Liquidity provision on Curve: how stable-to-stable pools earn fees with low impermanent loss

Most DeFi yield writing focuses on lending. Curve is the other major primitive in the vildX strategy, and it's a different kind of income source entirely. Instead of earning interest from borrowers, an LP earns trading fees from swappers. Instead of a utilization curve, the income comes from volume. Instead of credit risk, the dominant risk is impermanent loss.

For stable-to-stable pools, the IL risk is small enough that LPing is one of the cleaner yield sources in DeFi. This piece explains why.

What an AMM is, briefly

An automated market maker is a swap venue where prices are set by a formula rather than an order book. The simplest example is Uniswap V2's constant-product formula: x × y = k. The pool holds two tokens; their quantities multiply to a constant; the price of one in the other is determined by their ratio.

When someone swaps token A for token B, they add to the A side and remove from the B side. The ratio shifts. The price moves. The pool's invariant is preserved.

Curve's innovation was to design an invariant specifically tuned for assets that should trade at near-parity — like USDC/USDT or DAI/USDC/USDT. In those pools, the curve is much flatter than a constant-product around the parity point, which means swaps cause less price impact and the pool can quote tighter spreads.

Where the yield comes from

Three sources, in roughly descending order:

1. Swap fees. Every swap pays a small fee (usually 4 basis points for stable pools). The fee is added to the pool, increasing each LP's share of the underlying.
2. CRV emissions. Curve's governance token is emitted to LPs in pools that gauge weights direct to. The emissions are real yield; their dollar value depends on the CRV price.
3. Boosted yield from veCRV. Locking CRV grants voting power and a yield boost on LP positions. Larger holders run vote-buying markets (Convex, Yearn, etc.) to extract this efficiently.

For a managed strategy, the meaningful sources are swap fees and the dollar-weighted value of CRV emissions. We don't run a veCRV position directly; we capture boost indirectly by routing through aggregators where appropriate.

What impermanent loss actually is

Impermanent loss is the gap between holding two tokens passively and providing them as liquidity in an AMM. It exists because, when the relative price of the two tokens moves, the AMM rebalances toward the side that has fallen — exposing the LP to more of the falling asset.

For a pool of ETH and USDC, IL can be severe. If ETH doubles in price, an LP ends up with more USDC and less ETH than they started with — meaningfully underperforming a hold-and-do-nothing strategy. The IL is the cost the LP pays for the fees they earned.

For a pool of USDC and USDT, the math is the same but the magnitude is tiny. Both assets trade near $1.00. The pool rarely rebalances much because the price ratio rarely moves much. The most extreme historical events — USDC's depeg to $0.88 in March 2023, USDT's brief move to $0.95 in May 2022 — produced visible but bounded IL events. Most of the year, stable pools sit in a tight range and the LP captures fees with negligible IL.

A concrete example

Suppose you supply $10,000 to a 3pool position (USDC/USDT/DAI):

• Pool fee: 4 bps (0.04%).
• Daily volume on the pool: $200M (illustrative).
• Pool TVL: $250M.
• Your share: $10,000 / $250M = 0.004%.
• Daily fees: $200M × 0.0004 × 0.00004 = $32.

That's about $32/year on $10,000 from fees alone, or ~0.32% APY. That sounds low — and it is, by itself. The other components (CRV emissions, gauges, boosts) typically add 1–4% on top, bringing the total to a competitive range for a stable position with minimal IL.

The exact numbers shift constantly. The headline rate on Curve pools moves with volume, with CRV's dollar price, and with the gauge weights set by governance. We allocate to Curve when the blended rate makes the position competitive with lending; when it doesn't, we rotate weight back to lending.

When stable LPing breaks down

Two scenarios where the math stops being clean:

1. One of the pool assets actually depegs. If USDT broke its peg in a sustained way, the 3pool would end up holding mostly USDT (everyone would swap out of USDT into the other assets, which the AMM rebalances by absorbing the USDT they're dumping). The LP would be left holding the depegged asset. This is the largest tail risk in stable LPing.
2. The pool composition drifts persistently. If one asset trades consistently above $1.00 and another below, the pool ends up overweight the cheaper one. Fees can compensate for a while; if the drift is structural, the LP slowly bleeds.

The vildX strategy manages both risks by:

• Allocating to Curve pools where all underlying stablecoins have strong issuers and active redemption mechanics.
• Capping the Curve weight at 15% of total strategy, so a stress event in a pool is bounded.
• Monitoring pool composition daily; the strategy rebalances away from a pool that drifts past a threshold.

How LPing compares to lending

For a yield strategy, lending and LPing are complementary, not interchangeable.

• Lending pays a smoother yield, scales linearly with utilization, and the failure modes are credit-shaped (bad debt, oracle issues).
• LPing pays a more volume-sensitive yield, has IL as the primary cost, and the failure modes are price-shaped (depegs, persistent drift).

Holding both in a blend produces a return profile that's less correlated with any single primitive. When borrowing demand on Aave is low, lending APRs sag — but Curve volume might be steady because swaps don't depend on the same demand cycle. The blend captures both.

If you want to see how the full strategy mixes these together, the case for a blended multi-protocol stablecoin strategy walks through the allocation logic end to end.