I remember the first time I heard the pitch. It was late 2021, at a conference in Lisbon. A founder stood on stage, slides glowing with diagrams of validity proofs and recursive snarks. "Zero-knowledge rollups are the endgame," he said. "Unlimited scale, Ethereum security, and near-zero fees forever."
The crowd cheered. I nodded along. But something in my gut—the part that has been burned by every narrative cycle since 2017—whispered: What's the catch?
That catch is now playing out in real time. Over the past six months, I've been tracking the P&L of major ZK rollup teams. The numbers are ugly. One project I analyzed is burning through $2.3 million per month on proving costs alone. Another has seen its operator margin go negative despite 40% growth in transaction volume. The narrative says ZK rollups are the future. The data says they are bleeding cash.
Trust is no longer a promise; it's a protocol. And the protocol of ZK proving is an expensive one.
Context: The ZK Rollup Promise vs. Reality
Zero-knowledge rollups were supposed to solve the trilemma. By moving computation off-chain and submitting a cryptographic proof (a validity proof) on-chain, they offer scalability without the trust assumptions of optimistic rollups. No 7-day withdrawal delay. No fraud proof games. Just math.
Projects like zkSync, StarkNet, Scroll, and Polygon zkEVM have collectively raised over $1.5 billion in venture funding. The pitch to developers: build on us, and you get Ethereum security with L2 speed. The pitch to users: fees under a cent, instant finality.
But there's a dirty secret. The cost of generating those validity proofs—especially for general-purpose zkEVMs—is astronomically high. Each block requires a proof that can take minutes to generate on high-end GPU clusters. The hardware cost, electricity, and maintenance scale linearly with transaction volume. In a bull market, when gas fees on Ethereum were high, the trade-off made sense. Users paid fees that covered the operator's costs. But in a bear market, with Ethereum gas below 10 gwei, the economics flip.
Code is law, but empathy is the interface. The operators are people. And people don't like losing money.
Core: The Real Cost of ZK Proving
Let me take you through the math. I've been auditing ZK rollup economics since 2023, and I've built a simple model based on public data from three leading projects.
Assumptions: - Average transaction count per L2 block: 200 - Proving time per block: 5 minutes using a single NVIDIA A100 GPU - Cost per GPU hour: $3.00 (cloud rental) - Proving cost per block: $0.25 (5 minutes = $0.25) - L1 gas cost to post proof and data: ~500,000 gas per block, at 5 gwei = $0.15 (assuming ETH at $2,500) - Total operator cost per block: $0.40
Now, revenue per block. The operator collects fees from users. In the current bear market, average fees on these L2s are around $0.001 per transaction (0.1 cent). With 200 transactions per block, that's $0.20 in revenue.
Loss per block: $0.20.
That might sound small. But scale it up. A typical ZK rollup produces a block every 15 minutes. That's 96 blocks per day. Daily loss: $19.20. Monthly loss: $576. Per year: $7,008.
But that's for a single GPU. Real operators use clusters of 10-20 GPUs to parallelize proof generation and reduce latency. Their hardware costs are higher. And they also pay for sequencer infrastructure, monitoring, and team salaries. I've seen estimates from one operator that their total monthly run rate is $150,000, while fee revenue is only $40,000.
The pivot wasn't a pivot; it was a return to subsidy. Most ZK rollups are being propped up by venture capital. They are burning through their treasury to subsidize user fees. The narrative says "cheap fees forever." The reality is that fees are cheap only because operators are willing to pay for them.
This is unsustainable. When the VC money runs dry, fees will have to rise. Or the operators will shut down.
Contrarian: Maybe the Narrative Is Wrong
Here's the contrarian angle that few want to discuss: ZK rollups may never be economically viable for general-purpose computation.
Think about it. The entire value proposition of ZK rollups is that they compress thousands of transactions into a single proof. But compression requires computation. And that computation is expensive. The more complex the smart contract logic, the more expensive the proof. An ERC-20 transfer is cheap. A Uniswap swap with multiple hops? Pricey. A full EVM execution with state reads and writes? Very pricey.
Optimistic rollups, by contrast, have a much simpler verification cost. They just post the data and assume it's correct. No proof generation. The cost is entirely in L1 gas for data availability. With Ethereum's EIP-4844 (proto-danksharding), data availability costs will drop further, making optimistic rollups even cheaper.
So why do we keep hearing that ZK rollups are the future? Because the narrative is driven by VCs who have invested billions in ZK infrastructure. They need the story to hold. They need liquidity to exit. They need you to believe that the proving cost problem is a temporary engineering challenge, not a fundamental economic one.
Trustless systems require trusting relationships. Right now, the relationship between ZK rollup operators and users is subsidized by trust in venture capital. That's not a trustless system. That's a rent-seeking one.
Takeaway: The Bear Market Exposes the Truth
Bear markets are brutal for a reason. They strip away the narratives and reveal the underlying economics. The ZK rollup space is currently in a state of denial. Teams are burning cash to maintain the illusion of cheap fees. They are raising more capital at lower valuations to keep the lights on. But the math doesn't lie.
I believe we will see one of two outcomes by 2027:
- ZK rollups pivot to specific use cases where the cost of proving is justified, such as high-value institutional settlements or privacy-preserving transfers. General-purpose zkEVMs for retail users will fade away.
- A new proving paradigm emerges—perhaps fully homomorphic encryption or some breakthrough in hardware acceleration—that reduces proving costs by 100x. But that's a bet on future technology, not a solution today.
Until then, I'm watching the L2 data closely. I'm tracking which projects are still profitable and which are burning through their treasuries. The next 12 months will separate the sustainable protocols from the VC-funded mirages.
I learned to stop preaching and start listening. I'm listening to the numbers. And the numbers are telling me that ZK rollups, as currently implemented, are a luxury we can't afford in a bear market.
As for the founders who are still out there, raising rounds and tweeting about the future of scale: I see you. I respect the hustle. But I also see the P&L. And I'm not buying the narrative.
Trust is no longer a promise; it's a protocol. The protocol of ZK rollups is currently running at a loss. That's not a protocol. That's a charity.
