Impermanent loss is the central risk of providing liquidity to a constant-product AMM like Uniswap V2. If you deposit two tokens into a pool — say, 1 ETH and 3000 USDC when ETH is worth $3000 — and the price of ETH moves in either direction, the pool will automatically rebalance itself through arbitrage, and when you withdraw your liquidity you will end up with fewer of whichever token went up and more of whichever token went down. Compared to just holding the two tokens in your wallet, you are worse off. That gap is impermanent loss.
The name is misleading. It is called “impermanent” because the loss goes away if the price ratio returns to what it was when you deposited. In practice, prices usually do not come back, and the loss is as permanent as any other loss. The name was chosen to make the risk sound less alarming than it actually is, and it has been criticised ever since by people who think LPs should know what they are signing up for.
The Math
For a standard 50/50 constant-product pool, the impermanent loss as a function of price change works out to a formula you can look up. The rough intuition: a 2x move in one token’s price produces about a 5.7 percent loss relative to holding. A 3x move produces about 13 percent. A 5x move produces 25 percent. A 10x move produces about 43 percent. These numbers are measured against just holding the two tokens, not against holding cash, so an LP can be down relative to the market-neutral benchmark even when the total dollar value of their position has gone up.
The direction does not matter — the loss is symmetric. If ETH goes from $3000 to $6000, you lose against a buy-and-hold benchmark. If ETH goes from $3000 to $1500, you also lose against buy-and-hold, because the pool rebalanced you into more ETH on the way down. The loss is a direct consequence of the AMM’s rebalancing behaviour, not of any specific market direction.
What’s Supposed to Offset It
Liquidity providers get paid trading fees — typically 0.3 percent of each swap in the pool, split proportionally among LPs. The theory is that fees accumulate over time and offset the impermanent loss so that LPs end up ahead. Whether this actually works depends entirely on how much volume the pool does and how volatile the assets are.
For a stable-stable pair like USDC/DAI, fees almost always outweigh IL, because the prices barely move and the loss never gets meaningfully large. For a blue-chip volatile pair like ETH/USDC on a high-volume pool, fees sometimes outweigh IL and sometimes do not; results vary by period and pool. For long-tail pairs — a random altcoin against ETH — IL almost always exceeds fees, because the altcoin either moons (big loss against holding) or dumps (big loss in absolute terms), and the fees generated are rarely enough to compensate.
Several studies have tracked Uniswap LP returns and found that a majority of LPs, across most pools, would have been better off simply holding. The combination of IL, gas costs, and the opportunity cost of capital tends to make passive LPing a loser for unsophisticated participants.
The V3 Wrinkle
Uniswap V3 introduced concentrated liquidity, which lets LPs specify a price range where their capital is active. This amplifies both fees and impermanent loss within that range, and converts LPing from a passive strategy into something much closer to active market-making. V3 LPs can outperform buy-and-hold if they are good at picking ranges and managing positions, but the skill bar is substantially higher than on V2, and the risk of being badly wrong on range selection is real. For most retail LPs, V3 is harder to make money in than the V2 model was, not easier.