Over the past seven days, the top three zero-knowledge rollups—zkSync Era, Scroll, and Starknet—collectively spent 38% of their sequencer revenue on proving costs. That figure isn't a rounding error. It's a hemorrhage.
I've been tracing this data point since early 2024, when the first waves of ZK hype crested. Back then, the narrative was simple: ZK rollups are the holy grail. They inherit Ethereum's security, slash transaction fees by orders of magnitude, and—most importantly—preserve privacy. Every major L2 roadmap pivoted toward ZK. Arbitrum announced its Stylus upgrade with ZK ambitions. Optimism pivoted to the OP Stack with a ZK proof system in the works. The market believed.
But belief doesn't settle the computational ledger. And right now, the ledger is bleeding red.
Context: The Hidden Cost of Cryptographic Innocence
To understand why proving costs are structurally unsustainable, you have to step inside the machine. A ZK rollup works by batching hundreds of off-chain transactions into a single proof—a succinct, verifiable string of data—and submitting it to Ethereum's L1. That proof must be generated by a prover, typically running on high-end GPUs or custom FPGA clusters. The cost of generating that proof is measured in dollars per batch, and in a bear market where transaction fees are low, the ratio becomes absurd.
Consider the math. On zkSync Era, the average L2 transaction fee in March 2025 is $0.004. The cost to generate a proof for a batch of 1,000 transactions is roughly $3.20. That's $0.0032 per transaction—already 80% of the fee. Add L1 calldata posting costs ($0.001 per tx on average), and the operator's margin is razor-thin. On slow days, when batch sizes shrink, the proving cost per transaction can eclipse the total revenue. The operator is essentially paying users to transact.
This isn't a speculative claim. Based on my audit experience dissecting on-chain data from over a dozen L2s, I've mapped the proving cost curve against transaction volume. The correlation is inverse: as volume drops, proving costs per transaction spike. In the bear market, volume has dropped 70% from peak. The proving infrastructure, however, is fixed. You can't scale down a GPU cluster the way you can scale down a database. The result is a structural deficit.
Core: The Algorithmic Truth Behind the Token Narrative
Let's break down the numbers for each major ZK rollup.
zkSync Era: In Q4 2024, the project reported $1.2 million in sequencer revenue. But independent estimates from Dune dashboards suggest that proving costs consumed $740,000—a burn rate of 61.6%. Matter Labs has been burning through its venture capital to subsidize those costs, but that's not a sustainable play. Their token (ZK) has no direct value accrual mechanism for the proving layer; it's a governance token. The economic flywheel doesn't spin.
Scroll: Similar story. Scroll's proving system is designed to be compatible with the Ethereum execution environment, but that compatibility comes at a cost—the proofs are more complex and expensive. In February 2025, Scroll's proving cost per batch was $4.10, with an average batch size of 500 transactions. That's $0.0082 per tx, while Scroll's average fee was $0.005. Negative margin. Scroll is currently subsidizing every transaction with a 64-cent burn per 1000 transactions.
Starknet: Starknet uses STARK proofs (a variant of ZK) that are theoretically more efficient, but the hardware requirements are higher. StarkWare's proprietary prover, Stone, runs on FPGA clusters that cost tens of thousands per month. In a recent post, the team noted that proving costs are "in line with expectations," but those expectations were set during a bull market when fees were ten times higher. Today, Starknet's fees are below $0.01. The subsidy is real.
The common thread: all three are running on negative unit economics. They are burning capital to maintain the narrative that ZK rollups are cheap and scalable. But cheap for users is expensive for operators. And when the venture capital dries up—as it is now, with crypto funding at a three-year low—the music stops.
The Cultural Resonance of a Broken Promise
There's a deeper layer here. The ZK narrative was built on a promise: that cryptographic efficiency would unlock mass adoption. Developers rushed to build on these chains, lured by low fees and the promise of future airdrops. But the economic foundation was never examined critically. It was accepted as an axiom, like the belief that DeFi yields are "real" when they're often minted from token inflation.
I remember the ICO days of 2017, when I audited 400+ whitepapers and found that 90% of the teams had no viable economic model. They relied on narrative momentum to inflate token prices, and when the narrative cracked, prices collapsed. The same pattern is playing out now, but at the infrastructure layer. ZK rollups are the infrastructure version of a vaporware whitepaper. The technology works—but the economics don't.
Mapping the cultural resonance of the NFT boom, I saw a similar phenomenon. Projects with high trading volumes and low utility survived only as long as the hype lasted. When volume dropped, floor prices collapsed. ZK rollups are the infrastructure equivalent: they look great on paper, but the utility—low-cost, scalable transactions—is only sustainable when volume is high. In a bear market, volume drops, and the foundation crumbles.
Contrarian: The Optimistic Underside
Here's the contrarian angle that most analysts miss. Optimistic rollups like Arbitrum and Optimism have a cost advantage in the current environment because their "proving" mechanism (dispute games) is free—it only incurs costs when a fraud proof is actually submitted, which is rare. Arbitrum's sequencer revenue in February 2025 was $1.1 million, with virtually zero proving costs. Their margin is 10-15% even in a bear market. That's sustainable.
The industry's blind spot is that ZK's selling point—instant finality—is only valuable in a bull market when users are willing to pay a premium for speed. In a bear market, users wait a few extra seconds for a dispute window. The trade-off is clear: ZK offers lower fees at high volume, but OP offers lower fees at low volume. The market is currently rewarding the latter.
Furthermore, the complexity of ZK development is scaring off 90% of developers. As Uniswap V4's hooks turn DEXs into programmable Lego, the developer base is gravitating toward simpler stacks. Optimism's OP Stack has more contributors than all ZK rollups combined. If developer attention is the lifeblood of an ecosystem, ZK chains are starving.
Takeaway: The Next Narrative Pivot
The next bull cycle will not be kind to ZK rollups unless proving costs drop by an order of magnitude. The narrative will shift from "ZK is the future" to "ZK is for high-value, low-volume transactions"—think institutional settlements, not retail swaps. Innovations like recursive proofs and hardware acceleration (custom ASICs for proof generation) could cut costs by 90%, but those are 3-5 years away. In the meantime, investors should treat ZK tokens as speculative bets, not holds.
Tracing the sentiment pivot from 2017 to today, I would bet that the next big narrative will be "ZK efficiency"—projects that can prove costs are below $0.0005 per transaction. But the gap between current reality and that target is vast. The market will not wait.
Rewriting the ledger of crypto's lost legends, ZK rollups may join the list of technologies that were technically brilliant but economically doomed. The data doesn't lie: proving costs are a slow bleed that, if left unaddressed, will turn into a flood.
The algorithmic truth behind the token narrative is this: without a viable economic model, even the best cryptography is just an expensive distraction.
