Pump, dump, debug. Repeat.
I ran the numbers this morning — and they aren't pretty. zkSync Era spent roughly $1.7 million on proof generation in the last 30 days, based on their public sequencer data. That's about $0.23 per transaction in proving cost alone. Their average transaction fee? $0.18. You don't need a PhD in tokenomics to see the math doesn’t close.
This isn't a zkSync-specific problem. Starkware's SHARP aggregator is burning through $3M+ monthly on Ethereum calldata and proof verification. Scroll, Linea — same story with different shades. The bull market euphoria has everyone staring at TVL and price action, but the proving cost time bomb is ticking under every ZK rollup.
Why Now?
Right now, we’re in a full-on bull run. Bitcoin pushing $120k, ETH flirting with $8k, and L2 tokens pumping on “ZK narrative” hype. But here’s the dirty little secret: ZK proving costs scale linearly with compute, not with transaction value. When ETH gas spikes, proving costs don’t move in sync — they’re tied to the sequencer's hardware spend, not L1 demand. So as transaction volume explodes, the fixed cost per proof doesn't drop; it actually rises if they need more prover machines.
Meanwhile, L2 transaction fees are compressed to near-zero to attract users. Gas fees higher than the yield. Typical. Only this time, the yield is negative for the rollup operator.
Let me rewind for the newbies. A ZK rollup batches thousands of transactions off-chain, generates a single cryptographic proof (a zk-SNARK or STARK), and submits it to Ethereum. The cost has two parts: proving cost (generating the proof off-chain — computes resources) and verification cost (checking the proof on Ethereum — gas). In a bull run, verification cost goes up because ETH gas goes up. But proving cost also goes up if the rollup operator scales their prover fleet to handle more load.
The market expects that these costs will eventually be amortized over millions of transactions. That’s true — but only if transaction volume grows faster than proving cost growth. Right now, it’s not.
The Core: My Audit of ZK Rollup Economics
I went deep into the on-chain data. Here’s what I found for the top four ZK rollups (all data as of Q2 2026):
| Rollup | Daily Proving Cost | Daily Transaction Fees | Daily Loss (Subsidy) | |--------|---------------------|-------------------------|-----------------------| | zkSync Era | $56k | $42k | -$14k | | Scroll | $38k | $31k | -$7k | | Linea | $45k | $35k | -$10k | | Starknet | $110k | $63k | -$47k |
Bold: Every major ZK rollup is losing money on every transaction. They’re subsidizing the difference with token sales or foundation treasuries. Starknet is burning the most — almost $50k per day.
“But they’ll make it up in volume,” you say? Let’s test that. If transaction volume doubles, proving cost doesn’t double — it actually increases because prover hardware must be upgraded to keep latency low. The per-transaction proving cost might drop by 20% with better batching, but not below $0.15 given current hardware constraints. Meanwhile, users expect fees under $0.05. The gap only widens.
I’ve been coding smart contracts since the 2017 ICO sprint. Back then, I learned to trust code, not whitepapers. So I pulled the actual proving contracts for zkSync — their Boojum prover. The circuit is optimized, but the recursion overhead is massive. Every proof triggers a recursive verification that costs roughly 300k gas on L1. At current ETH price, that’s $12 per batch. They batch 300 transactions — that’s $0.04 per tx just for L1 verification. Add the off-chain proving compute: a GPU cluster running 24/7 at $200/hr — that’s $4,800/day. Spread over 1.2 million daily transactions? $0.004 per tx. So total cost per tx: $0.044. Fees? $0.018. t check.
The Contrarian: What Everyone Misses
The common narrative is that ZK rollups are the inevitable future — trustless, fast, cheap. They’ll eat Optimistic rollups for breakfast. But the unreported angle is that this bull market is actually their worst enemy.
Why? Because high ETH gas inflates verification costs, but more importantly, high token prices encourage operators to subsidize losses with token inflation. Every ZK rollup has a native token that’s pumping. They can afford to burn cash today because the token sale fills the treasury. But that’s a Ponzi-light structure: sell tokens to users, use proceeds to pay for proving, keep fees low to attract more users, which pumps the token further. Rinse, repeat. The moment the bull market turns, token prices collapse, the subsidy disappears, and either fees skyrocket or the chain slows to a crawl.
Optimistic rollups like Arbitrum and Optimism have a different cost profile — they only pay L1 data costs (calldata or blobs), which are proportional to data size, not compute. Their per-tx cost is currently around $0.01. They’re profitable. The ZK camp is not, and may never be, unless proving hardware improves by an order of magnitude.
I interviewed a lead engineer from one of these rollups at a Buenos Aires tech meetup last week. Off the record, he admitted: “We’re burning $2M per month. The foundation is fine with it for now because the token is up 400%. But if the market drops, we either raise fees or dilute.” He laughed nervously. I didn’t laugh. I’ve seen this movie before — the 2020 DeFi yield farming collapse, the 2022 Celsius implosion. When the music stops, the one with the highest burn rate dies first.
Another blind spot: the centralization of proof generation. All major ZK rollups currently run their own prover nodes — a single entity generates all proofs. That’s a single point of failure and a censorship vector. The narrative says “decentralized ZK rollup,” but the proving layer is still a permissioned cluster. If that cluster goes down, the entire rollup halts. We saw that with zkSync’s two-hour outage in April 2026. They blamed a “prover bug,” but the community had no recourse because no one else can generate proofs.
Takeaway: What to Watch Next
I’m not saying ZK rollups are doomed. The technology is brilliant. But the economics are propped up by bull market subsidies. Bold: If ETH gas remains above 500 gwei for six more months, expect at least two ZK rollups to pivot to a fee hike or a hybrid model using optimistic execution.
The key metric isn’t TVL or TPS — it’s proving cost per transaction vs. average fee. If that ratio stays above 1.0 for a sustained period, the project is essentially a charity. And charities don’t last long in crypto’s kill-or-be-killed cycles.
Keep your eyes on the next generation of proving hardware — custom ASICs or GPU clusters specifically for zk-SNARKs. If a team announces a 5x cost reduction, they win. Otherwise, the ZK rollup narrative might hit a very hard wall when the next bear market arrives.
Pump, dump, debug. Repeat. The debugging isn’t done yet.
