Tracing the ghost in the liquidity protocol. The chain says solvency, the order book says panic. We are seven months into a bull run where Bitcoin has breached $70,000, ETF flows are hitting records, and yet the on-chain data for ZK-Rollup operators tells a story of quiet hemorrhage. I’ve been staring at the gas cost structures of zkSync Era and Scroll for the past three weeks, and the numbers are uncomfortable. Code is law, but narrative is leverage — and right now the narrative is masking a structural flaw that will surface when the next macro liquidity squeeze hits.
Context: The Bull Market Mirage
The bull market euphoria has papered over a crucial technical reality: ZK Rollups are expensive to run. In theory, they are the holy grail of scaling — validity proofs that compress thousands of transactions into a single batch, inheriting Ethereum’s security. In practice, the proving cost per batch for a modest throughput chain like zkSync Era sits at roughly $18,000–$25,000 in ETH gas fees today, according to my tracking of the batch submission data on Etherscan. At current ETH prices (~$3,800), that’s about 5–6 ETH per batch. With the network processing around 1.5 million transactions per day and submitting batches every 15 minutes, the daily proving cost exceeds $400,000. Compare that to the revenue generated from user fees (which are less than $50,000 per day for zkSync Era), and you see the gap. Volatility is the price of admission, but this gap is not volatility — it’s a design arbitrage that only works if transaction volume stays absurdly high or ETH gas stays low.
During the bear market, when ETH gas hovered around 5–10 gwei, the economics looked different. Proving costs were manageable. But today, with base fees often above 50 gwei and ETH price up 150% year-to-date, the cost per proof has quadrupled. Operators are bleeding. They are subsidizing users with token rewards and hoping that future volume growth will close the gap. But as a macro watcher, I see a classic liquidity trap: the very success of the bull market is making the infrastructure less sustainable.
Core: The Unacknowledged Subsidy Model
Let me walk through the math with Scroll, a relatively newer ZK-Rollup. Based on my audit-style breakdown of their batch submission data, Scroll submits a batch roughly every 3–5 minutes, containing about 1,000–2,000 transactions. Each batch requires a verification proof on Ethereum mainnet. The cost? Roughly 0.25 ETH per verification, plus the gas for calldata. At current gas prices, that’s about $950 per batch. With 12 batches per hour, that’s $11,400 per hour, or $273,600 per day. Their revenue from user fees? According to public data, Scroll collected about $30,000 in fees yesterday (mostly from swap and transfer fees). The subsidy rate is over 90%.
Now, proponents will argue that this is temporary — that batching efficiency improves, that EIP-4844 (proto-danksharding) will reduce calldata costs by 90%, and that volume will grow. Let me address each. First, batching efficiency: yes, you can pack more transactions per batch, but the proving cost scales sub-linearly; you still need to generate a validity proof for each batch, and the prover hardware is not free. Second, EIP-4844 is not live yet, and even when it arrives, it helps calldata but not the verification cost. The verification cost is a fixed per-batch cost that will not drop significantly unless SNARK algorithms improve by orders of magnitude. Third, volume growth: during a bull market, transaction count rises, but so does ETH gas price. The cost of proving is denominated in ETH, not USD. If ETH goes to $10,000, even a small gas spike becomes painful. The architecture of digital scarcity here is not the token supply of the L2, but the scarcity of Ethereum block space for proofs.
Contrarian: The Decoupling Thesis Myth
The prevailing narrative is that Bitcoin ETFs have decoupled crypto from traditional macro factors. I disagree. The liquidity cycle still rules. ETF inflows are a lagging indicator of global central bank liquidity. Real rates are still positive, and if the Fed is forced to cut due to a recession (which my models suggest is a 60% probability for H2 2025), risk assets will rally initially, but then the liquidity premium will vanish as credit spreads widen. In that scenario, ETH gas will drop (because retail activity wanes), and ZK-rollup economics will suddenly look viable again — but only because the underlying business model is dead. A solution that only works in a bear market is not a solution; it’s a counter-cyclical asset.
Decoding the signal from the hype: the real test for ZK-rollups is not whether they can process a million transactions per day at peak. It’s whether they can do so at a profit without token subsidies. Right now, no L2 — ZK or optimistic — is profitable on a standalone basis. Arbitrum and Optimism are sustained by massive token emissions that dilute holders. Those tokens are propped up by narrative, not cash flow. When the macro music stops, narrative is the first thing to evaporate.
Takeaway: Position for the Structural Fallout
My fund has been reducing exposure to L2 tokens that rely on subsidy models, and increasing positions in L1 blockchains with real fee revenue (Ethereum itself, Solana) and in infrastructure tokens that benefit from ZK-rollup failure (e.g., data availability layers like Celestia). Why? Because when the proving cost crisis hits, L2s will be forced to either centralize their sequencers (to reduce batch frequency) or raise fees (killing UX). Both outcomes are bearish for their tokens.
The market doesn’t price in what it thinks is temporary. But I’ve seen this movie before: DeFi summer’s liquidity traps, the NFT liquidity vacuum, the 2022 derivatives cascade. The ghost in the liquidity protocol is always the same — unacknowledged cost structures masked by bull market volume. Watch the on-chain proof costs, not the tweet threads. That’s where the architecture of digital scarcity will reveal its true nature.
