DeFi TVL slides 14% as KelpDAO exploit chills risk

DeFi’s total value locked has fallen 14 per cent to $148 billion since the KelpDAO exploit, turning what first looked like a protocol breach into a broader test of crypto risk appetite.

According to The Block’s May 26 report, the drop from about $172 billion followed an April 18 attack that reportedly drained $293 million through KelpDAO’s LayerZero bridge. Five weeks later, capital had not snapped back. That is the tell. Bull-market crypto often treats hacks as isolated once reimbursements, patches or post-mortems arrive. This breach is being priced more like a balance-sheet shock to the plumbing beneath restaking and cross-chain yield.

For DeFi investors, the live question is not only whether KelpDAO recovers. It is whether the next marginal dollar wants bridge exposure at all when token prices, validator assumptions and off-chain verification can become part of the same risk trade. DefiLlama’s hacks database now reads less like a crime blotter than a capital-allocation screen.

Capital is voting with its feet

Start with the simplest read: TVL moved lower and stayed there. A 14 per cent sector drawdown is large enough to matter even in crypto, where collateral moves fast and denominator effects can exaggerate the shift when token prices fall. In this case, the timing puts the KelpDAO breach near the centre of the repricing.

Analysts trying to separate a broad risk-off move from a protocol-specific scare do not yet have a perfect category map from the public numbers. That missing map matters. Concentrated outflows from restaking, bridges and lending protocols with cross-chain assumptions would send one message. A slide spread across DeFi would send a harsher one: investors are applying a security discount to the whole stack.

CoinDesk offered the other reading. Its analysis of the TVL drop framed the decline as a possible stress test rather than a thesis-breaker, with DeFi Technologies president Andrew Forson pointing to Treasury-backed collateral behind USDT and USDC. In that builder-optimist version, risk is migrating rather than disappearing, and stronger collateral rails could absorb the shock.

For now, the market is taking the colder side. DeFi users are not just marking down a token. They are reassessing whether yield from complex routing is worth the tail risk that one weak verification path can erase months of returns. That makes this a different kind of drawdown.

The bridge problem

KelpDAO is uncomfortable because the weakness described by researchers was not a simple smart-contract bug that could be patched and forgotten. The Block reported that attackers manipulated RPC nodes and validators connected to the LayerZero bridge. In plain English, the exploit sat in the infrastructure that tells protocols what has happened elsewhere.

Bridge risk keeps returning as a market-structure issue for that reason. A lending protocol can disclose loan-to-value ratios. A decentralized exchange can show liquidity depth. A bridge has to convince users that messages, validators, relayers and node infrastructure are all telling the same truth at the same time. When one layer goes wrong, the user sees a single loss.

“all of DeFi” unsafe

Manuel Aráoz, OpenZeppelin co-founder, quoted by The Block

Aráoz’s line in The Block’s interview was deliberately blunt. It also captured the asymmetry security teams are facing. Composability is the user promise, but the risk model is cumulative. Every extra protocol, bridge, oracle, validator path and wallet permission adds another surface that has to work.

Skeptics land on a harder question from there. If attackers need only one weak verification route, better code review by itself is not enough. The upgrade has to touch monitoring, key management, validator design, circuit breakers and the incentives for pausing flows before losses compound. Otherwise, DeFi’s security stack remains a collection of strong components joined by fragile assumptions.

April changed the baseline

KelpDAO’s loss also arrived in a rough month for DeFi security. Aráoz cited nearly $630 million stolen across 27 reported exploit cases in April, making it the worst month for hacks since early 2025. The number was large. More important, it was clustered.

Clustered losses change how investors think. One breach can be dismissed as a one-off failure of diligence. A month of repeated incidents gives capital a reason to leave before the next post-mortem arrives. Yield markets are especially vulnerable to that reflex because the incremental spread looks attractive until the hidden insurance cost appears all at once.

Follow-on examples have not helped. The Block later reported that the Cosmos-based Gravity Bridge was drained of $5.4 million in a suspected key compromise. The amount was much smaller than KelpDAO’s loss, but the root concern rhymed: bridges remain a repeated point of failure.

“Coding agents are superhuman at finding vulnerabilities, and smart contract security is too asymmetric,”

Manuel Aráoz, OpenZeppelin co-founder, quoted by The Block

That quote cuts two ways. AI-assisted audit tools may find bugs faster, but attackers can use the same speed. Defenders still have to be right across the full system. An attacker needs one overlooked assumption, one compromised key, one stale dependency or one validator path that does not behave as expected. The balance of labour still favours the person looking for the crack.

Security firms can narrow that gap, but they cannot erase it for a sector built on permissionless routing. A more useful near-term benchmark is whether protocols can make losses smaller and slower once something breaks. Investors will look for withdrawal throttles, independent monitoring and public incident dashboards before they treat the April cluster as last month’s problem.

What brings capital back

Capital probably needs more than a better press release from KelpDAO before it returns. It needs a clearer risk price. Traditional markets recover from operational failures when investors can see who absorbed the loss, what changed in controls and whether the same route can fail again. DeFi often gives users a faster post-mortem but a blurrier liability chain.

The builder-optimist case is still credible. Stress can force better design. Protocols can harden bridge verification, publish clearer dependency maps, cap exposure to any one cross-chain route and use monitoring that turns abnormal validator or RPC behaviour into an automatic brake rather than a Telegram alert. Those are not cosmetic upgrades. They mark the difference between composability as an asset and composability as contagion.

A portfolio-level answer is likely, too. Investors may not abandon DeFi outright. They may demand simpler yield, shorter lock-ups, stronger collateral and less exposure to restaking strategies whose risk depends on several systems behaving correctly at once. That would leave DeFi functioning, but with a lower multiple on complexity.

KelpDAO’s exploit is now bigger than KelpDAO. A five-week, $24 billion fall in TVL says security has become a funding cost. Protocols that can prove their controls may still attract liquidity. Those asking users to trust opaque infrastructure will pay for that trust in lower deposits, higher incentives or both.

Near-term flows will show whether the market is already differentiating between protocols with visible controls and protocols still leaning on brand trust. Until then, DeFi’s recovery narrative is capped by a harsher question than token prices: whether the yield is high enough to compensate for infrastructure that can fail outside the smart contract everyone thought they were auditing.

That is a colder DeFi market, not necessarily a dead one. It is also a market starting to ask the same question lenders ask after any credit event: what else is hiding in the collateral?