Yield farming is the practice of putting crypto capital wherever it earns the most, and moving it when somewhere else earns more. In practice that means supplying liquidity to an exchange, lending to a money market, staking, or depositing into a vault that does one of those things, and collecting whatever combination of fees, interest and token rewards results.
Where the practice came from
Compound began distributing its COMP governance token to users in June 2020, paying it to both lenders and borrowers in proportion to their activity. The distribution turned borrowing into a profitable act, because the token received could exceed the interest paid, and capital arrived in volume. The period that followed became known as DeFi summer, and the pattern it established has repeated in every cycle since.
That pattern is mercenary capital. Incentive programs attract deposits quickly and lose them just as quickly, because the capital was never attached to the protocol. A protocol that launches with a subsidized rate is buying attention and hoping some fraction of the users stay once the subsidy ends. Most do not.
The two kinds of yield
Every farming return decomposes into revenue the protocol actually earned and tokens the protocol newly issued. The distinction determines whether a rate can persist.
| Source | Paid out of | Persists while |
|---|---|---|
| Trading fees | Swap volume through a pool | Traders keep trading that pair |
| Lending interest | Borrower payments | Someone wants leverage or working capital |
| Staking rewards | Network issuance and transaction fees | The underlying network keeps paying validators |
| Funding payments | Perpetual futures longs paying shorts | The derivatives market leans long |
| Token incentives | Newly issued governance tokens | The emissions schedule runs and the token holds value |
The first four are described collectively as real yield. They are constrained by how much economic activity the protocol supports, which is why they rarely reach spectacular levels. Stablecoin lending on major venues paid roughly 3.6 percent on Aave's Ethereum USDC market and about 5 percent on a large institutional credit venue on 8 September 2026, according to DefiLlama. Ether staking paid a low single-digit rate on the same date.
Incentive yield has no such constraint, because a protocol can print as many tokens as it likes. A 40 percent advertised rate composed mostly of emissions is a statement about the emissions schedule and the token's current price, not about the protocol's economics. When emissions taper or the token falls, the rate falls with it, and the capital leaves.
Impermanent loss, properly
Providing liquidity to an automated market maker carries a cost that is easy to overlook because it never appears as a transaction. As the relative price of the two pooled assets changes, the pool automatically sells the appreciating asset and accumulates the depreciating one. The result is a position worth less than simply holding both tokens.
The standard formula for a constant-product pool gives impermanent loss as 2 times the square root of the price ratio, divided by one plus the price ratio, minus one. A worked example makes the size concrete. Deposit 1 ether at $3,000 alongside 3,000 USDC, a $6,000 position. If ether doubles to $6,000, holding the two assets separately would be worth $9,000. The pool position rebalances to roughly 0.707 ether and 4,242 USDC, worth about $8,484. The shortfall is roughly 5.7 percent.
Two properties matter. The loss is symmetric, so a halving produces the same percentage shortfall as a doubling. And the loss is only impermanent in the sense that it reverses if prices return to their starting ratio, which for a pair that has genuinely repriced they will not. Fees can outweigh the drag in a busy, range-bound market and frequently do not in a trending one.
How the practice evolved
Simple liquidity mining gave way to vote-escrow designs, pioneered by Curve, in which users lock the governance token for years in exchange for voting power that directs where emissions flow. That created a secondary market: protocols wanting emissions directed to their pool would pay locked-token holders to vote that way, an arrangement that became a substantial industry in its own right.
From 2023 onward, points programs largely replaced direct emissions. A protocol awards points for deposits, promises that points will matter at a future token launch, and specifies nothing binding. Capital arrives on the strength of an expectation. The mechanism is cheaper for the protocol than emitting tokens and correspondingly less certain for the depositor, who is accepting an unpriced claim in exchange for a real deposit.
Where farming capital actually sits now
Most farming today happens through vaults rather than through positions a user opens directly. A depositor puts stablecoins into a vault, and a curator or strategist allocates them across lending markets, adjusting as rates move. The depositor has outsourced the farming rather than stopped doing it, and the fee they pay is the difference between the gross yield and what lands in their share price.
That shift changes which questions matter. Selecting pools yourself makes the pool the unit of analysis. Depositing into a vault makes the operator the unit of analysis, because the operator will move the capital between pools without asking. The vault and curator guides cover how to evaluate that layer.
The risks that recur
Farming risk is cumulative, because a farmed position usually touches several protocols at once. A vault that supplies a liquidity pool that is itself deposited into a lending market exposes the depositor to every contract in the chain, and to the possibility that the composition breaks in a way none of the individual protocols anticipated.
- Smart contract risk at every layer the position passes through, not just the front one.
- Impermanent loss on any position that provides two-sided liquidity.
- Emission decay, where the advertised rate falls sharply once incentives taper.
- Token price risk on rewards that must be sold to be realized.
- Exit risk, where the position cannot be unwound at the assumed price because everyone is leaving at once.
A useful discipline is to write down, before depositing, what the yield would be with incentives removed and what the position is worth if the least liquid asset in the chain falls by half. If both answers are acceptable, the position is being entered on its economics rather than on its advertised rate.