The bubble isn't the disappearance of on-chain lending. The bubble is the story selling it as an extinction event. On-chain lending didn't die in 2022. It was never alive โ not as a sustainable business, anyway. It was a liquidity subsidy wearing a collateralized-loan costume.
Let me frame the scale before the narrative gets ahead of the data. At the peak of the 2021 cycle, DeFi lending protocols held around fifty billion dollars in total value locked, depending on which dashboard you trust. By the 2023 trough, that number had collapsed by roughly seventy percent. Aave and Compound, the two surviving giants, learned to scrape by on a fraction of their former volume. Fantom and Avalanche, chains that had built entire economic identities around lending flywheels, watched their TVL evaporate into single digits.
And yet Ethereum never stopped settling tens of billions of dollars per day. Coinbase still routes institutional flows through the same L1 rails. The L2s kept shipping blocks. The industry, in its usual narcissistic way, decided the whole episode was a "DeFi winter" and moved on to points, restaking, and AI agents.
That's the story selling it. The reality is uglier. The disappearance of the lending market didn't reveal an L1 industry that lost one leg. It revealed an L1 industry that was never standing on more than one leg โ and that leg was subsidized. Friction reveals the fault lines no one else sees. The friction was the liquidation cascade of 2022. The fault line was the architecture itself.
From my seat as an exchange market lead, I watched institutional liquidity providers pull risk from one chain and rotate into another based on the depth of the borrowing book. That is not a technical metric. That's a bank-run signal. If a chain's entire economy depends on how much money is available to borrow at a given moment, it stops being infrastructure and starts being a leveraged balance sheet with a consensus layer bolted on top.
So when I see articles asking "What remains for the L1s after the lending market disappears?" I don't read it as a rhetorical question. I read it as a governance-first audit of what those L1s were actually selling. The answer is uncomfortable.
Context: The Lending Market Was Not a Market. It Was a Mechanism.
Let's reconstruct what on-chain lending actually was before we talk about what it left behind.
The standard protocol design is elegantly simple. You deposit a crypto asset as collateral. You borrow a different asset against that collateral. You pay a variable interest rate determined by the utilization of the pool. If the ratio of collateral to debt falls below a threshold, a liquidator is allowed to seize your collateral and close your position. No credit score. No income verification. No legal entity. Just math, oracles, and the threat of liquidation.
That model was the first genuinely crypto-native financial primitive. DEX trading had existed for years, but it was a closed loop between two currencies. Lending was different: it let users own a position in an asset while simultaneously borrowing against it. It created leverage. It created yield. It created a reason for capital to stay inside a single chain's ecosystem instead of being transferred out after a trade.
The technical prerequisites are what matter for this story. On-chain lending needs a reliable price oracle, a liquidation auction mechanism, and a pool of assets deep enough to absorb slippage during stress. All three of those prerequisites depend on L1 infrastructure. The oracle has to publish price updates as transactions. The liquidation has to execute in the same block as the price drop. The pool has to be composable enough that other protocols can integrate it without permission.
That is why lending became the benchmark application for every L1. It wasn't just an app. It was a stress test for the entire consensus and execution stack. If a chain could run a liquidation war without falling over, it had proved its technical worth. If it couldn't, it was dead on arrival. In 2021, every chain wanted to run that test and pass it, because passing meant the capital would flow.
The problem is that the capital did not flow because of superior technology. It flowed because of yield subsidies. This is where my own history with the industry gets uncomfortable.
In 2020, when the first DeFi summer was consuming the industry, I was a junior researcher obsessed with governance failures. I spent six weeks dissecting the bZx exploit, not because I cared about the stolen dollars, but because the governance token distribution behind it was an even bigger disaster. What I learned from that period is that most DeFi protocols were not designed to attract users. They were designed to attract capital. The two are not the same, and the distinction gets hidden when capital is cheap and attention is expensive.
A bull market is a terrible time to audit incentives, because everything appears to work. Deposits generate yield. Yield attracts more deposits. More deposits push the native token higher, and the native token becomes the most convenient collateral. So users borrow the native token, sell it for a stablecoin, deposit the stablecoin back into the lending pool, and repeat. The chain sees transaction volume, fee revenue, and TVL. The lending protocol sees utilization rates that justify the yield. The native token sees a buy wall that is supported by nothing but the loop itself.
I called this yield scaffolding. It looked like an ecosystem under construction. In reality, the scaffolding was the only structure. I remember running a liquidation simulation for one of these forks and realizing the protocol could not survive a simultaneous drop in the native token and the collateral asset. The design assumed the token would only go up. That is not a lending protocol. That is a Ponzinomics simulator.
Then 2022 happened. Terra collapsed. Celsius froze withdrawals. Three Arrows Capital defaulted. BlockFi filed for bankruptcy. The price of every liquid collateral asset dropped at the same time, because the leverage was not distributed across independent strategies. It was concentrated in the same loop, repeated at different scales. Liquidations triggered more liquidations. Withdrawals triggered more withdrawals. The oracle prices were not wrong, but the order book was empty on the way down.
The lending market did not literally vanish. Lending still exists on Ethereum, on Solana, and on a few other chains. But the kind of lending that had driven L1 growth โ the reflexive, token-collateralized, leverage-yield-loan loop โ collapsed in a matter of weeks. What remained was not a market. It was a mechanism that had lost its fuel source.
The activity that replaced lending did not replace its economics. Perpetual futures took over as the primary fee generator on many chains, but perps are a different beast. They are a zero-sum trading venue, not a capital formation primitive. A perp exchange can generate more fees than a lending protocol, but it does not create the same network effect, because users do not leave deposits behind. They leave position sizes that disappear when the trade closes.
That correction was never truly processed. The industry moved on. But the structural question remained: if a chain's value was driven by lending, and lending's value was driven by subsidized yield, what was the chain actually worth when the subsidy ended?
Core: Anatomy of an L1 After the Flywheel Stops
This is the part that needs clear technical decomposition, because the casual answer โ "L1s are still useful for transfers" โ is lazy and hides a deeper problem.
An L1's value can be broken down into five layers. The first is settlement and finality: the guarantee that once a transaction is included, it will not be reversed. The second is data availability: the guarantee that the data needed to reconstruct the chain is publicly accessible. The third is execution: the ability to run smart contracts that alter state. The fourth is composability: the ability for different applications to build on the same state without permission. The fifth is monetary premium: the token's role as the native unit of settlement, collateral, and gas.
Lending touched all five layers. Settlement, because each borrow and liquidation produces a transaction with economic consequences. Data availability, because every oracle update has to be stored somewhere. Execution, because liquidation must happen within a narrow time window. Composability, because lending pools are the connective tissue between other protocols. And the monetary premium, because the native token was the most convenient collateral on the chain.
Remove lending and the five layers start to thin. Settlement still handles transfers, but transfers are low-frequency events compared to borrowing and liquidations. Data availability still stores blocks, but there is less meaningful data to store. Execution still runs contracts, but the most demanding contract โ the liquidation โ is gone. Composability still works, but with fewer interconnected apps, the network effect weakens. The monetary premium becomes a gas token and a governance token, and neither of those guarantees enough demand to keep a validator set secure.
There is a specific metric that captures this better than TVL: utilization. Utilization is borrowed supplies divided by total supplies. A healthy lending protocol runs at high utilization, because high utilization means capital is being put to work. But high utilization is also what makes a protocol vulnerable to bank runs. When a large borrower gets liquidated, other borrowers rush to repay, utilization crashes, and yield disappears. Capital then leaves, and the cycle feeds on itself.
I saw that dynamic in real time during the 2022 unwind. When Aave's USDC lending pool utilization hit one hundred percent, the interface showed a modest borrow APY, but the actual cost of borrowing during a supply squeeze was ten times higher. That is not a bug. That is the mechanism. The protocol was designed to bid for liquidity under stress. But it also meant that the chain's fee revenue was hostage to supply shocks in a stablecoin market that had nothing to do with the chain's technology.
For an L1, the equivalent metric is base-fee revenue per active address. This is the amount of actual economic value users are willing to pay to have their transactions included. In 2021, lending drove that number up because every liquidation was a race to be first in the block. In 2023, after lending had collapsed, that number dropped across most L1s. It was still propped up by airdrop farmers, but the organic demand for block space was a fraction of what the TVL charts suggested.
This is why I keep saying that the bull market is a hiding mechanism. When token prices are rising, nobody wants to subtract the points-farming demand from the fee chart. But the subtraction is exactly what matters. If a chain's base-fee revenue after removing airdrop farmers is negative, the token is not a currency; it's a fundraising vehicle.
Let me show you what this looks like in fee data. Between 2022 and 2023, I ran a rough correlation between lending protocol TVL and L1 fee revenue across several chains. I want to be clear that this wasn't a formal academic study. It was an operational tool I used to decide which chain liquidity to allocate to. But the results were striking. For Ethereum, the correlation was present but not dominant, because Ethereum's fee market was increasingly driven by L2 settlement, NFT activity, and the early stages of restaking. For Avalanche, the correlation was brutal. For Fantom, it was closer to one-to-one. A chain whose fee revenue moves in lockstep with a single lending app is not a blockchain. It's a structured credit product with a consensus layer on a marketing budget.
The bull market since 2023 has not solved this problem. It has only given it a new uniform. The new "growth" on L1s and L2s is dominated by points programs, airdrop farming, and restaking. Let's be honest about what points are. A user deposits an asset into a protocol, receives a score based on the size and duration of that deposit, and hopes the protocol eventually issues a token that can be sold. There is no organic yield. There is no underlying loan. There is no real economic activity. The point is a promise, and the promise is exactly as good as the airdrop's future value. That is not lending. But it is structurally identical to the yield-scaffolding loop that collapsed in 2022: capital is being paid to park itself in a useless position, with the payment denominated in a future token.
There is a name for an economy where the only organic demand is the expectation of a future token: pre-rent. It is not lending. It is not fee generation. It is a pre-revenue company selling equity in a future airdrop, with the chain as the venue. That is not a sustainable base for L1 valuation.
In my audit work, I've seen this pattern repeat. Every time a new L1 or L2 launches a points campaign, the same chart appears. TVL spikes. Fee revenue spikes. Token price spikes. Then the incentive ends, and the metric that matters โ base-fee revenue from users who are not farming a token โ remains flat. The market doesn't move on what the chart looks like during the campaign. The market moves on what the chart looks like after the campaign ends.
And there is a second technical pressure coming that will make this worse. Post-Dencun, rollups started publishing compressed data to blob spaces instead of storing everything in the L1 calldata. That was a beautiful upgrade: the cost of rollup data dropped by ninety percent or more, and L2 fees collapsed. But the supply of blob space is not infinite, and the demand for it is growing far faster than the supply schedule. We are already seeing blob prices spike during periods of heavy L2 usage. If that trend continues, and I believe it will, the cost of settlement for every rollup in the ecosystem will double again within two years. L2 fees will stop being invisible. They will become a factor that every user has to think about, and at that point the L1s that survived the lending collapse will have to answer an awkward question: "Why are you charging me this much when the actual loan that used to justify this fee no longer exists?"
The core insight is not that lending was the only useful thing an L1 can do. The core insight is that lending was the only thing that created a self-reinforcing demand for block space with a large enough surface area to justify a native token's market cap. If you remove lending, you have to find another demand source of equal intensity. Payment settlement alone is not enough, because people don't transact billions of times a day for groceries on-chain. Gaming and social are not enough yet, because they are still chasing attention rather than generating fees. AI-agent payments might be the next candidate, but the infrastructure for those is still embryonic.
The next experiment is AI-agent economies. I have spent the last year looking at decentralized compute networks and zero-knowledge proofs as verification layers. The honest view: it is too early to know whether agents will transact on public rails or on private infrastructure. But the market is already pricing some L1s as if the agents will come. That is the same mistake as pricing L1s as if the lenders would stay.
That leaves the L1s in a strange, uncomfortable place. They are not dead. They are not going to zero. But the valuation architecture that was built on the assumption that every L1 would grow its own Compound fork and its own yield loop is gone. The token price might still recover in the next bull market. The fundamental demand for block space will not, unless the chain finds a reason to exist that is not a variation on borrowing against a native token.
Contrarian: The Real Threat Isn't Disappearance. It's Export.
Now for the part that everyone in the industry refuses to say out loud. The conventional framing is that on-chain lending was killed by the bear market, by regulation, or by the collapse of bad actors. All of those are true, but they are surface-level. The more dangerous story is that lending didn't die. It was exported. And it was exported precisely because public L1s were never the right place for it to live in the first place.
Think about where on-chain credit actually works today. The largest and most liquid credit markets in crypto are not on Aave or Compound. They are on centralized exchanges, where margin lending happens against institutional collateral without any public block explorer watching. They are in prime brokerage desks that net positions across venues and settle on private infrastructure. They are in over-the-counter desks that offer leverage to counterparties they have KYC'd and credit-scored. These entities are not interested in on-chain lending for the borrower experience. They are interested in the audit trail. They want a public record of the final settlement, perhaps for compliance, perhaps for proof of solvency, but not because they need the chain's smart contract to mediate the loan.
This is the uncomfortable truth about RWA that I keep coming back to. On-chain real-world asset lending has been a three-year storytelling exercise. We've tokenized treasuries. We've tokenized money-market funds. We've tokenized private credit funds. Every announcement is greeted with the same breathless enthusiasm, and every announcement misses the same point: traditional institutions do not need your public chain to originate a loan. They need your public chain as a notary. They need a timestamp and a cryptographic signature that says the asset existed, that the transfer happened, and that the audit trail is clean. That is a settlement service, not an economy. And it is a service with a price that trends toward zero, because the same notarization can be done on a private chain or a centralized database with much lower friction.
I have a simple test I use when evaluating RWA integrations. I ask the team: if the underlying asset were cash, would you still keep the tokens on-chain? The ones who pause and say "for compliance, yes" are telling you the truth. The ones who answer with a marketing slide are not. That pause is the difference between a tokenized treasury and a notarized database.
I have seen this pattern from the inside. When my exchange was discussing tokenized treasury products, the institutional counterparties were always enthusiastic about the technology and completely indifferent to the chain. What they wanted was a clean legal wrapper, a reliable custodian, and the ability to issue a report at the end of the quarter. The public chain was a nice-to-have. It was a decoration. That should terrify anyone who believes that RWA is going to "save" the L1s from the loss of lending. If the lender does not need your chain, the chain has no leverage.
The same logic applies to Bitcoin. I have watched the BRC-20 and Runes debates with a mixture of fascination and exhaustion. The idea that you can take a deliberately restrictive, security-first base layer and turn it into a playground for meme-coin inscriptions is backwards. BRC-20 and Runes on Bitcoin are like using a Rolls-Royce to haul cargo. It insults the car and doesn't carry much. Bitcoin's scarcity is in block space and settlement security, not in the ability to mint dog-themed tokens. If the industry wants to preserve Bitcoin's role, it should treat it as the final settlement layer, not as a new venue for leveraged token markets.
What this means for the L1s is more complicated than "lending left so we lose fees." The real risk is that public L1s are being pushed into a corner where their only defensible role is as a settlement and finality layer. That role is valuable but it is also brutally competitive and thin-margin. Settlement is a commodity. Finality is a commodity. The chain that wins is the one that can provide those commodities at the lowest cost with the highest guarantee of non-corruption. That is Ethereum's current position, but it is not guaranteed forever.
I also think about the ETF approval cycle of 2024 in this context. When the Bitcoin ETFs launched, I spent months mapping the custody flows between Coinbase wallets and traditional brokerage accounts. The lesson was not that institutions had finally embraced blockchains. The lesson was that they had adopted Bitcoin as a position, not as an economy. They wanted exposure to the asset without touching the infrastructure. That is the same instinct that will drive institutional lending off-chain: take the yield, avoid the chain. The market doesn't reward cathedrals; it rewards toll booths. And so far, the most successful toll booth in this industry has been the first one that convinced everyone it was boring enough to trust.
The contrarian conclusion is this: the disappearance of the lending market is not the signal you should be watching. The signal is whether new credit instruments are settling on public chains under terms that are visible and inspectable, or whether they are settling inside closed infrastructure where the public chain is reduced to a notary. If the former, the L1 has a future as a financial settlement layer. If the latter, the L1 becomes a decorative blockchain, and the market will eventually price it accordingly.
Let me stabilize the panic before we close. The data does not say that every L1 is worthless. It says that every L1 whose value model depended on lending is now unanchored. The chains with diversified fee streams are fine. The chains with one lending app and a token buyback are not. The difference is measurable, and it will become visible as soon as the next airdrop cycle ends.
Takeaway: The Toll Booth Test
So what do you do with this analysis? Stop watching TVL. TVL is a vanity metric that counts other people's subsidies and gives no information about the sustainability of demand. Watch base-fee revenue. Watch the number of distinct addresses that are paying for settlement because they need it, not because they are farming a token. Watch whether lending returns in a new form โ undercollateralized, agent-verified, institutional โ and whether the settlement of those loans happens on a public L1 or inside an exchange's private ledger.
The question was never "What does an L1 have left after lending disappears?" The question is "What did it have before the subsidy arrived?" If the answer is nothing, then the disappearance is not the disaster. The exposure is. The bubble isn't the story. The story is the story selling it. Friction reveals the fault lines no one else sees. The market doesn't reward cathedrals; it rewards toll booths. Time to find out which chains are toll booths, and which are just ruins wearing blockchain costumes.
