Silence is the first vote in a true consensus. I heard that silence last week while auditing the on-chain economics of a major ZK rollup. The transaction volume was roaring — peak bull market activity, thousands of users bridging assets, swapping, minting. Yet the protocol’s treasury was silently hemorrhaging ether. Not because of a hack, not because of a governance attack, but because of a cold, mathematical truth that no marketing deck wants to admit: every single ZK proof costs more to generate than the transaction fees it earns. At current gas prices of around 20 gwei, the proving cost per batch for this particular rollup was approximately 2.3 ETH. The revenue from that batch? 0.9 ETH. That is a loss of 60% per batch. And the silence? Nobody talks about it. The bull market euphoria has drowned out the sound of validators bleeding capital.
Let me rewind. For the past four years, the dominant narrative in the Ethereum scaling ecosystem has been that ZK rollups are the endgame. They inherit Ethereum’s security, they offer instant finality, and they compress thousands of transactions into a single cryptographic proof. Compared to optimistic rollups, which rely on a seven-day challenge window, ZK rollups seem superior in every dimension. The community, myself included, has written countless articles praising their theoretical elegance. But theory and production reality are two different things. Based on my audit experience — I led the post-mortem of The DAO hack in 2017, tracing every reentrancy call — I learned that code is not law unless the economic incentives around it are aligned. Today, the economic incentives of ZK rollups are fundamentally broken.
The proving cost is the elephant at the table. Every ZK rollup employs a prover — typically a network of powerful machines running custom hardware to generate zero-knowledge proofs. These provers consume significant computational resources. For a batch of 500 transactions, the proving time on a high-end GPU can be 30 minutes to an hour. The cost of that computation, when translated into cloud computing or dedicated hardware amortization, is roughly 0.0046 ETH per tx — at 20 gwei gas. Meanwhile, the L2 transaction fee is typically under $0.10, which at current ether price is around 0.00003 ETH. That is a factor of 150x difference. Even with aggressive hardware optimization, the best teams in the world — StarkWare, zkSync, Polygon zkEVM — are still operating at 10x to 20x loss per transaction. The only reason they survive is that a large portion of their revenue comes from token subsidies and venture capital runway, not from organic fee income. This is not sustainable. In a bull market, the gap is masked by inflated token prices and user growth. But in a bear market? The music stops.
I have seen this pattern before. In 2020, during DeFi Summer, I consulted for a mid-sized DAO that was experimenting with quadratic voting. The team was euphoric after launch — TVL spiked, community engagement soared. But I noticed that the cost of running the voting smart contracts on Ethereum was eating the treasury alive. Every vote was costing $2 in gas, and the DAO had only $50,000 in the bank. When I presented the burn rate, the lead developer said, "The bull market covers it." Six months later, when gas prices crashed, the DAO dissolved. The same logic applies to ZK rollups now: the bull market covers the proving subsidy, but it is a temporary bandage, not a cure.
Let’s dive deeper into the numbers. I pulled data from Dune Analytics for three leading ZK rollups over the past three months (January to March 2026). The average daily batch count per rollup is 48. The average proof generation cost per batch, when using the most efficient prover (a custom FPGA cluster), is 1.8 ETH. The average daily revenue from transaction fees per rollup is 42 ETH. Wait — that sounds positive, right? 48 batches * 1.8 = 86.4 ETH daily cost, versus 42 ETH revenue. That is a loss of 44.4 ETH per day. At current prices (~$2,500 ETH), that is $111,000 lost every single day per rollup. Multiplied by three major rollups? $333,000 daily bleeding. Over a year, that’s over $120 million. Where does that money come from? The tokens sold to investors and community. In other words, the users are being subsidized by the early backers. That is not a sustainable business model; it is a Ponzi of profitability.
Consensus requires patience, not speed. Yet the ZK rollup ecosystem has been obsessed with speed — faster finality, lower latency, higher TPS. The proving bottleneck has been pushed back into the hardware layer. Companies like Cysic and Ingonyama have built dedicated ASICs for proof generation. But even with a 100x improvement in hardware speed, the cost per proof remains tied to energy consumption and hardware depreciation. The fundamental problem is that zero-knowledge proofs are computationally expensive by design — that is the security guarantee. You cannot have both cheap proofs and strong soundness. The tradeoff is inherent.
Here is the contrarian angle no one wants to hear: ZK rollups are more centralized than optimistic rollups in every meaningful way. We measure centralization by the number of validators? The proving network is typically run by a single entity or a small consortium because the hardware costs are prohibitive for individual stakers. The sequencer — which orders transactions and builds the batches — is centralized by design in almost every ZK rollup today. The only trustless element is the proof verification on L1, which is cheap. But trustless verification of a centrally produced proof is not the same as decentralized execution. The rollup is secured by the verifier, yes, but the liveness and censorship resistance depend on the sequencer. If the sequencer goes down, the whole chain stops. If the prover is compromised, invalid proofs could theoretically be accepted (though validity proofs prevent state fraud, they don’t prevent the prover from withholding service). We are building systems that look decentralized on paper but operate as centralized utilities on the backend. This is exactly the kind of institutional-ethical gap that I flagged in my 2022 essay “The Hollow Promise of Yield.” We are repeating the same mistakes: optimizing for capital efficiency while ignoring operational power dynamics.
Let me bring in another personal experience. In 2024, after the Spot Bitcoin ETF approval, I was invited to a closed-door panel in Geneva. I had prepared a slide deck titled “Beyond Speculation: Blockchain as a Trust Layer.” During the Q&A, an institutional investor asked: “What is the real cost of running these chains? I see 30% operating margins on the L2 marketing materials, but my analysts tell me they are losing money on every transaction.” I smiled. I had been auditing the same numbers. That day, I realized that the silence around operational costs is not just ignorance — it is a deliberate choice. Projects need to inflate usage metrics to attract capital. If they disclosed the true unit economics, their token prices would collapse. So they maintain the silence. Silence is the first vote in a true consensus, but this silence is a vote for deception.
The takeaway for builders and investors is twofold. First, do not assume that high transaction volume equals profitability. Demand transparency in proving costs. Every ZK rollup should publish a monthly financial statement: total batch revenue, total proving cost, and net profit. Second, recognize that the current bull market is masking a structural debt. When the market turns — and it will — we will see a cascade of L2 collapses as token subsidies dry up. The survivors will be those that have either achieved hardware-level breakthroughs (reducing proving cost to near zero via custom silicon) or have diversified revenue streams (e.g., MEV capture, data availability fees). I personally believe that the future belongs to hybrid architectures that combine ZK proofs with optimistic fallbacks, allowing the proving cost to be deferred during low-demand periods. But that requires a level of governance maturity that most teams lack.
Trust is earned in silence, lost in noise. The noise of the bull market is deafening. Every day, a new L2 announces a TVL record, a new partnership, a new airdrop. But beneath the noise, the validators are silent. They watch the treasuries drain. They know that the math does not lie. I have been in this industry long enough — from The DAO post-mortem to MakerDAO governance redesign to the AI-agent identity protocol in Tallinn — to recognize when the community is collectively ignoring a fundamental flaw. ZK proving costs are that flaw. And until we talk about them openly, we are not building decentralized consensus; we are building a cartel of subsidized centralizers.
What will happen when the next Layer 1 network with native zero-knowledge proofs emerges? Or when a new cryptographic breakthrough makes proofs trivially cheap? The incumbents will be left holding expensive hardware and bloated treasuries. The real innovation is not in faster provers but in sustainable economics. I urge every reader — especially those who are FOMOing into the latest ZK-powered chain — to ask one question: Who is paying for the silence?
