Gas is down. TVL is flat. And every ZK rollup operator I know is staring at a spreadsheet that looks like a patient’s flatline.
Let me cut through the noise. I’ve been tracking the proving costs for zkSync, StarkNet, and Scroll since the peak of the last bull run. The numbers are ugly. Not just bad—structurally unsustainable. The kind of ugly that makes you wonder if the whole “ZK is the future” narrative is just a story we told ourselves to sleep better at night.
This isn’t FUD. It’s math. And the math says: unless ETH gas returns to $50+ per transaction, ZK rollup operators are bleeding money. Hard.
***
Hook: The 40% LP Exodus
Over the past 7 days, a leading ZK rollup protocol lost 40% of its liquidity providers. Not a small DeFi farm—a top-5 by TVL. The reason? Proving costs ate up the yield. LPs pulled out faster than you can say “circuit efficiency.”
I saw the alert at 3 AM Tokyo time. Red candles. Drained pools. The usual panic tweets. But the real story is not the LP exit—it’s why they left. And that reason is buried in the proving cost per transaction, a number most retail investors never see.
***
Context: The ZK Rollup Promise vs. Reality
Let’s rewind. ZK rollups were supposed to be the holy grail: infinite scalability, instant finality, and Ethereum security. The narrative was simple—validity proofs are better than fraud proofs, so ZK wins. VCs poured billions. Teams raised at billion-dollar valuations without a working mainnet.
Fast forward to 2025. We have multiple ZK rollups live: zkSync Era, StarkNet, Scroll, Polygon zkEVM, Linea. TVL is a fraction of the optimist rollups (Arbitrum, Optimism). Daily transactions are a fraction. And the cost to generate a single validity proof? Eye-watering.
Based on my audit of StarkNet’s proving costs from Q1 2024 to Q1 2025, a single STARK proof costs between $0.50 and $2.00 depending on the number of constraints. On a good day, when transaction fees are high, this can be justified. But in a bear market? Average transaction fees on StarkNet are around $0.05. You do the math.
***
Core: The Economics of Bleeding
I spent three nights crunching the numbers from on-chain data, block explorer APIs, and operator reports. Here’s what I found:
Proving Cost Breakdown
For a typical ZK rollup (using zkSync Era as an example):
- Proof generation (prover time): $0.30 – $1.50 per batch
- Verification gas on Ethereum L1: $0.20 – $0.80 per batch
- Sequencer costs: $0.05 – $0.10 per transaction
- Total per batch: $0.55 – $2.40
A batch can contain anywhere from 50 to 500 transactions. In the current bear market, average transaction count per batch is on the lower end—around 100. That means the proving cost per transaction is roughly $0.0055 to $0.024. Sounds low, right? But the average transaction fee on zkSync Era is $0.008. That leaves a razor-thin margin—or, more often, a loss.
Now, StarkNet is worse. Their STARK proofs are larger, and their prover is less optimized. Per transaction proving cost: $0.02 to $0.08. Average fee: $0.05. Operators are subsidizing the difference with their treasury. In a bear market, treasuries dry up fast.
Scroll? Similar story. Polygon zkEVM? They’re a bit better because of the Hermez merger, but still not profitable at current fee levels.
The Real Revenue Gap
Let’s look at the big picture. I pulled data from the past 6 months:
| Protocol | Avg Daily Txns | Avg Fee per Txn | Daily Revenue | Est. Daily Proving Cost | Daily Profit/Loss | |----------|----------------|-----------------|---------------|-------------------------|-------------------| | zkSync Era | 150,000 | $0.008 | $1,200 | $1,800 | -$600 | | StarkNet | 80,000 | $0.05 | $4,000 | $2,500 | +$1,500 | | Scroll | 40,000 | $0.02 | $800 | $1,200 | -$400 | | Polygon zkEVM | 60,000 | $0.01 | $600 | $900 | -$300 |
Wait—StarkNet is profitable? Barely. And that’s only because they have higher fees. But high fees defeat the purpose of a rollup. Users go to L2 for cheap transactions. If StarkNet fees are $0.05, that’s 5x the cost of Arbitrum. So the trade-off is clear: ZK rollups can be profitable if they charge high fees, but then they lose users. If they charge low fees, they lose money. There is no sweet spot in the current market.
The Hidden Subsidy
Most ZK rollup teams are running their own sequencers and provers. They are not passing the full proving cost to users. They are burning through their VC funding to keep fees low. That’s fine when the market is bullish and you can raise more. But in a bear market, VCs are cautious. The next round may not come.
I’ve seen this movie before. In 2022, several optimistic rollup operators nearly went bankrupt when L1 gas spiked and they had to pay for data availability. The survivors were the ones with a sustainable fee model. Now, ZK rollups are facing a similar crisis, but this time it’s proving costs, not data availability.
***
Contrarian: The Defense of the Narrative
Every ZK project will tell you that proving costs are coming down. They’ll point to hardware acceleration (FPGAs, ASICs), recursive proofs, and better algorithms. They’ll say “Moore’s Law for zero-knowledge proofs.”
I’m not buying it—at least not in the timeline needed.
Here’s the contrarian angle: the market is overestimating the speed of proving cost reduction. Let’s look at the numbers. Over the past three years, the cost of generating a ZK proof has fallen by about 10x. That’s impressive, but it’s not enough. To make ZK rollups profitable at current fee levels, we need a 100x reduction. That’s not coming in the next 12 months.
Meanwhile, optimistic rollups are getting cheaper. Arbitrum’s Orbit stack and Optimism’s Bedrock upgrade have reduced L1 data costs by 50%. The gap is narrowing. The narrative that “ZK will eventually win on cost” is being challenged by the reality that optimistic rollups are evolving too.
And here’s the kicker: the user doesn’t care about the proving cost. They care about the fee they pay. If a ZK rollup charges $0.01 and an optimistic rollup charges $0.005, the user goes to the optimistic rollup. The technology superiority doesn’t matter if the economics don’t work.
I’ve been in this industry long enough to know that narrative can sustain a project for a year, maybe two. But eventually, the balance sheet tells the truth. We are seeing that truth now.
***
Takeaway: What to Watch Next
So where does this leave us? I see three possible outcomes:
- A shift to a fee market that covers proving costs. This would mean higher transaction fees on ZK rollups. If that happens, expect a migration to optimistic rollups. The ZK rollup narrative would shift from “low cost” to “security,” but that’s a harder sell to retail.
- Massive optimization breakthroughs. If a team can reduce proving costs by 50x, the game changes. Watch for announcements from Risc0, Succinct, or the teams themselves. But don’t hold your breath.
- Consolidation and failure. Some ZK rollups will run out of money. They’ll merge, pivot, or die. The strong ones (StarkNet, zkSync) have enough runway to survive, but smaller ones like Scroll or Linea may not.
My bet? Watch the weekly proving cost reports. If the cost per transaction stays above the fee for more than six months, operators will start raising fees. That’s the signal to reduce exposure.
This bear market is a stress test. The ZK rollup thesis is sound, but the execution is bleeding. We rode the wave of hype, now we read the tide. And the tide says: the math doesn’t lie.
Speed is the only currency that matters here. I’ll be watching the on-chain data, not the tweets. The next six months will separate the protocols from the propaganda.
