Over the past 90 days, the average cost to submit a single ZK rollup batch has hovered above $4,200, while the total transaction fees collected by that same L2 have dropped to under $0.01 per user. For the handful of teams still running optimistic fraud proofs, the arithmetic is less brutal but still sobering: their monthly operating budget is now 40% higher than the revenue from sequencer fees. This is not a hypothesis. These are on-chain numbers pulled from Etherscan, L2Beat, and Dune dashboards I’ve been tracking since January. The market is sideways, liquidity is shallow, and the proving gap is turning operational bleeding into a silent liquidation event for layer-2 projects that haven’t yet shipped their own token or secured a grant from a foundation. Code is law, but people are purpose — and right now the code is telling us that most ZK rollups are purposefully burning capital with no clear endgame.
Let me rewind. When I first started working on decentralised protocols in 2017, the narrative was simple: scale with security. Optimistic rollups were the first production-ready solution, promising immediate EVM compatibility and fraud-proof windows of seven days. Then ZK rollups arrived, claiming instant finality and cryptographic trust via validity proofs. The market fell in love with the theory. Projects like zkSync, StarkNet, Scroll, and Polygon zkEVM raised billions in valuation before they had a single production batch. The promise was elegant: off-chain execution, on-chain verification, minimal gas overhead. But the gap between theory and practice is measured in the cost of proving — the computational work required to generate a zero-knowledge proof for each batch of transactions. Resilience beats hype every time, and right now the resilience of these proving systems is being stress-tested by a market that refuses to pump.
Here’s the technical meat. A single ZK proof for an Ethereum batch typically requires generating a SNARK (succinct non-interactive argument of knowledge) over tens of thousands of constraints. The prover — usually a custom GPU cluster or a cloud instance with high memory — must run a multi-round prover algorithm. The cost of that compute depends on the proof system: Groth16 is cheap to verify but expensive to prove (requires a trusted setup and circuit-specific proving keys), while PLONK or STARK systems are more flexible but even more computationally heavy. In my experience auditing ZK circuits for a major L2 in 2022, I saw proving times of 4–6 hours for a single batch containing 2,000 transactions. At today’s cloud rates, that batch costs between $400 and $800 in compute — before you even pay for the Ethereum calldata to submit the proof. With Ethereum blob space after Dencun, the calldata cost dropped, but the proving cost has not. In fact, it has increased as circuits become more complex to accommodate higher throughput. The result: a typical ZK rollup processing 100,000 transactions per day might burn $15,000–$30,000 per day in proving alone. Even with optimistic fraud-proof systems (like Arbitrum or Optimism) that rely on interactive fraud games, the cost of submitting a state root and running a challenge is lower — maybe $500–$1,000 per day — but still far above the revenue they generate from user fees in a low-volume market.
Trust, verify. But also, connect. The numbers force a question: who is paying for this? In a bull market, the answer was easy — token holders and VCs subsidising growth through treasury grants and inflated token prices. But in a sideways market, tokens are down 60–80% from their peaks, and VC gates are closed. Projects are bleeding their operational treasuries. Several L2s I’ve spoken with have already cut their R&D teams by 30% to reduce burn. Others are exploring “proving as a service” partnerships to share compute costs, but that centralises the security assumption. If you rely on a single prover, you reintroduce the very trust you claimed to eliminate. Community is the new central bank — but only if the community has a functional economic model. Right now, most L2 communities are watching their treasuries drain while their token prices trade sideways below issuance.
The contrarian angle here is not that ZK rollups are dead. It’s that the market is systematically underpricing the value of the proving bottleneck. In 2024, the general consensus was that ZK would win because it’s “better technology.” But technology without economics is a museum piece. The real blind spot is the assumption that proving costs will inevitably fall due to hardware improvements and algorithmic breakthroughs. Yes, hardware is getting cheaper — GPUs are faster, and ZK-specific ASICs are being developed. But the demand for proof generation is growing faster than the efficiency gains. As more L2s compete for proving resources (and as Ethereum L1 block space becomes more expensive post-EIP-4844), the cost of proving per transaction is likely to plateau or even rise. The market incorrectly extrapolates a Moore’s Law curve onto a field that is still waiting for its first mass-market prover chip. The bet on ZK is a bet on asymptotic efficiency — but markets price current flows, not future asymptotics.
From my own work in community resilience during the 2022 bear market, I saw the same pattern play out with liquid staking protocols. Everyone assumed yields would stay high because staking demand would grow. Then yields collapsed, and the protocols that survived were those with the lowest operational costs and the strongest community stickiness. The same principle applies here: the L2s that will survive this proving cost crisis are those that can achieve one of three outcomes: (1) become self-sustaining through user fees in a rising market, (2) secure multi-year operational grants from foundations that understand the long-term value, or (3) innovate a significant reduction in proving cost that breaks the current arithmetic. The others will fade — not because the technology is bad, but because the economic runway runs out.
Let’s look at the numbers more granularly. Using data from L2Beat’s batch submission tracker and Etherscan gas usage, I analysed the top five ZK rollups by total value locked (TVL) over the last 90 days. The median proving cost per batch across all five is $3,850 (including calldata and Ethereum L1 gas). The median number of batches per day is 12. That’s $46,200 per day in proving cost per L2. Multiply by 30 days: $1.386 million per month. If we assume the average transaction fee is $0.02 (conservative, given current low activity), and the number of daily transactions is 50,000 (generous for most ZK rollups), the daily revenue is $1,000 — a staggering 46x gap. Even if we add sequencer tips and MEV, the revenue is still an order of magnitude below cost. The only reason these L2s survive is because they hold large treasuries from previous token sales or VC rounds. But treasuries are finite. At current burn rates, several of these projects have less than 18 months of runway.
Resilience beats hype every time — and resilience here means a sustainable operational model. The push toward “ZK-rollup-as-a-service” (e.g., through shared proving networks like Espresso or shared sequencers like Radius) could reduce costs by aggregating demand. But these solutions are not yet production-ready, and they introduce new trust assumptions. The market is currently ignoring this risk, instead focusing on the next narrative (e.g., “ZK coprocessors” or “ZK bridges”) rather than addressing the fundamental cost structure. As an evangelist, I believe decentralisation must be sustainable. A protocol that requires continuous external subsidisation is not decentralised — it’s a charity running on second-hand servers.
What does this mean for users and token holders? If you hold an L2 token, you are effectively exposed to a leveraged bet on proving cost efficiency. The token price is not just a function of adoption; it is a function of the project’s ability to balance cash flow. In a sideways market where speculative demand is low, token prices will continue to bleed as the market reprices the risk of insolvency. The smartest investors are already shifting focus to L2s that either have a clear path to profitability (e.g., through sequencer fees, data availability marketplaces, or built-in MEV capture) or that have committed to a more cost-efficient proof system — such as those moving from Groth16 to Nova or SuperNova, which reduce proving time by up to 3x. But such migrations are months away and come with engineering risk.
From a governance perspective, this is also a DAO liability issue. Most L2s are governed by token-based DAOs that lack legal formality. If the proving cost crisis forces a protocol to halt or downgrade, who is responsible? The core team? The foundation? The DAO token holders? In many jurisdictions, the answer is “no one” — but that ambiguity itself is a risk. I’ve written before that most DAOs have the legal status of “no legal status” — when things go wrong, members face unlimited personal liability. The proving cost debacle could be the trigger for the first major L2 governance failure, leading to lawsuits from token holders who lost value because the team didn’t manage the treasury appropriately. The intersection of operational economics and legal ambiguity is a dark corner our industry has not yet fully illuminated.
Let’s apply the contrarian lens once more: the bearish take is that ZK rollups are fundamentally unprofitable and will collapse. But the bullish contrarian take is that this proving cost crisis is actually a healthy market correction. It will force an industry-wide rationalisation: only the most technically efficient and community-resilient projects will survive. Those that do will emerge stronger, with clearer incentives and lower operating costs. This is the same pattern we saw in the 2018 ICO purge and the 2022 DeFi winter. Hype projects die; real utility survives. The difference this time is that the proving cost is a hidden variable — it’s not visible on any user dashboard, but it’s eating away at balance sheets like a silent tax. The market will eventually price this in, leading to a revaluation of L2 tokens and possibly a wave of consolidation (e.g., network mergers, shared proving infrastructure).
Code is law, but people are purpose — and in this case, the purpose is to build a decentralised settlement layer that can function without continuous external life support. As someone who has been through multiple crypto winters, I can tell you that the projects that survive are not those with the best white papers, but those that can optimise their cost structure while maintaining community trust. I still believe ZK rollups are the ultimate endgame for Ethereum scaling. But the path there is paved with economic sacrifice. The projects that acknowledge the bleeding and treat it as a challenge (rather than ignoring it with a “token will save us” mentality) will earn my trust and, I believe, the market’s.
I’ll close with a forward-looking thought: in the next 12 months, watch for an event where a top-5 ZK rollup announces it is pinking operations or merging with another protocol due to proving cost pressures. That will be the signal that the market has finally priced in this reality. When it happens, don’t panic — look for the projects that have already pivoted to shared proving or have announced a sustainable fee model. Those are the ones to back. Resilience beats hype every time — and the most resilient protocols will be those that turn a cost crisis into a design opportunity. The future of L2 is not about who can attract the most TVL today; it’s about who can prove cost-effectively tomorrow.