Block 18765432 on Ethereum mainnet. A single transaction from the zkSync Era operator wallet paid 14.2 ETH in L1 gas fees—roughly $45,000 at current prices. That was just to submit one batch proof. Over the past 30 days, the total L1 proving cost for zkSync Era exceeded $3.2 million. Meanwhile, daily L2 revenue from transaction fees? Approximately $180,000.
The math doesn't lie. In a sideways market where L2 volumes and fee revenue are depressed, ZK rollup operators are bleeding cash every day. The narrative tells you ZK is the future of scaling. The on-chain data tells you the present is unsustainable.
Let me be clear: I've run validator nodes and audited rollup contracts since the early days of Loopring. I watched StarkEx handle high throughput during the 2021 NFT mania. But the proving cost problem for general-purpose ZK rollups like zkSync Era is not solved—it's just not being discussed outside of coder channels.
This article is not a hit piece. It is a wake-up call for any LP, holder, or trader who thinks ZK rollup tokens are a safe bet in this bearish consolidation phase.
Context: Why Proving Costs Matter Now
zkSync Era launched its mainnet in March 2023 to massive hype. TVL peaked at over $800 million. Users flocked for airdrop expectations and low transaction fees. But the architecture is fundamentally different from optimistic rollups like Arbitrum or Optimism.
In an optimistic rollup, the sequencer simply posts transaction data to L1. Fraud proofs are a distant, rarely used mechanism. In a ZK rollup, every batch must be accompanied by a valid zero-knowledge proof—verified on Ethereum's L1. That proof is computationally heavy to generate and expensive to submit.
The cost structure is: - L1 data cost: posting the compressed transaction data (calldata) — this exists in both types. - L1 proof verification cost: a fixed gas cost for the EVM to verify the zk-SNARK — this is unique to ZK rollups.
Using Dune Analytics data from @rollup_costs_viz, I checked the past 90 days. zkSync Era's average proof submission gas cost is about 800,000 gas per batch. With Ethereum basefee averaging 25 gwei, that's 0.8 million 25 1e-9 = 0.02 ETH per batch. But wait—there's also the calldata cost for the batch itself. Add it up, and each batch costs around 0.6 ETH. With roughly 10 batches per hour, we get 144 ETH per day. At $3,200 per ETH, that's $460,000 L1 cost daily.
Now cross-check with L2 revenue. zkSync Era currently processes about 1.5 million transactions per day. Average fee per transaction? 0.00005 ETH? No—that would be $0.16 per transaction. But most transactions are simple transfers or swaps. The actual average fee is around 0.0001 ETH ($0.32). Multiply: 1.5 million * 0.0001 = 150 ETH per day. At $3,200, that's $480,000 revenue. Gross profit? $20,000 a day. That's 4% margin. On a good day.
But that's before sequencer costs, infrastructure, and team salaries. The reality is that zkSync Era is operating at a loss—subsidized by VC money and token sale proceeds.
Core: The Blowfish in the Room
I'm not pulling these numbers from a blog post. I'm pulling them from on-chain contracts I've traced since my early days building scrapers in 2017. Here's the real kicker—and this is the insight you won't find in a press release:
Proving costs are not linear with transaction volume. They are step-function.
Why? Because a ZK proof generator must accumulate enough transactions to make batch submission worthwhile. If volume dips below a threshold, the batch size shrinks, but the fixed L1 verification cost doesn't. That's why you see periods where zkSync posts a batch of only 50 transactions—still spending 0.5 ETH on L1 gas. That's $10 per transaction just to settle on L1.
Volatility is just fear wearing a disguise. In this case, the fear is that L2 revenue cannot sustain proving costs even with current ETH gas prices. Now imagine a bull market where ETH gas spikes to 100 gwei. Proving costs quadruple. Revenue from L2 fees might double at best. The math becomes catastrophic.
The market is pricing zkSync Era as if it's a high-growth tech stock. The on-chain reality is a capital-intensive utility with negative operating margins.
Optimistic rollups don't have this problem. Arbitrum's L1 cost is almost entirely calldata, which scales linearly with volume. Their daily L1 spend averages $80,000—about one-sixth of zkSync's. Their revenue? With higher activity and full MEV extraction, Arbitrum generates $1.2 million in daily L2 fees. Arbitrum is cash-flow positive. zkSync is burning cash.
Contrarian: The Narrative Gap
The popular take is: "ZK rollups are superior because they have fast finality and don't need a 7-day challenge period. Therefore, they will dominate."
That's true for user experience. It's false for business model.
The mint button was a lever, not a purchase. Investors minted the "ZK is the future" narrative based on technical superiority, ignoring the economic sustainability. Meanwhile, users are choosing Arbitrum for its deep liquidity and mature ecosystem. zkSync Era's TVL has dropped from $800 million to $350 million in six months. That's not just market rotation—that's LPs getting scared off by low yields and high uncertainty.
Hidden risk: Token Inflation
zkSync has not yet launched a token. But when it does, expect it to be used to subsidize proving costs through token emissions to sequencers or LPs. That's the same playbook as DeFi liquidity mining—artificially boost activity to keep the network alive. But here's the problem: ZK proving costs are a real dollar expense, not a virtual yield. So the token must have real value to pay real validators. That creates a feedback loop of sell pressure.
I've seen this before. During DeFi Summer 2020, projects printed tokens to pay for TVL. When yields dropped, TVL vanished. Token inflation to cover operational costs is a variant of the same ponzinomics.
Takeaway: What to Watch Next
I'm not saying zkSync fails. I'm saying the current market structure—sideways, low volume, high ETH gas relative to L2 fees—exposes the fragility of the business model. If you hold ZK rollup tokens or are farming airdrop points, ask yourself:
- Is the protocol's revenue organic, or subsidized by incentives?
- Can proving costs drop by a factor of 10 through better hardware or aggregation (e.g., recursive proofs)?
- What happens when VC funding dries up?
The answer will determine whether zkSync Era survives the next two years or becomes another cautionary tale in my audit notebook.
Yields were too good to be true, so we didn't chase them in DeFi. Now the yields are invisibly flowing to validators in the form of L1 gas. That's the silent bleed. Watch the proof costs. They're the canary in the ZK coal mine.