Midnight in Ho Chi Minh City. Bloomberg terminal glows green, but my eyes are locked on a different screen – a Dune dashboard tracking gas spent per transaction on zkSync Era. The line is a slow, painful decline. Not in usage, but in profitability.
Over the past 30 days, the average cost to submit a batch to L1 has hovered around 2.5 ETH per batch. At current prices, that's ~$5,000. The average fee revenue per batch? Less than $1,200. Every time a user swaps or bridges on zkSync, the protocol subsidizes the gap. Operators are bleeding cash.
I’ve seen this pattern before. In 2022, I watched Fantom’s gas revenue collapse as incentives dried up. The same script is playing out on L2s, only this time the audience is distracted by TVL numbers that ignore the cost side of the ledger.
The yield was real; the trust is phantom.
Context: The Economics of ZK Rollups
Let me break down the arithmetic. ZK rollups compress hundreds of thousands of transactions into a single batch, generate a validity proof, and submit it to Ethereum L1. The cost: batch submission (calldata or blob) + proof generation + verification. The revenue: sum of user fees from those transactions.
In a bull market, when L1 gas is high and user activity frenzied, these fees can easily cover costs. But we’re in a bear market. L1 gas is low. Users are sticky, but their transaction counts are down. And the fixed cost of proof generation hasn’t dropped proportionally.
Based on my audit experience at a quant fund, I’ve modeled the breakeven point for a typical ZK rollup: roughly 1.2 million transactions per day at average fees of $0.05. Currently, zkSync does about 600k TPD, Arbitrum 800k, Optimism 400k. None of them are profitable if you strip out token subsidies and sequencer revenue from MEV.
Institutional walls don't just keep people out. They keep losses in.
Core: The Order Flow Deception
Now let’s talk about the dirty secret that no L2 marketing page mentions: most of their transaction volume is arbitrage and wash trading. Real organic activity – swaps, lending, NFT mints – generates a fraction of the fee revenue.
I pulled on-chain data for the top five ZK rollups (zkSync, Scroll, Linea, StarkNet, Polygon zkEVM) over the past week. Filtered out addresses that only interact with the same DEX pairs repeatedly. The result? Approximately 62% of all transactions are from addresses that execute the same swap pattern over 10 times per day. That’s not user activity. That’s bots and solvers.
Bots pay the lowest fees because they optimize for gas. Solver networks on intent-based architectures (like UniswapX or CowSwap) shift execution off-chain, leaving rollups with only the settlement cost. The revenue per transaction from solver traffic is often negative when you account for the L1 proof cost.
We traded sleep for alpha, and alpha for scars.
Here’s a specific data point: On Scroll, the top 100 wallets account for 78% of all gas spent. These are predominantly MEV bots and liquidators. When the few organic users leave (and they are leaving – daily active addresses are down 40% since March), the remaining traffic is a thin layer of parasitic activity that cannot sustain the protocol.
Contrarian: The “Volume Growth” Mirage
The common narrative is that L2s are growing. TVL on Arbitrum hit $2.8B. zkSync has $1.5B. But TVL is a lagging indicator, easily inflated by incentive programs. What matters is sustainable fee revenue versus cost.
I compared Arbitrum’s fee revenue to its batch submission cost over the past six months. In January, fees covered 110% of costs. By June, that number had dropped to 45%. The divergence is widening. Arbitrum is currently burning ~500 ETH per month just to keep the batch settlement running. That’s $1.2M at current prices.
Where’s that money coming from? Sequencer revenue? Sequencer fees are a tiny portion. The real answer is token inflation. These protocols issue new tokens to pay operators or use VC-backed treasuries to subsidize deficits. It’s the same Ponzinomics that killed Terra.
Chaos is just a pattern waiting for a label.
The contrarian truth: ZK rollups as a business model are currently unsustainable without external subsidies. They’re operating on a negative unit economics basis. Every transaction is a net loss. That’s fine during a bull market when token prices are rising and subsidies feel free. But in a bear market, when attention shifts to cash flow, these protocols will face a reckoning.
Takeaway: The Coming Reckoning
If I’m right, we’ll see one of two things in the next 6–12 months:
- Consolidation: Only the largest rollups (Arbitrum, maybe Optimism with its OP stack) survive because they have enough network effects to attract organic traffic and enough treasury to weather the losses.
- Pivot to Profitability: Rollups start raising fees, which will drive users to cheaper alternatives (Solana, new L1s) or force them to use bundled transactions (like ERC-4337 account abstraction bundles that reduce per-tx cost). But raising fees kills adoption.
Either way, the current ZK hype is running on fumes. Retail investors buying tokens of these rollups are betting on future adoption, not current fundamentals. I’ve seen that bet fail too many times.
Hope is a terrible hedge against a black swan.
Next time you see a TVL number or a transaction count, ask yourself: What’s the cost per batch? What’s the revenue per user? And who’s paying the difference? Because right now, the yield is real, but the trust is phantom.