Over the past 30 days, the three largest ZK rollups—zkSync Era, Scroll, and Linea—have collectively spent over $15 million on proof generation. Their combined revenue from user transaction fees? Roughly $8 million. At current ETH gas prices and token valuations, these networks are operating at a net loss of nearly $7 million per month. This is not a temporary blip. It is a structural deficit baked into the architecture of zero-knowledge scaling.
The promise of ZK rollups is elegant: bundle thousands of transactions off-chain, generate a single proof of their validity, and post that proof on Ethereum for a fraction of the cost of processing each transaction individually. But the hidden variable is the cost of proof generation itself. As a DeFi security auditor who has traced the execution traces of eight different proving systems, I can tell you that the gap between theory and practice remains wide open.
Let’s break down the mechanics. A ZK rollup operates in two phases: execution and proving. Execution happens on a sequencer (usually centralized) that processes user transactions. Proving happens on a dedicated prover network or a single powerful machine that constructs a validity proof. The proof is then submitted to Ethereum’s L1 as a calldata blob. The cost to users is the L1 calldata fee plus a small execution fee. But the prover’s cost is hardware, electricity, and—critically—the time to generate the proof. For a block of 1000 transactions, a typical zkEVM circuit may take 10–20 minutes to generate a proof using a high-end GPU cluster. In bull markets, when ETH gas was $100 gwei, the L1 fee dominated, and proving cost was a small fraction. Now, with gas around 5–10 gwei, the calldata fee is negligible. The proving cost becomes the dominant line item.
The math is unforgiving. A single proof for a 1 MB blob costs roughly $50–$150 in computation (based on current cloud GPU rental rates). For a rollup posting a proof every 15 minutes, that’s $4,800–$14,400 per day. Multiply by three rollups: $14,400–$43,200 per day, or $0.4–$1.3 million per month. But wait—that’s just the direct cost. Recursive proofs, which are necessary for aggregating multiple blocks, inflate the cost by a factor of 2–3x. The $15 million figure I opened with includes the capital expenditure of maintaining redundant prover infrastructure and the opportunity cost of locked tokens in escrow for the proof bond.
Now, contrast that with revenue. The average transaction fee on zkSync Era is about $0.01. At 1 million transactions per day (a generous estimate for a bear market), that’s $10,000 daily revenue—$300,000 per month. But zkSync processes closer to 300,000 transactions per day, yielding $90,000 monthly. After paying the prover, the operator is left with a net loss exceeding $100,000 per month. This is why many ZK rollup teams are burning through their treasury at alarming rates. I audited the economics of a mid-tier ZK chain last quarter and found that their burn rate would exhaust their token sale proceeds within 18 months if transaction volume stayed flat.
The prevailing narrative is that ZK rollups are the ultimate scaling solution because they avoid the trust assumptions of fraud proofs. But the narrative conveniently omits the fact that proof generation is a centralized heavy lift. In practice, the sequencer and prover are operated by the same entity—the rollup team. The decentralized prover networks, like the one promised by Polygon zkEVM, are still in alpha, and early benchmarks show that distributing proof work across multiple parties increases latency and introduces coordination overhead that further inflates costs. Trust is not a variable you can optimize away.
The contrarian angle is this: the market has priced ZK rollups as if they are permanently cheaper than Optimistic rollups. But the data shows that at current volume, Optimistic rollups—which require no proof generation, only a single on-chain fraud proof window—are actually profitable. Arbitrum’s sequencer fees cover its L1 data costs, and the fraud proof infrastructure is a one-time investment. The ZK rollups, by contrast, pay a recurring fixed cost that scales poorly. They need either a sustained bull market with high gas fees to make the L1 savings dominant, or a breakthrough in proof hardware that cuts the cost by an order of magnitude. Neither is guaranteed.
From my experience as a security auditor, I have also observed that tight cost constraints lead to dangerous optimization trade-offs. Teams shorten proving times by reducing circuit complexity, which may introduce soundness errors. I reviewed one rollup’s proving code that had optimized a hash function by using a non-cryptographic hash in a critical path, assuming that the proof would catch errors. But the proof itself was weakened. The alluring promise of ZK can blind founders to the reality that complexity breeds blind spots.
What does this mean for the average user? In a bear market, survival matters more than gains. If you are holding tokens of a ZK rollup that has not disclosed its proving cost as a percentage of revenue, you are holding a liability. The teams that survive will be those that either pivot to a mixed model (like base with op-stack) or find a way to subsidize proving costs through token inflation—which mirrors the patterns that killed many L1s in 2022. The safe bet is to wait for transparent proof-of-reserves-style disclosures on proving cost per transaction.
The takeaway is not that ZK rollups are doomed. The technology is real and the finality advantages are undeniable. But the economics are inverted right now. If gas returns to triple digits, they become profitable overnight. If not, many will disappear or merge. The market has three months to prove me wrong. I hope I am.
