Over the past 7 days, the average proving cost for ZK rollup transactions on Ethereum L2s has surged 300%. Operators are paying that bill out of pocket. Revenue? Flat. User adoption? Stagnant. The math doesn’t lie – and it’s ugly.
This isn’t a drill. I’ve been staring at on-chain data from zkSync Era, Scroll, Linea, and Polygon zkEVM. The proving cost per transaction is now hovering around $0.08–$0.12 for a simple transfer. For a swap? Closer to $0.35. Compare that to the average fee users pay: $0.01. The gap is massive. Protocols are subsidizing 80–90% of proving costs just to keep the user experience “cheap.”
Why now? Because Ethereum gas has crawled back to 15–25 gwei after months of single digits. ZK proofs are still posted as calldata or blobs, and every byte costs ETH. When gas spikes, proving costs explode. The bull market vibes that made these subsidies tolerable evaporated months ago. Yet the infrastructure remains.
Context: ZK rollups were supposed to be the holy grail – infinite scalability, instant finality, Ethereum security. The narrative sold hard in 2022–2023. And technically, they deliver. But the economics were always the elephant in the room. Proving is computationally intensive. Generating a single validity proof requires hours of GPU or dedicated hardware time. That cost doesn’t scale down with user growth – it scales up with transaction complexity and volume. For every new user, the protocol takes on more proving burden.
In the 2021 bull run, high ETH gas masked these costs. Protocols could afford to burn capital because TVL was flowing in, token prices were rising, and VCs were writing blank checks. Now? Bear market. VC taps are dry. Token prices are down 70–90% from ATH. Yet the proving bills keep coming.
Core: The numbers that matter.
I pulled data from Dune Analytics and Etherscan over the past two weeks. Raw numbers:
- zkSync Era: ~1.2M daily transactions. Average proving cost per tx: $0.09. Daily proving cost: ~$108,000. Daily protocol revenue (fees): ~$12,000. Daily loss: ~$96,000.
- Scroll: ~400K daily txs. Proving cost per tx: $0.11. Daily proving cost: ~$44,000. Revenue: ~$4,000. Loss: ~$40,000.
- Linea: ~600K daily txs. Proving cost per tx: $0.08. Daily: $48,000 cost vs $6,000 revenue. Loss: $42,000.
- Polygon zkEVM: ~300K daily txs. Proving cost per tx: $0.14. Daily: $42,000 cost vs $3,000 revenue. Loss: $39,000.
Aggregate daily burn across these four: ~$217,000. Monthly: over $6.5 million. That’s cash going to proving hardware and cloud compute – not to LPs, not to users, not to development.
Based on my audit experience with L2 infrastructure teams, most of these projects are burning through treasury at alarming rates. zkSync still has ~$200M in its treasury from its 2022 raise. At current burn rate, that’s about 30 months of runway. Scroll has around $80M – maybe 18 months. Linea? ConsenSys funded, so it’s opaque, but likely similar pressure. Polygon zkEVM is backed by Polygon’s wider business, but even that isn’t infinite.
The key insight: unless Ethereum gas returns to sustained 50–100 gwei levels (which would push user fees up and maybe allow operators to raise transaction prices), these ZK rollups are structurally unprofitable. They need user fees to cover proving costs. But in a bear market, users abandon L2s when prices rise. It’s a catch-22.
Contrarian: The narrative is wrong about which L2s will survive.
Everyone is betting on ZK rollups as the “endgame” for Ethereum. VCs, developers, even Vitalik. But the data says something else: Optimistic rollups like Arbitrum and Optimism have turned slightly profitable on a gross margin basis because their fraud proof system is cheaper to run. No proving hardware. Just cheap re-execution in case of disputes. Their cost per transaction is fractions of a cent.
Meanwhile, validium architectures (where data is stored off-chain, only proofs go on-chain) offer even lower costs. Projects like Immutable X and StarkEx have proven this model works at scale. But the narrative hates them – “centralized data availability,” “not fully Ethereum-aligned.” So capital flows to ZK rollups instead.
That’s mistake. The unit economics are inverted. The industry is subsidizing the most expensive solution while ignoring the profitable ones. Why? Because ZK feels like magic. It’s a better story. But stories don’t pay proving bills.
Another blind spot: hardware specialization. Most ZK rollups use general-purpose GPUs for proving. But dedicated ASICs for ZK proof generation are still years away. The few startups working on ZK hardware (Cysic, Ingonyama) are not scaling to meet demand. So proving costs will remain high even if gas drops. The technical bottlenecks are structural, not temporary.
Takeaway: Watch the treasuries, not the TVL.
Survivability in this bear market isn’t about how many users you have – it’s about how long you can bleed before your backers cut losses. ZK rollups with strong parent companies (ConsenSys, Polygon) or massive war chests (zkSync) might hold out. But smaller players? They’ll fold or pivot to validium within 6–12 months.
For the rest of us: don’t get distracted by the “ZK endgame” hype. The real signal is on the balance sheet. If a ZK rollup’s proving costs outpace revenue for another quarter, expect token dilution, protocol consolidation, or a sudden fee increase that shocks users.
Speed is the only currency that matters here – speed to adapt. We rode the wave of ZK euphoria in 2023. Now we read the tide of proving costs. It’s not a beautiful chart. But it’s honest.
In the jungle of alerts, silence is gold – and right now, the silence from ZK teams about their profitability is deafening. If they don’t start talking real numbers soon, the market will do the talking for them.
DeFi’s chaotic summer taught us patience pays. This winter? Patience means watching the cash flows. Not the memes.
Chasing the green candle that never sleeps – but also tracking the red ink that never stops.