What restaking is, and why EigenLayer now secures $7 billion

Restaking lets people reuse the same staked Ethereum to help secure other services and earn extra rewards. EigenLayer pioneered it. The protocol, now part of a wider platform called EigenCloud, held about $7 billion in restaked assets on 27 September 2026, according to DefiLlama.

That total is down sharply from a peak near $20 billion in 2024. The fall is not a collapse. It followed two changes: EigenLayer switching on real penalties for bad behavior, and the free reward points that pulled in early money running dry.

How restaking works

Staking is the base layer. To help run Ethereum, holders lock up ETH and earn a yield, the reward at the heart of proof of stake. Restaking adds a second job for that same stake. Through EigenLayer, a staker can pledge locked ETH to back other services too, from data availability layers to price oracles, and collect fees from each.

These extra services are called actively validated services, or AVSs. Picture a guard who already watches one building agreeing to watch three more on the same shift, for extra pay. The upside is clear. So is the danger. If the guard slips at any single building, the pay turns into a fine.

Those services vary widely. Some post data so rollups can prove what they ran. Others feed prices to lending apps or timestamp events for other chains. Each one needs honest computers running it, and each pays the operators who put up restaked ETH as a bond. EigenLayer sits in the middle, matching stake to services and tracking who backs what.

None of this is free money. Base staking on Ethereum pays a low yield, a few percent a year, and restaking rewards sit on top as extra fees from the services an operator agrees to back. Early on, much of the draw was not fees at all. It was points, a promise of future tokens. Steady payment from live services is the part still being built.

Slashing turned the idea real

For its first two years, restaking ran without teeth. Operators could act badly and keep their collateral. Not anymore. On 17 April 2025 EigenLayer turned on slashing, the feature that lets a service seize an operator's funds when they break its rules. CoinDesk reported the protocol secured about $7 billion across 39 services at the time.

Sreeram Kannan, EigenLayer's founder, said the switch delivered the original pitch. 'We are happy to say now that the whole promise has been delivered,' he told CoinDesk. Real penalties changed the math for everyone holding a restaked position.

EigenLayer tried to fence the risk in. Operators can assign set portions of their stake to each service, so a penalty from one does not automatically drain the funds backing another. A separate 2025 upgrade cut the proof size and gas cost of managing many validators by up to 64 times, the company said, which makes running at scale cheaper.

A later step went further. In July 2025 EigenLayer made slashed funds reusable instead of burned, so a penalty can be routed to the users it was meant to protect. One lending protocol plans to use redirected funds as a backstop for stablecoin holders when an operator breaks a loan promise, the company said on its blog.

Why the money cooled

The retreat from $20 billion has a plain cause. Much of the first deposit wave chased points, a loyalty score early users expected to convert into free tokens. Once the airdrops landed and the points stopped, a lot of that capital moved on to the next incentive.

Live slashing did the rest. With penalties now real, stakers had to weigh each service's added risk against its reward. Backing more services means more ways to lose the stake, so some operators retreated to the risks they could actually judge.

The danger is not hypothetical. In April 2026 attackers drained about 116,500 units of rsETH, a liquid restaking token, worth around $292 million, from a bridge run by Kelp, then borrowed against the stolen collateral on Aave, according to CoinDesk. The loss rippled into other lending markets, and Aave's deposits fell by about $6.6 billion in the days that followed. Restaking spreads security. It can spread failure too.

Analysts had expected some of this. Before slashing shipped, the fear was that operators would back every service while there was no downside, then pull back once real punishment arrived and re-delegate their stake to match the true risk. That is roughly what happened. Lower, choosier deposits are the mark of a market pricing danger it once waved through.

The pivot to verifiable services

EigenLayer no longer sells itself as restaking alone. In 2025 the project folded its work into EigenCloud, a wider push to sell verifiable computing, where a task runs off the chain but anyone can check it was done right and slash whoever cheats. Its recent writing points hard at artificial intelligence. EigenCloud describes what it offers as verifiable execution, verifiable data, and slashing-backed enforcement for autonomous agents.

That framing answers a question the numbers raise. Restaked ETH is only worth something if real services want to pay for security. Data availability layers and price oracles were the first buyers. Whether AI agents that need provable, punishable computation become the next big one is the bet EigenCloud is now making, and it is far from settled.

EigenLayer is not the only contender either. Rivals such as Symbiotic and Babylon chase the same shared-security idea, yet EigenLayer still holds the largest pool by a wide margin, ranked first among restaking protocols by DefiLlama on 27 September 2026. Size brings its own weight. The more value one protocol secures, the more a single failure could shake.

What to watch

A bigger test is whether services on EigenLayer pay operators enough to justify the risk now that the risk is priced in. Reward data across the services is still thin, and it is not clear how much operators actually take home for the extra work. EigenCloud, the platform EigenLayer now sits inside, is betting the next wave of demand comes from verifiable services for AI agents, where a task can be checked and a cheat can be punished.

Watch three things. First, whether restaked value climbs back or settles at a lower, steadier level. Second, whether a big service starts paying real fees rather than token rewards. Third, whether the contagion risk seen in 2026 gets contained or repeats. For now the experiment is live, the penalties are real, and the market is still working out what shared security is worth.