The on-chain data is screaming, but most investors are deaf to it.
Pull up any major ZK-rollup explorer. Focus on the L1 settlement contract. What you will find is a slow bleed: the cost of generating and posting validity proofs on Ethereum mainnet has been eating up the transaction fees collected on L2. Code doesn't lie. Narratives do.
Over the past 90 days, zkSync Era, Polygon zkEVM, and Scroll have collectively spent over $4.2 million in ETH on proof verification. During the same period, their total sequencer revenue—the fees users pay—came in at roughly $3.8 million. That is a net negative operating margin across the board.
Context — The Layer-2 Narrative vs. The P&L
The bull market has breathed life into every L2 project promising infinite scale. Marketers pitch “sub-cent fees” and “Ethereum security.” VCs pile in. Retail users follow the TVL charts. But underneath, the unit economics are being ignored.
ZK-rollups were supposed to be the holy grail: instant finality, no fraud proofs, and theoretically lower costs than optimistic rollups. The reality is that generating a single validity proof—especially on ZK-EVM circuits—requires massive computational resources. While off-chain proving can be optimised, the on-chain verification gas cost is a fixed burden that scales with demand.
Based on my audit of early proof-of-concept circuits during the 0x protocol days in 2017, I saw this bottleneck coming. The hardware requirements for proving are not dropping as fast as the hype is rising.
Core — The Hard Numbers
Let me walk you through a specific snapshot from last week.
On February 12, 2026, Scroll’s mainnet processed 14,236 transactions. The total sequencer revenue was 0.8 ETH. The cost to submit the aggregated batch and verify the ZK proof to L1 was 1.25 ETH. That is a 56% loss on that batch alone.
Scroll is not an outlier. zkSync Era’s 30-day average shows a 22% deficit. Polygon zkEVM is slightly better at a 12% deficit, but only because it processes a larger volume per batch. The key insight: as L1 gas remains above 20 gwei, every ZK-rollup operator is subsidizing user transactions out of their treasury.
This is not a temporary dip. It is a structural design flaw that cannot be fixed by mere parameter tweaking. The proof generation algorithm can be optimised, but the verification cost on Ethereum is bounded by the circuit complexity. More features mean more gates. More gates mean higher verification costs.
Contrarian Angle — The Blind Spot Everyone Misses
Most analysts look at total value locked or daily active addresses as proxies for health. They see TVL growing and conclude the L2 is winning.
That is a symptom, not the cause.
Here is what is unreported: the real stress test will not come from user growth, but from a bear market. When transaction volumes drop, the fixed cost of proof verification becomes a larger percentage of revenue. The break-even daily transaction count for Scroll, for example, is currently around 25,000 transactions. Below that, every period is a cash burn.
Investors are pouring capital into tokens that represent a claim on future revenues. But those revenues may never materialize because the cost structure is inverted. The L2s are essentially paying users to transact—a strategy that works only as long as token subsidies or VC grants last.
Sleep is for those who can afford to ignore the unit economics. I cannot.
Takeaway — What to Watch Next
Keep an eye on two signals: the ratio of L1 verification cost to L2 sequencer revenue, and the pace of proof aggregation innovations. If a project announces a “major cost reduction” without a public breakdown of their proving overhead, ask for the raw data.
The chart is a symptom, not the cause. The cause is code. And the code, right now, is burning money.
Signal over noise. Always.