I sat with a DeFi founder in Tel Aviv last Thursday who pitched what he called the "fourth-generation restaking" thesis. His framing was sharper than the way most VCs are currently describing the category, and it crystallised something I'd been circling for a few months. Restaking's first act, the one that produced the $20B EigenLayer TVL headline in mid-2024, is functionally over. The second act has different mechanics, different winners, and a different question being asked of every protocol in the stack. I want to write down what that question is, because I think the largest funds in this space are pricing the wrong end of it.
What The First Act Actually Was
The first act of restaking was a TVL race. EigenLayer launched, the points programme worked exactly as designed, and capital piled in faster than any DeFi primitive had seen since the original DeFi summer. Symbiotic followed. Karak followed. Babylon launched the Bitcoin-side version. By the time the music stopped, the four-protocol race had attracted somewhere north of $35B in restaked capital across the major venues.
The pitch that pulled the capital was elegant. You stake ETH (or stETH, or BTC) and earn the base staking yield. You also "restake" the same underlying — pledging it as economic security for a separate set of services, the so-called Actively Validated Services or AVSes, and earn an additional yield on top. The same dollar of capital earned two yields. The market priced that double-yield aggressively. The protocol that aggregated the most restaked capital won the first act.
That's the part that's over. The TVL game has been played. The points have been distributed. The protocols are large and the moats are roughly settled. What hasn't been settled, and what is going to define the next 18 months in this category, is whether the AVSes actually generate the yield the model implied.
The Yield-Attribution Problem
Here's where I think most of the analysis goes wrong. Total restaked TVL is a number that's easy to measure. Total fees paid by AVSes to restakers is a number that almost nobody is measuring credibly. The gap between them is the second-act problem.
In early 2026, the dollar-yield being paid by genuine, customer-paying AVSes to restakers — separated out from points programmes, founder-team-funded subsidies, and within-ecosystem rebates, is a fraction of what the original mental model implied. I've now seen unit-economics work from three different AVS founders that confirm a version of this. The AVS revenue line isn't yet large enough to support the restaked-yield expectations that the points era set.
That gap will close one of two ways. Either AVSes generate enough real revenue to fund competitive yields, or restakers reprice the security they're providing and a meaningful chunk of TVL exits the system. Both are happening simultaneously. The protocols that survive the second act are the ones that move quickly to the first outcome, credible AVS revenue, rather than the ones that keep playing the TVL race.
What Real AVS Revenue Looks Like
The AVSes I'm most bullish on in 2026 are the ones with a non-restaking-internal customer. That sounds obvious. It is not how the first generation of AVSes was designed.
The first wave of AVSes were largely Web3-internal services — bridge security, oracle networks, sequencer decentralisation for L2s, DA-layer security for app-chain stacks. All of those are real and important. None of them generated direct customer revenue. The yield they paid restakers came either from token emissions of the AVS itself or from infrastructure subsidies from the L2s and ecosystems that benefited.
The second wave looks different. The AVSes I'd actually invest in now are ones with credible external customers paying in actual cashflow. The most interesting category, to my read: cross-chain settlement and bridging, projects like LayerZero's security work, the Wormhole guardian set, the new generation of intent-settlement protocols. These have real, paying customer transactions flowing through them, and they're starting to route some of that fee revenue back to economic-security providers rather than just to the protocol token.
Disclosure here: Kima Network, where I serve as Chief Marketing Officer, is in the cross-chain settlement category, and the second-act restaking framing maps onto how Kima thinks about its own validator economics. I'm conflicted on the category. I'm also closer to the unit economics than I'd be otherwise, and the numbers I see there reinforce the framing rather than undermine it.
The other category I'd watch in the second act is RWA-adjacent settlement and oracle work. As tokenized real-world assets scale into the trillions over the next decade (I wrote about the bottlenecks a few weeks ago), the security and oracle infrastructure underneath them becomes one of the largest customer-pays demand surfaces in DeFi. Restaked economic security plugged into RWA-adjacent oracles and settlement venues, paid for by issuers and counterparties, not by the protocol itself, is the cleanest version of the second-act bet.
What's Different About How I'm Underwriting
The frame I now use when a restaking-adjacent founder pitches us at PRIM3 in 2026 is the AVS-revenue-attribution question. Before tokenomics, before TVL projections, before the security model: what is the customer-paid revenue line of the AVSes you're either operating, securing, or competing with, on a 12-month forward basis with no subsidies and no points? If the founder can answer crisply, the rest of the pitch tends to compose. If the founder can't, the TVL projection isn't doing the work it pretends to.
Two years ago I'd have weighted the security-model design more than the customer-revenue line. I was wrong about that. The security model is necessary. It is not sufficient. The second act decides itself on whether the AVSes plugged into the security become real revenue businesses, not on whether the security design itself is technically elegant.
The Honest Caveats
What this thesis doesn't explain: liquid restaking tokens (LRTs) like ether.fi, Renzo, Kelp, Puffer. Those are partly an AVS-revenue bet and partly a points-aggregation bet and partly a structured-product bet, and they each have their own unit economics. The biggest LRTs will probably do fine even if the broader AVS-revenue story takes longer than the market is pricing, they have other revenue lines. They're the safer corner of the second act.
It also doesn't explain Bitcoin-side restaking (Babylon and the second wave coming behind it). The customer set there is partly the same, bridges, settlement, oracles, and partly a separate category specific to BTCFi's emerging needs (BTC-backed lending, BTC-native L2s, BTC-LRT structuring). I'd treat that as a parallel second act, with its own AVS-revenue question.
And it doesn't explain why some pure-EigenLayer-shaped TVL plays will still compound from here — the moat from being the largest restaking venue is real, and the second-act AVS revenue, if it ever does scale to match the original expectations, will disproportionately benefit the largest venues first.
The Forbes-Quotable Line
If I had to put a single sentence to it — and this is the line I keep ending up at on calls with allocators: the first act of restaking was about how much capital you could attract; the second act is about how much revenue you can attribute back to that capital, and almost nobody is measuring that number honestly yet.
The next Bridge Notes will turn to a piece I've been chewing on for months — why the Middle East cohort is reshaping the founder map for everyone in this space. If you're building in the AVS-revenue layer and you've got real customer-paid numbers you'd be willing to share, push the conversation: LinkedIn or pitch us at prim3.vc.