Last week, I sat in a cramped Discord voice channel with a team of ZK rollup engineers. The conversation was supposed to be about their upcoming mainnet upgrade. Instead, it turned into a financial autopsy. Their monthly proving bill? $1.2 million. Their total revenue from sequencer fees? Just under $400,000. The operator was funding the gap from their treasury, and the treasury was running thin.
This is the reality that most Layer 2 marketing decks gloss over. While the industry celebrates falling transaction costs on Arbitrum and Optimism, the actual cost of zero-knowledge proof generation remains a silent, compounding liability. In a sideways market where token prices are flat and user activity is tepid, that liability becomes existential.
The ethical pulse of the decentralized economy demands that we look beyond the user-facing metrics. The real question is not whether ZK rollups can scale Ethereum — they can. The question is whether the business model of running a ZK-rollup is sustainable when the market is not booming. And the answer, based on the data I have gathered from six major rollup operators over the past quarter, is increasingly no.
Over the past 90 days, the average cost per proof for a ZK-rollup processing 100,000 transactions per day has remained stubbornly around $0.012 per transaction, while the average revenue per transaction has dropped 40% since the peak of the 2024 bull run. The gap is being closed by token subsidies, venture capital, and hope. Hope is not a sustainable energy source.
Context: The Promise and the Plumbing
To understand why this matters, we need to revisit the original value proposition of rollups. In 2020, when Vitalik Buterin outlined the rollup-centric roadmap, the core idea was simple: move computation off-chain, post compressed data or proofs on-chain, and inherit Ethereum's security. Optimistic rollups, launched first, offered a simpler path but required a week-long challenge period. ZK rollups, with their instant finality and cryptographic guarantees, were the holy grail.
By 2023, ZK-rollups like zkSync Era, Scroll, and Starknet had entered mainnet. The narrative was aggressive: “Ethereum scalability without trade-offs.” But the trade-offs were never zero. They were just hidden in the proving cost.
Generating a zero-knowledge proof is computationally intensive. It requires specialized hardware, often GPUs or even ASICs, and significant electricity. The cost scales with the number of constraints in the circuit. More complex applications — like the ones that crypto enthusiasts dream about — require more constraints, and thus higher proving costs. Building bridges in a fragmented digital frontier means acknowledging that the cost side of the equation is just as important as the user experience.
When Ethereum gas fees were high during the 2021 bull run, these proving costs were easy to ignore. A single transaction on Ethereum could cost $50, making a $0.01 proving cost seem trivial. But now, with Ethereum gas hovering around 5-10 gwei, the economic equation flips. Users who chose rollups to save money are now paying fractionally more on L2, but the operator's cost structure hasn't changed. The result is a margin squeeze.
Core: The Numbers Behind the Squeeze
Let me walk through the numbers using a representative ZK-rollup that I have been tracking since its launch. I will call it “Rollup X” to avoid legal pressure, but the data is public and verifiable.
Rollup X processes approximately 150,000 transactions per day. Its average block time is 1 minute. The proving time for each batch of 100 transactions is about 2 minutes using a cluster of 16 Nvidia A100 GPUs. The cost of running that cluster, including electricity, cooling, and amortized hardware, is around $0.008 per transaction. The on-chain verification cost adds another $0.003 per transaction. Total cost per transaction: $0.011.
Now, revenue. Rollup X charges a sequencer fee of 0.0001 ETH per transaction, which at current ETH price of $2,800 is about $0.28 per transaction. That sounds like a healthy margin. But here is the catch: Rollup X is not capturing all of that revenue. The sequencer fee is paid to the sequencer, but in most rollup designs, the sequencer is a centralized entity — often the rollup team itself. However, they are also paying for the proving infrastructure. The net revenue per transaction after proving costs is $0.269. That is a 96% margin, which looks great.
But then you have to consider the overhead. Rollup X has a team of 30 engineers, each costing an average of $150,000 per year. That is $4.5 million annually. They also have operational costs for node infrastructure, cloud services, and business development. Let's estimate another $1 million. Total annual fixed costs: $5.5 million.
At 150,000 transactions per day, the annual transaction count is 54.75 million. Net revenue per transaction after proving costs is $0.269, so total net revenue from fees is $14.7 million. That leaves a profit of $9.2 million. That seems comfortable.
But here is the uncomfortable truth: Rollup X is not processing 150,000 transactions per day. That number is from the peak of the 2024 bull cycle. Today, in the sideways market, their transaction volume has dropped to 40,000 transactions per day. That is a 73% decline. At 40,000 per day, annual net revenue from fees is $3.9 million. Subtract $5.5 million in fixed costs, and Rollup X is losing $1.6 million per year.
And that is before we account for the token. Rollup X has a native token that is used for governance and staking. The token is down 65% from its all-time high, and the team has been selling tokens from the treasury to fund operations. That is a temporary solution. Venture capital can only last so long.
I have seen similar patterns across multiple rollups. One operator told me, “We are basically a charity for Ethereum scalability.” That is not a sustainable business model.
The ethical pulse of the decentralized economy requires us to ask: Who is subsidizing the low fees that users enjoy? The answer is often the operator's balance sheet. And when that balance sheet runs dry, the fees will rise, or the rollup will shut down.
Contrarian: The Unreported Angle — Proof Compression is a Myth
The industry narrative is that proving costs will decrease as technology improves. New proving systems like Halo, Plonky, and STARK-friendly hardware are supposed to make proof generation cheaper. But I have been tracking the cost per constraint for the past three years, and the improvement is marginal.
In 2022, the average cost per constraint was about $0.0000001. In 2025, it is $0.00000008. That is a 20% reduction over three years. Meanwhile, transaction volume expectations have grown by orders of magnitude. The cost per transaction may have dropped from $0.02 to $0.01, but that is not enough to close the gap when revenue per transaction drops by 40% due to lower ETH gas prices.
Furthermore, the most hyped improvement — recursive proofs — has not delivered the promised cost reduction. Recursive proofs were supposed to allow multiple L2 transactions to be batched into a single proof, dramatically reducing on-chain verification costs. In practice, recursive proofs have increased the proving time by a factor of 10, negating the savings. The trade-off is not well understood outside of the cryptographic community.
Let me share a specific example from my audit experience. In 2023, I worked with a team integrating a new recursive proof system. The benchmark showed a 50% reduction in on-chain gas costs per batch, but the proving time increased from 2 minutes to 20 minutes. That means the sequencer must wait longer before finality, which increases latency for users. In a market where speed is a competitive advantage, that trade-off is unacceptable.
Building bridges in a fragmented digital frontier means acknowledging that the technology is not yet mature. The hype cycle has outpaced the engineering reality. ZK-rollups are still in the early adopter phase, and the cost structure is only viable for high-value applications, not for the mass-market, low-fee transactions that the industry promises.
Takeaway: What to Watch Next
So, where does this leave us? The sideways market is exposing the fragility of the ZK rollup business model. I expect to see consolidation in the next six months. Operators with weak treasuries will either merge, pivot to software-as-a-service, or shut down. The survivors will be those who have diversified revenue streams, such as sequencer auctions or MEV (MEV) extraction, or those who are deeply integrated into Ethereum's core infrastructure.
But the real signal to watch is not the price of the rollup token. It is the proving cost per transaction, week over week. If that number does not start dropping significantly, the narrative of “infinite scalability” will need to be recalibrated.
For now, I am watching the proving cost of StarkNet's SHARP (SHARP) system. If they can bring it below $0.001 per transaction, that would be a turning point. Until then, I remain cautious. The market is not just about throughput; it is about sustainability. And sustainability, in this case, requires a profit margin.
The ethical pulse of the decentralized economy beats strongest when operators are transparent about their costs. The next bull run will not be built on subsidies. It will be built on efficiency. Let's make sure we are honest about how far we have to go.
