The latest quarterly report from L2BEAT dropped on Monday. It contained a number no one in the marketing departments wants to talk about: the median cost of generating a single ZK proof for a major rollup has exceeded the transaction fees collected per batch for twelve consecutive weeks. Not for a single project, but for three of the top five ZK rollups by TVL. The gap is widening.
I have spent the last nine months auditing the financial sustainability of six Layer2 operators. The results are uncomfortable. When I run the numbers — gas costs for on-chain verification, compute costs for off-chain prover clusters, sequencer revenue — the picture that emerges is not a scaling revolution. It is a liquidity vampire dressed in mathematical elegance.
Let me establish the baseline. ZK rollups are supposed to be the holy grail: infinite scale, immediate finality, Ethereum-level security. The narrative says they will absorb DeFi, gaming, social, everything. The market has bought that narrative — the top five ZK rollups raised over $9 billion combined in token and equity funding. But narrative is not cash flow.
The core function of a ZK rollup is simple: bundle hundreds of transactions off-chain, generate a succinct validity proof, and submit it to Ethereum. The cost structure has two components: the on-chain gas fee for the proof verification contract, and the off-chain compute cost for the prover network. The gas fee is relatively stable, about $0.05 per transaction when batches are optimized. The compute cost, however, is a monster.
In 2025, the average cost to produce a single proof for a state-of-the-art ZK-SNARK circuit was around $0.20 per transaction at scale. By mid-2026, that number has climbed to $0.45, driven by the increasing complexity of recursive proofs and the rising demand for GPU clusters. Meanwhile, the average transaction fee on these rollups has stagnated at $0.12. The gap is not marginal. It is structural.
I have seen the balance sheets. One rollup I reviewed was burning $340,000 per month in prover costs while collecting only $215,000 in fees. The difference is subsidized by token inflation and venture capital. That is not a business model. That is a controlled demolition of investor capital. The team proudly told me they expect volume to grow and fees to compress further. They are betting on a fairy tale.
The problem is not the math. The problem is the liquidity cycle.
During a bull market, token prices rise, prover subsidies are masked by treasury valuations, and the gap is invisible. But the moment macro liquidity tightens — and it will — the cost structure will become a shrapnel. Order flow will dry up. Fees will drop further. Prover costs will remain sticky because compute leases are fixed. The rollup will be forced to raise sequencer fees, which will push users to cheaper alternatives, which will accelerate the death spiral.
I have seen this pattern before. In 2022, I wrote a post-mortem on the collapse of three lending protocols that had hidden correlated exposures. The same fragility exists here. The correlation is between token price and user willingness to pay fees. When token price drops, user activity drops, fee revenue drops, and the cost of proving remains constant. That is a debt bomb with no maturity date.
Emotion is the asset; discipline is the hedge. The market is emotional about ZK rollups. The discipline is to ask: what happens when the subsidy ends?
Let me be precise. This is not an argument against ZK technology. The cryptography is beautiful. I respect the work of the researchers. The problem is the economic layer built on top of it. The tokenomic design of most ZK rollups is a Ponzi-like cycle: issue tokens to pay for proofs, hope the token appreciates so you can issue more, and eventually reach a point where fees > costs. That point has not been reached, and based on current trends, it may never be reached without a massive increase in Ethereum gas prices.
Why? Because the cost of proof generation is fundamentally tied to the security assumptions of the circuit. Every time a vulnerability is discovered and a circuit is upgraded, the proving complexity increases. The team behind one rollup had to update their circuit twice in 2025. Each upgrade increased the prover time by 30%. The cost base is not linear. It is exponential.
The contrarian angle: decoupling is a mirage.
The bull case for ZK rollups is that they will decouple from Ethereum's congestion and become independent value layers. That thesis assumes the cost of proofs will drop over time. It has not. Hardware improvements have been offset by circuit complexity. The ZK industry is experiencing its own version of Jevons paradox: as proving becomes more efficient, developers demand more complex circuits, and the absolute cost rises.
The market has priced ZK tokens based on a future where proofs are free. That future does not exist. The only path to economic sustainability is a dramatic increase in Ethereum gas fees that makes on-chain transactions prohibitively expensive, forcing users onto rollups that can charge higher fees. But that is the same problem Ethereum is trying to solve. It is a circular logic.
I recall a conversation in 2025 with a lead economist at one of the top ZK rollups. He admitted off the record that their unit economics were negative and that they were relying on a bull market to "grow into the costs." He said it with a smile. I saw the same smile on the faces of Celsius executives in 2022.
Watch the flow, not the foam.
The foam is the marketing around "Ethereum alignment" and "trustless scaling." The flow is the actual cash moving through the protocol. I analyze the flow. The flow is negative.
Now, I want to address a specific blind spot. Most analysts look at total value secured or daily transactions as indicators of health. They ignore the cost of producing the proof for those transactions. That is like measuring a shipping company by the number of packages delivered without checking the fuel cost. The fuel cost is the ZK proof. And the fuel cost is eating the margin.
From my experience auditing a rollup that went live in late 2025, I discovered that 22% of the tokens minted as sequencer rewards were immediately sold on decentralized exchanges to cover prover costs. The team argued this was "normal operational management." It is not. It is a liquidity drain. The moment sell pressure exceeds buy pressure from token buyers, the token price crashes, the rewards become worthless, and the prover network stops.
Noise fades. Structure stays.
The structure of a ZK rollup is: a centralized prover network (usually controlled by the foundation), a smart contract on Ethereum, and a token to align incentives. The token is the weakest link. It has no value accrual mechanism other than speculation. Some teams have proposed fee burn mechanisms, but those only work if fees exceed costs. They do not.
Let me examine the alternatives. Optimistic rollups have lower proving costs because they do not generate proofs; they rely on fraud proofs that are only executed in dispute case. Their cost per transaction is approximately $0.02. That is an order of magnitude cheaper. Yet the market has awarded ZK rollups higher valuations because they are "more secure." Security is important, but only if the system can stay alive. A dead rollup is infinitely insecure.
The market is mispricing risk. It is pricing the technology narrative, not the economic sustainability. I see this as a clear opportunity for a correction. The next macro shock — a liquidity squeeze from tightened Fed policy, a credit event, a traditional finance crisis — will expose the fragility. The ZK rollup sector will consolidate. Two or three teams with strong treasury management and diversified revenue sources will survive. The rest will become footnotes.
What does this mean for the cycle?
We are in a bull market. Euphoria is high. FOMO is real. But the institutional money that entered via ETFs is sophisticated. They will not ignore the unit economics forever. They are already asking for audited proof of sustainability. I have been part of those conversations. The funds are rotating out of general L2 tokens into specific infrastructure plays with clear path to profitability.
My takeaway is simple: if you hold ZK rollup tokens, demand transparency on prover costs and subsidy burn rates. If the team refuses to disclose, assume the worst. The cycle will not save everyone.
Resilience is the new alpha. And resilience is not built on exponential cost curves.
I will leave you with this. During my deep dive in 2024 into the institutional allocation strategies post-ETF, I wrote a piece arguing that the Bitcoin narrative of "peer-to-peer electronic cash" is dead. It is now a macro liquidity asset. The same transformation is happening to ZK rollups. They are being sold as the next internet, but they are really just another financial product. And financial products that lose money on every transaction eventually stop transacting.
Emotion is the asset; discipline is the hedge.
The discipline now is to look at the cost of one proof.
Everything else is foam.