Ethereum L2s Hit a 2-Year Low. ZK Rollups Are Empty.

Total value locked across Ethereum's layer-2 networks has slid back to roughly $5 billion on L2Beat's measure, a level last seen in 2023. That erases most of what accumulated during 2024, when Optimism, Arbitrum and ZKsync were the entire conversation.

The headline number is bad. The distribution underneath it is worse, and it is not what "the L2 sector is shrinking" implies. We pulled DefiLlama's chain-level TVL on the morning of July 29 to see how the remaining money is actually spread, and the answer is that the sector has not shrunk evenly. It has collapsed into one chain.

RelatedLayer 2 rollups: how blockchains finally scaled

Two numbers, two definitions

Before reading anything into the figures, it matters which one you are looking at, and most coverage this week has quietly mixed them.

L2Beat measures value bridged into a rollup: canonically bridged assets, externally bridged assets and natively minted tokens. That is the roughly $5 billion, two-year-low number. DefiLlama measures value sitting inside DeFi protocols deployed on the chain. That is our $6.58 billion. Neither is wrong. Bridged value counts a stablecoin sitting idle in a wallet; DeFi TVL does not, but does count assets that arrived by other paths. Our figures below are DefiLlama's unless stated otherwise.

Where did the concentration come from?

Base holds $4.55 billion of the $6.58 billion we tracked, which is 69 percent on its own. Add Arbitrum at $1.20 billion and OP Mainnet at $305 million and the top three reach 92 percent. Everything else, and that is roughly eighteen more chains, splits the remaining half-billion.

The chains at the bottom are the story. ZKsync Era, which raised and launched as one of the two flagship zero-knowledge rollups, holds $14 million. Scroll holds $12 million. Polygon zkEVM and Zora round to zero at this precision. These are not early-stage networks that never got traction. They shipped, they had incentive programs, they had capital, and the capital left.

Why did the ZK rollups empty out?

The obvious answer is that ZK proving was expensive and optimistic rollups were cheap, so users went where fees were low. That was true in 2023 and it stopped being the explanation some time ago, because proving costs fell substantially and the money did not come back.

The better explanation is that after EIP-4844 made data availability cheap for everyone, cost stopped being a differentiator at all. When every rollup is fractions of a cent per transaction, "we are cheaper" is not a reason to bridge. What is left as a reason is distribution, and that is a completely different competition. Base has Coinbase's user funnel attached to it. Arbitrum has the incumbent DeFi deployments and the liquidity that comes with them. A technically excellent rollup with neither of those things is competing on a dimension the market stopped pricing.

Liquidity also compounds in one direction. Traders go where slippage is lowest, which is where liquidity already is, which attracts more liquidity. Once a chain falls below the depth needed for serious size, the remaining users are the ones who cannot easily leave. Fourteen million dollars spread across a general-purpose chain's protocols is below that line.

What does this mean for Ethereum itself?

Less than the L2 numbers suggest, and this is where the doom framing overreaches. Ethereum L1 holds $41.3 billion in DeFi TVL, more than six times the entire L2 sector combined. The rollup-centric roadmap was pitched as the way Ethereum scales, and on activity terms it broadly worked. On value-retention terms, serious capital largely stayed on mainnet.

RelatedLitecoin's LitVM Brings ZK Smart Contracts to LTC

The honest read is that L2s won the transaction count and lost the balance sheet. That is a real outcome, just not the one the 2021 roadmap described, and it has consequences for anyone whose token accrues value from L2 sequencer revenue. Cheap transactions on a chain with $14 million in it do not generate meaningful fees.

What it means for the market

The signal for anyone holding rollup governance tokens is that TVL and token value have decoupled from the technology entirely. A ZK rollup's proving system can be genuinely superior and its chain can still be empty, because the thing being competed for is distribution. When evaluating an L2 token now, the questions worth asking are who sends it users, what fee revenue the sequencer actually books, and whether any of that revenue reaches the token. For most of the list above, the third answer is currently no.

This is analysis, not investment advice, and the numbers move daily. But the structural point holds: consolidation of this shape rarely reverses without an external distribution shock, and there is no visible candidate for one.

Our take

Calling this the death of layer twos gets it wrong in an important way. Base is a layer two, and Base is doing fine. What died is the assumption that a rollup is a business simply because it is a rollup.

Between 2021 and 2024 the sector built dozens of general-purpose chains on the theory that block space was the scarce good and whoever produced it cheaply would capture value. EIP-4844 made block space abundant, and the scarce good turned out to be users. The teams that had a user pipeline attached to something else, an exchange, an existing DeFi ecosystem, kept their liquidity. The teams that had only excellent cryptography are now running very fast, very cheap, very empty chains. That is not a failure of engineering. It is a failure of the thesis the engineering was funded against.

Original analysis by GenZTech. Per-chain figures are our own pull of the DefiLlama chains endpoint on 29 July 2026 and will drift; the two-year-low framing follows The Block.