Curve has done something rare in DeFi over the past three years: stayed out of the headlines while the protocols around it imploded, pivoted, or rebranded. The panda has been watching crvUSD quietly compound, and the mechanism behind it deserves a fresh audit before another vault economy claims to have reinvented credit.
According to DefiLlama's chains dashboard, total DeFi TVL sits at $71.21B on June 12, 2026, with Ethereum at $37.26B. Curve's crvUSD is a small slice of that, but a structurally interesting one. It is the only major decentralized stablecoin that does not liquidate borrowers in one shot. Instead, it gradually converts collateral to crvUSD as the price drops, then back to collateral as the price recovers. That choice has consequences, good and bad, and most coverage skips straight to the chart.
What Is LLAMMA and How Does Curve Soft-Liquidate?
LLAMMA stands for Lending-Liquidating AMM Algorithm. The name is ugly. The idea is interesting. According to Curve's official LLAMMA documentation, a borrower's collateral, typically ETH or wstETH, is split across a range of price bands. When the collateral price drops into a band, the protocol's internal AMM swaps the collateral in that band for crvUSD. When the price recovers, the AMM swaps the other way.
Three things follow from that design. First, there is no all-or-nothing liquidation moment. The borrower never gets wiped by a 10 percent flash candle. Second, the borrower realises losses gradually, as a worsening collateral-to-debt ratio, instead of a sudden zero. Third, the protocol carries the price-impact cost of running its own internal AMM, paid in worse fills than an external DEX would have produced.
Compared to Aave's hard-liquidation model and Morpho's curator-driven markets, LLAMMA is a different bet entirely. Aave optimises for solvency under stress. LLAMMA optimises for borrower survival through volatility. Both are defensible. Neither is free.
crvUSD's Quiet Run Through 2026
According to DefiLlama's Curve DEX protocol page, Curve's combined TVL across pools and lending markets sits comfortably in the multi-billion-dollar range as of mid-June 2026. The crvUSD lending markets, tracked separately under LlamaLend, continue to absorb wstETH, sfrxETH, and tBTC as primary collateral types.
The growth story is unspectacular, which is exactly the kind of compounding worth watching. No incentive farms, no points campaign, no airdrop carrot. crvUSD's expansion has been driven by the simple fact that a borrower who collateralises wstETH can earn the underlying staking yield while paying a borrow rate set by Curve's PegKeeper-controlled monetary policy.
For context on how decentralized stables are being repositioned around yield this cycle, the Sky Protocol USDS savings rate piece covers the supply-side push. crvUSD takes the opposite route. It does not pay holders. It pays borrowers, in the form of softer liquidations and lower forced-sale slippage. Same problem, different side of the table.
Coverage from The Defiant on Curve's stablecoin evolution has noted that crvUSD has not depegged meaningfully since launch. The PegKeeper system, four contracts that mint and burn crvUSD into Curve pools, has handled minor deviations without external intervention. The numbers say yes. The mechanism is doing its job.
Risks: Oracle, Governance, and veCRV Concentration
A defi-watch piece without a risks section is marketing. Here are four that matter.
Oracle risk is the first. LLAMMA's soft-liquidation depends on a price feed. Curve combines Chainlink with its own EMA-smoothed pool prices to limit oracle manipulation. The smoothing reduces flash-loan attacks. It also means that during a fast crash, LLAMMA's view of the price lags reality, and the AMM converts collateral at stale prices. Borrowers near the bottom band can end up holding more crvUSD debt than the spot value of their original collateral. This is the bad-debt scenario that hard-liquidation systems are built to avoid.
Governance risk is the second. crvUSD's parameters, including interest rates, debt ceilings per collateral, and the PegKeeper pool list, are set by veCRV holders. Voting power is concentrated in a small set of lockers, with Convex chief among them. A capture of veCRV by a hostile or merely incompetent voter base could push parameters into a state that bleeds the protocol. This has not happened. It also has not been stress-tested at scale since the 2023 founder-loan saga.
Smart-contract risk is the third. The LLAMMA contracts are audited and battle-tested, but the design is novel by 2026 standards. Any bug in the band-pricing logic does not produce a single bad liquidation. It produces a structural mispricing that compounds for as long as nobody notices.
Stablecoin-collateral risk is the fourth. The PegKeeper pool list assumes a healthy counterparty stable on the other side. If USDC or USDT ever loses its peg materially, the PegKeeper mechanism inherits that loss. The crvUSD float would be the canary.
For a wider view of how DeFi accounting itself is now stretched, the DeFi TVL triple-count piece covers the broader credibility problem. crvUSD is not immune to that critique. It does, however, count cleanly: collateral in, debt out, no looped LST-LRT-lending exposure inside its own books.
What to Watch Next
Three things over the next quarter. First, whether the PegKeeper system handles a real USDC or USDT scare without breaking. Second, whether crvUSD's borrow rate stays competitive when USDS, GHO, and USDe all chase the same wstETH-collateral borrower. Third, whether Curve ships LLAMMA v2 with the cross-margin and multi-collateral hooks that have been on the DeFi pillar roadmap discussion since late 2025.
Curve is no longer the protagonist of the 2026 DeFi story. That is fine. Boring infrastructure that survives without rewrites is the whole point. The panda watches, the panda judges, and the verdict on crvUSD is a slow, deliberate nod.
