The simplest conclusion to draw from the $290 million exploit and the subsequent $13 billion decline in DeFi's total value locked is that decentralized finance has failed once again. However, this perspective is likely overly simplistic. The KelpDAO exploit, which occurred over the weekend, was indeed severe and appears to have originated from a targeted attack on the infrastructure used in LayerZero's verification stack, rather than a smart contract bug.
According to preliminary findings by LayerZero, the incident can be linked to North Korea's Lazarus Group, and the attack succeeded because Kelp had opted for a single-verifier setup despite repeated recommendations to use a more secure configuration. The exploit resulted in rsETH, a liquid staking token issued by KelpDAO, becoming unbacked, triggering concerns that bad debt would spread to lending markets, particularly Aave's WETH pool. Nevertheless, the more intriguing story is not the fact that DeFi was attacked, but rather that DeFi remains operational. In the aftermath of the breach, capital fled rapidly, with Aave alone experiencing $8.45 billion in outflows over a 48-hour period, while DeFi's total value locked plummeted to the mid-$80 billion range, roughly equivalent to its value at this time last year.
This sharp repricing of risk, rather than a catastrophic collapse, is a more accurate description of the situation. Aave, the largest DeFi lending market, had accumulated significant amounts of rsETH as collateral in the weeks leading up to the exploit, as users built leveraged positions. The scale of the TVL drop also warrants some context, as a $292 million theft does not directly result in a $13 billion decline unless a substantial portion of that TVL was already recycled collateral.
Much of Aave's ETH exposure prior to the weekend was concentrated in looping strategies, where users deposit liquid restaking tokens, borrow ETH against them, swap for more restaking tokens, and repeat. This leverage inflates TVL on the way up and unwinds sharply during events like this, making the actual net capital loss likely a fraction of the headline figure. Those strategies were themselves partly a product of a yield environment that had already stopped making sense, with Aave offering 2.61% APY on USDC deposits as of early April, below the 3.14% available on idle cash at Interactive Brokers.
The risk premium that historically justified DeFi's complexity and smart contract exposure had largely disappeared, and with organic yield insufficient, leverage filled the gap. Crypto has survived worse incidents, including the collapse of Terra, which vaporized confidence across the sector, and the losses suffered by Wormhole and Ronin, which totaled roughly $1 billion each. More recently, Bybit suffered a significant loss of around $1.5 billion, yet it continued to operate and process a surge in withdrawals. The repricing of trust is a significant but unlikely existential loss, according to 0xNGMI, founder of DefiLlama.
Aave has many resources to cover the loss, including its treasury and taking loans, and those will have to be used to protect the protocol. The biggest issue will be the impact on risk premiums assigned to DeFi, which are a real and lasting cost.
Capital will demand more compensation for participating in on-chain systems whose attack surface now extends beyond code. Still, repricing is not the same as collapse, and some of the money will return, as seen before in Aave when rumors of a hack appeared. There is also evidence that capital is not simply leaving DeFi but rotating, with Spark's TVL jumping from $1.8 billion to $2.9 billion over the weekend, demonstrating clear capital rotation.
The more interesting critique is not that DeFi failed but that it has become too timid, and if the sector is going to ask users to bear infrastructure risk, smart contract risk, and governance risk for low single-digit yields, the product set starts to look less compelling. With that in mind, Kelp is not the end of DeFi but a wake-up call for builders to build safer systems while continuing to offer real-world use cases.