stops, taking the pool’s depth, and your fee income, with it.

Hooks and the 2026 LP Toolkit

The most consequential design change for liquidity providers since concentrated liquidity is Uniswap v4’s hooks, which shipped when v4 launched in early 2025, per Uniswap. A hook is custom code that runs at defined moments in a pool’s life, before or after a swap, on a deposit, on a withdrawal, which turns a pool from a fixed formula into a programmable one.

For an LP, hooks open up several tools that did not exist before:

• Dynamic fees that rise with volatility, so the charge tracks LVR in real time instead of guessing at it with a fixed tier.
• Auction-managed AMMs, or am-AMM, which replace the fixed fee with an on-chain auction for the right to be the pool’s temporary manager; the winner sets the fee and captures arbitrage that would otherwise leak away as LVR, and the winning bid flows back to LPs. The design comes from Austin Adams, Ciamac Moallemi, Sara Reynolds, and Dan Robinson (arXiv), and Bunni runs a live version on v4.
• JIT-resistance and MEV-aware routing, hooks built to keep just-in-time bots and sandwich attackers from skimming the pool at the providers’ expense.

At the other end of the spectrum sits CoW AMM, which sidesteps arbitrage by refusing to quote continuously at all. Trades clear in batches at a single price set by competing solvers, so there is no stale quote for an arbitrageur to exploit, a design meant to eliminate both LVR and sandwich losses for its providers, as the CoW documentation explains. It runs on Balancer.

The catch is significant. A hook is arbitrary code sitting directly in the swap path, and a buggy or malicious one can trap funds or reroute them. Concentrated liquidity made providing liquidity a skill game; hooks make pool selection a code-audit question as well. Favor audited, widely used hooks, and treat a brand-new one with the same suspicion you would give a brand-new token.

A Pool-Selection Framework

Put the pieces together and the decision to provide liquidity comes down to four questions. How correlated are the two assets? How much real volume does the pool do per dollar of liquidity? Where does the yield come from, fees or emissions? And how much contract risk are you taking on? The table below maps common pool types against those questions.

The pattern is the one you would expect: risk and reward move together. The two mistakes that account for most LP losses are putting passive money into a tight range on a volatile pair, and mistaking emissions for fee income. Sidestep those two, match the pool to your appetite and your attention span, and you have already avoided most of the ways this ends badly.

The Risks That Are Not on the Yield Dashboard

Impermanent loss and LVR are the market risks, and they get all the attention. The risks that actually take an LP position to zero are usually the other kind, the ones no APR figure shows you.

Smart-contract risk sits underneath everything. Pools, routers, hooks, and manager vaults are all code, and code gets exploited. Even audited protocols have been drained, and cleaning up afterward has become its own industry, as we covered in our look at crypto’s post-hack recovery race. Favor battle-tested contracts with meaningful TVL and time in the market.

Malicious tokens are the flip side. Anyone can spin up a pool for any token, and a token with a hidden mint function, a transfer tax, or a honeypot can drain everything you pair against it. If you have never heard of the token, assume the pool is bait. Approval risk rides alongside this: providing liquidity means signing token approvals, and a malicious front end or a wallet drainer can turn one signature into a theft. Verify what you are signing and revoke stale approvals; we tested how well the major wallets catch this in a head-to-head on drainer defenses.

Regulation, Taxes, and Whether It Is Worth It

Two forms of friction get underrated by liquidity providers: regulation and taxes.

On regulation, the tone in the United States has softened. The SEC sent Uniswap Labs a Wells notice in 2024, then closed the investigation in February 2025 with no enforcement action, under Acting Chair Mark Uyeda and Commissioner Hester Peirce, as CoinDesk reported. The agency’s current posture treats the interface and the underlying protocol as distinct, and providing liquidity from a self-custodied wallet is not, for now, a registered activity. That is a policy stance rather than settled law, and it can shift with the next chair.

Taxes are the more immediate headache, and the messier one. The IRS has not issued clean, LP-specific guidance, so practitioners disagree on whether entering or exiting a pool counts as a taxable disposal, how to treat LP tokens, and when fee income is recognized. Under some readings, every swap the pool executes on your behalf is a taxable event. Keep detailed records of your cost basis; a yield that looks attractive before tax can look very different after it, and worse if you never tracked the numbers.

So is providing liquidity worth it in 2026? The people who build these systems are optimistic about the design. Uniswap founder Hayden Adams has argued that AMMs are “only just getting started,” rejecting the claim that the model is unsustainable, on the grounds that on-chain liquidity is cheaper to run and composable in ways an order book can never match, in remarks reported by CryptoPotato.

Both things can be true. The AMM is a genuine breakthrough in market structure, and most casual liquidity providers still lose to holding. The difference between the two outcomes is not luck; it is the design of the pool you choose and the honesty with which you count fees against impermanent loss and LVR. Read the pool before you fund it, treat emissions as a bonus and not a wage, and remember the researchers’ warning: a pool only works if its LPs are happy, and no one is going to check whether you are except you.

Frequently Asked Questions

What is impermanent loss in simple terms?

It is the difference between what your deposited tokens are worth inside an AMM pool and what they would have been worth if you had just held them, and it grows as the two assets’ prices move apart. It only becomes permanent if you withdraw while the prices are still diverged; if they return to your entry point, it disappears.

Can you actually make money providing liquidity?

Yes, but only when fee income beats impermanent loss plus the arbitrage bleed known as LVR. Stablecoin and correlated pools manage this reliably at modest single-digit yields, while volatile concentrated positions can pay far more but are a skill game; one widely cited study found roughly half of Uniswap v3 providers underperformed simply holding.

What is the difference between impermanent loss and LVR?

Impermanent loss compares your position to holding the two tokens, while loss-versus-rebalancing compares the pool to a strategy that trades at the true market price, isolating exactly what arbitrageurs extract from a stale AMM quote. Impermanent loss is a point-in-time snapshot; LVR is the continuous cost that accrues every time the price moves.

Are stablecoin liquidity pools safer than volatile ones?

Generally yes, because two dollar-pegged assets barely diverge, so impermanent loss and LVR stay small. The main risk shifts to a depeg and to the contract itself: if a stablecoin loses its peg, the pool ends up holding the broken one, because the formula keeps buying it on the way down.

Do you owe taxes on liquidity provider rewards in the US?

Almost certainly, although the IRS has not issued clean, LP-specific rules. Fee income is generally taxable, and depending on interpretation, entering or exiting a pool and the pool’s own swaps can also be taxable events, so keeping detailed records of your cost basis is essential.

By Yuki Tanaka, senior DeFi correspondent, HOGE Wire.