What Is Yield Farming and How Does It Work?

It represents the accumulated deployment of capital into a set of financial mechanisms the ecosystem has spent years refining: the conversion of idle tokens into usable liquidity, credit, market-making capacity, and settlement infrastructure, with the resulting revenue distributed back to the capital providers.

Yield farming sits at the center of that loop. In its broadest sense, it means allocating digital assets to a protocol or strategy in exchange for a financial return. That return may come from trading fees, lending interest, staking rewards, protocol incentives, or a combination of several sources. The label is broad because the underlying strategies are broad. A liquidity provider supplying a token pair to an automated market maker, a depositor earning interest in a lending market, and a user placing assets into an auto-compounding vault may all describe what they are doing as yield farming.

Stripped of its marketing language, the question of what is yield farming reduces to a structural inquiry: how does a decentralized system attract the capital it needs to function, compensate that capital for the service it provides, and do both without a central operator negotiating either side of the arrangement? The answer is encoded in a few mechanisms, each of which carries its own source of return and its own form of risk.

The Mechanics of Liquidity Provision and Fee Generation

The most recognizable venue for yield farming is the automated market maker, or AMM. An AMM is a smart contract that holds reserves of two or more tokens and prices trades according to a mathematical formula, rather than matching buy and sell orders through a traditional order book.

The original and still influential formulation is the constant-product invariant, written as:

x × y = k

Here, x and y represent the quantities of two tokens held in the pool, while k is the product the contract seeks to preserve after each trade. If a trader adds one asset to the pool and removes the other, the reserves change in a way that moves the quoted exchange rate. The larger the trade relative to the pool's available liquidity, the more its execution price moves along the curve.

That movement is price impact. It is often described loosely as slippage, although the two ideas are not identical. Slippage is the difference between the expected execution price and the price the trader actually receives; price impact is the portion of that difference caused by the trade itself changing the pool's reserve ratio. Neither term should be confused with the swap fee.

The fee is a separate charge specified by the protocol or by the particular pool tier. It is collected from the trade and, depending on the AMM's design, distributed to liquidity providers, directed partly to a treasury, or allocated according to another governance-defined rule. Price impact does not become LP revenue. It is the economic effect of moving through the pool's pricing curve. The fee is the explicit compensation for using the pool's liquidity.

Liquidity providers are the parties who deposit the reserves in the first place. In a traditional two-asset pool, an LP generally contributes both tokens in proportions determined by the pool's current composition. In return, they receive a claim on the pool's assets and fee growth. Some AMMs represent that claim through transferable LP tokens; others use account balances or non-fungible position receipts, particularly when liquidity is concentrated within a selected price range.

Those receipts or position claims can sometimes be held, transferred, or redeposited into another protocol. This composability gives the broader yield-farming stack its layered character. A user may provide liquidity in one contract, deposit the resulting LP token into a rewards market, and then place the additional reward token into an auto-compounding vault. Each layer can add potential return, but it can also add smart-contract risk, liquidation risk, withdrawal restrictions, and additional fees.

Yield farming is not a single product. It is a way for capital to find productive work inside an otherwise idle ledger.

The LP's core earnings come from their share of eligible trading fees paid by swaps through the pool. On Uniswap V2-style contracts, that share is generally proportional to the provider's liquidity relative to the pool, subject to the protocol's fee design and the timing of deposits and withdrawals. If trading volume rises while the pool's liquidity remains relatively stable, fee revenue per unit of liquidity may increase. If liquidity grows faster than volume, the same activity is divided among more providers.

That relationship is the first calculation an LP should understand. A high advertised yield does not necessarily mean that a pool has strong underlying demand. It may simply have a small liquidity base, a temporary incentive program, or a reward token whose market price has risen sharply. Fee-derived yield is connected to the protocol's utility: traders pay to access liquidity, and part of that payment can flow to the people who supplied it.

Yield farming, staking, and lending are not the same position

The language of passive income can make several different strategies sound interchangeable. They are not.

Staking usually involves committing a network's native asset to help secure or validate a proof-of-stake blockchain. The reward is tied primarily to the network's issuance policy, transaction fees, validator economics, or penalties. The staker is not necessarily providing an exchange market for traders.

Lending involves depositing assets into a money market where borrowers pay interest for access to capital. The primary variable is borrowing demand, along with utilization, collateral rules, and the possibility of bad debt or protocol failure.

Liquidity provision involves quoting an exchange between assets. The provider earns fees when traders use that liquidity, while accepting the possibility that the pool's rebalancing will leave them with a less favorable asset mix than if they had simply held the tokens.

All three can be described as yield-generating activities. Their return sources and risks are different enough that comparing only their headline APR is nearly meaningless.

Evolution from DeFi Summer to the $94 Billion Ecosystem

The current architecture of yield farming was not designed in a single act. It emerged through a sequence of incentive experiments, the most consequential of which occurred during the summer of 2020. That period, quickly named DeFi Summer, was triggered in part by Compound's distribution of COMP governance tokens to users who lent or borrowed on the platform.

The mechanism was simple enough to understand immediately: deposit assets, use the platform, and receive the protocol's token in addition to any native interest. The strategy was no longer limited to earning the revenue produced directly by a financial service. Protocols could also direct their own token supply toward the users they wanted to attract.

Annualized returns on major platforms briefly exceeded 1,000% during this period. Such figures pulled capital into the sector at remarkable speed, but they also exposed the fragility of emission-driven yield. A reward denominated in a volatile governance token can rise or fall independently of the activity that supposedly supports it. A pool can show an impressive annualized rate while its underlying fee revenue remains modest or its reward token loses liquidity.

The first wave of yield farming also established a vocabulary that remains in use: liquidity mining, farming multipliers, reward gauges, LP tokens, harvests, vaults, and composability. These were not just marketing terms. They described a new distribution system in which protocols competed for liquidity and users moved capital between venues in response to changing incentives.

What followed was a long consolidation. Exploits drained several large pools, regulatory scrutiny increased, and many of the least sustainable strategies unwound. Some protocols retained meaningful activity after incentives declined; others had attracted capital only for as long as the reward rate exceeded the perceived risk and transaction cost.

The approximately $94 billion in total value locked across DeFi by early 2026 is therefore qualitatively different from the liquidity that poured into the sector during the first incentive rush. The composition of that capital matters more than the aggregate number. Some of it supports swaps, lending markets, derivatives settlement, staking, and stablecoin infrastructure. Some remains dependent on token incentives. TVL measures how much capital is present, not whether that capital is generating durable revenue or whether it can leave without creating a disorderly unwind.

Yield farming today is best understood as the operational layer beneath that TVL: the set of strategies by which depositors align their capital with a protocol's transactional function. The stronger strategies make the source of return legible. The weaker ones hide it behind an annualized number.

Navigating Impermanent Loss and Capital Efficiency

The most misunderstood cost of providing liquidity is not the protocol fee, nor necessarily the gas required to enter a position. It is impermanent loss, or IL: the divergence in value between depositing two assets into an AMM and simply holding those same assets outside the pool.

IL arises because the constant-product formula continuously rebalances the LP's position as prices move. When one asset outperforms the other, arbitrage traders adjust the pool until its price is brought back in line with the wider market. In the process, the pool sells some of the appreciating asset and accumulates more of the depreciating one. The LP ends up with more of the underperforming asset and less of the outperforming asset than they would have held by simply keeping the original pair in their wallet.

For a standard 50/50 constant-product pool, the relationship between price movement and IL is mechanical:

These figures follow from the constant-product relationship:

IL = 2 · √price ratio / (1 + price ratio) − 1

The result is negative because the comparison is against simply holding the assets, not because the LP necessarily loses money in absolute terms. Trading fees and incentives may more than compensate for the divergence. Conversely, a pool can generate fee revenue and still underperform a passive holding strategy if the price move is large enough.

The label impermanent is precise in one sense: if prices return to their original ratio, the divergence can disappear on paper. But if the LP withdraws while the ratio remains different, the relative loss is realized. Waiting does not erase IL by itself. Only a return toward the original price relationship, together with the pool's fee income and any other rewards, can change the outcome.

Impermanent loss is not a fee charged by the protocol. It is the structural cost of providing liquidity through a rebalancing mechanism.

The asset pair matters. Two stablecoins that remain close to their intended values may experience limited divergence, although depegging and contract risk remain real. A volatile asset paired with a stablecoin can generate substantial trading fees during active markets, but it also exposes the LP to one-sided price movement. Two volatile assets may move together, reducing relative divergence in some periods, or separate sharply in others.

Concentrated liquidity changes the position

The introduction of concentrated liquidity on platforms such as Uniswap V3 changed the geometry of the problem. Instead of spreading capital across the entire price curve, an LP can allocate liquidity within a selected range. When trades occur inside that range, the position can generate more fees per unit of capital than a comparable full-range position.

That capital efficiency is conditional. If the market price moves outside the selected bounds, the position becomes one-sided and stops earning fees until the price returns to the range. The LP may need to rebalance or select a new range, which turns liquidity provision into an active management strategy rather than a purely passive deposit.

The trade-off can be expressed simply:

• Wider range: less capital-efficient, but more resilient to price movement and less demanding to manage.
• Narrower range: potentially higher fee income while active, but more exposed to going out of range.
• Stable or correlated pair: generally lower relative-price risk, but potentially lower fee rates and different depeg risks.
• Volatile pair: greater opportunity for fee generation, alongside a larger possibility of impermanent loss.

A V2 LP is, in effect, a passive participant in price discovery across the available curve. A concentrated-liquidity LP is an active underwriter of a specific price corridor. The higher apparent efficiency is not free; it is purchased with a greater need to forecast volatility, volume, and the future location of the market price.

Sustainable Yields vs. Temporary Token Emissions

Not all advertised yield is the same. The headline figure on a liquidity pool is usually a composite of several components: trading fees earned by LPs, lending or staking income where relevant, and tokens distributed as an additional incentive.

The fee component is tied to trading activity and the rules of the pool. It can rise when volume increases and fall when demand disappears. It is not automatically stable, guaranteed, or risk-free. A volatile market may produce substantial volume for a short period and then become quiet. A pool may also lose activity to a competing venue or suffer from an adverse shift in its asset pair.

The incentive component has a different economic character. Rewards may consist of newly issued tokens, tokens released from an existing treasury, tokens already circulating but allocated to a rewards program, or some combination of these sources. The relevant question is not simply whether tokens are being distributed. It is how the distribution is funded, who bears the dilution or opportunity cost, and what rules govern its duration.

An emission schedule may be short, long, renewable, governance-adjusted, or dependent on a continuing treasury allocation. It may be reduced over time, maintained at a target rate, redirected between pools, or terminated. None of those outcomes should be assumed in advance. A reward program can be temporary, but it is not necessarily temporary merely because it is paid in a protocol token.

The arithmetic of emission-funded yield still requires care. If rewards are paid in a liquid token, the recipient may be able to sell them, but selling creates market pressure. If the token comes from an existing treasury, the distribution may not increase total supply immediately, yet it still transfers value or control from the treasury to recipients. If the token is newly issued, the distribution can dilute existing holders. These are different mechanisms, but none should be treated as free income.

A high reward rate can also attract more liquidity. As the pool grows, the same reward budget is divided among more participants, reducing the yield per unit of capital unless the incentive rate changes. The token's market price may move in the opposite direction as farmers sell rewards to recover their original capital or convert income into a preferred asset.

This is why fee revenue and token incentives should be separated before evaluating a strategy. Fee revenue indicates that users are paying for a service. Incentives indicate that someone is trying to influence where liquidity goes or how users behave. A protocol may rationally use incentives to bootstrap a market, reward governance participation, or subsidize a public-good function. That does not make the rewards fraudulent. It does mean the yield depends on a policy decision rather than solely on transaction demand.

The difference between productive yield and speculative carry is visible in what happens when the reward program changes. If a protocol continues to attract traders, borrowers, or stakers after incentives are reduced, its underlying service may be valuable. If the capital leaves immediately, the previous return was largely a payment for temporary participation. Both models can be profitable during the right period. Their risk profiles are not comparable.

Distinguishing APR from APY in Compound Interest Strategies

A final layer sits on top of the yield calculation itself: the difference between Annual Percentage Rate, or APR, and Annual Percentage Yield, or APY.

APR represents a simple annualized rate. It describes the return over a year without assuming that rewards are reinvested. APY incorporates compounding and therefore assumes that earned rewards are periodically added back to the position. The difference is not cosmetic, especially when the displayed rate is high or the strategy compounds frequently.

For a strategy where rewards sit idle until manually claimed, APR is generally the more useful starting point. For an auto-compounding vault that harvests rewards, converts them into the required asset, and redeposits them, APY may better describe the intended mechanism.

But APY is not a promise of realized performance. The calculation may assume that the displayed rate remains constant, that every harvest occurs at the same cost, that the reward token keeps its price, and that the strategy can reinvest without material slippage. Those assumptions can fail one by one.

Yield aggregators exist precisely to automate this process. They collect rewards from an underlying pool, swap them into the required token or token pair, and redeposit them, sometimes several times per day. The additional return comes from compounding, but it is reduced by gas costs, swap costs, vault fees, and the market impact of converting rewards. The vault also introduces another smart contract and another layer of strategy logic.

The gap between APR and APY widens with both the compounding frequency and the underlying rate. At moderate returns, the difference may be limited. At the triple-digit APYs occasionally seen in newly launched incentive programs, the difference can look dramatic, but that result is often more sensitive to changing token prices and reward schedules than to the mathematics of compounding itself.

For an LP evaluating a strategy, the relevant calculation is not the largest number on the dashboard. It is the expected net return after:

• the pool's trading-fee revenue;
• token incentives and their funding source;
• impermanent loss or one-sided price exposure;
• gas, swap, and withdrawal costs;
• vault or performance fees;
• changes in liquidity and reward allocation;
• the probability that a smart-contract or oracle failure interrupts the position.

A strategy that compounds several times per day may outperform manual harvesting when the position is large enough and the underlying yield is durable. On a smaller position, the same automation can spend too much of the return on execution costs. Compounding is a mechanism, not a source of value by itself.

A Structural Question for the Next Cycle

Yield farming, viewed from the level of protocol architecture, is a coordination mechanism. It aligns capital with transactional need and compensates capital providers through a combination of fees, interest, network rewards, and token incentives.

Its mechanics are now familiar: AMMs use liquidity pools to facilitate swaps; LPs earn a share of eligible fees; concentrated liquidity makes capital more efficient while increasing management demands; lending markets pay depositors for supplying credit; staking directs assets toward network security; and vaults automate the reinvestment of returns. None of these mechanisms removes risk. They distribute it differently.

The most important distinction is between a return generated by activity and a return generated by distribution policy. A swap fee exists because a trader paid to use liquidity. A lending rate exists because a borrower demanded capital. A staking reward follows the rules of a network or protocol. A token incentive may be an effective way to bootstrap participation, but its durability depends on the source of the tokens, the schedule, governance, and the market's willingness to value them.

The approximately $94 billion deployed across DeFi by early 2026 demonstrates that decentralized financial infrastructure has moved beyond a single speculative experiment. It does not demonstrate that every pool, vault, or reward program is sustainable. Aggregate TVL cannot show whether capital is earning real fees, whether positions are exposed to impermanent loss, or whether liquidity will remain after incentives change.

That is the question worth holding open as the ecosystem matures: can protocols retain capital through the revenue their services generate, rather than through the size of the reward they distribute? The answer will not be found in the highest APY on a dashboard. It will be found in the composition of returns, the resilience of the underlying activity, and what happens when the incentive program changes.