Everyone thinks ZK Rollups are the holy grail of Ethereum scaling. The narrative is loud: infinite throughput, absolute security, and a post-EIP-4844 utopia where gas costs vanish. But the on-chain data tells a different story. Over the past three months, I have been tracking the transaction profiles of the top five ZK Rollup projects by TVL. What I found is a pattern eerily reminiscent of the NFT wash-trading days—volume without intent, liquidity without users. Let me show you the evidence.
But first, a confession. I was a ZK true-believer in 2021. I audited one of the earliest ZK-snark contracts for a now-defunct ICO. That experience taught me to respect the math but distrust the marketing. So when I saw the latest batch of ZK Rollups touting '100,000 TPS' and 'sub-cent fees,' my scepticism kicked in. I set up a Python script to pull live data from Etherscan, L2Beat, and Dune Analytics, filtering for unique contract interactions, gas expenditure, and wallet clustering. The results were sobering.
Let’s get the methodology out of the way first. I defined 'active users' as wallets that initiated at least two non-trivial transactions (value > $1) on the L2 within a 7-day window. I excluded CEX deposit addresses and bridging contracts because they skew the signal. I then measured 'real throughput' by counting only transactions that executed user-intent logic (swaps, transfers, interactions with non-infrastructure contracts). I call this the 'Intent-to-Transaction Ratio' (ITR). A high ITR means users are actually using the chain for meaningful actions. A low ITR means the chain is just a conduit for bots and bridge operators.
Now, the core discovery. For the top ZK Rollup by TVL (let's call it 'Project A'), the ITR over the last 30 days is a paltry 0.12. That means 88% of its 'transactions' are either bridge operations, batched fee payments, or spam from a cluster of 15 wallets that appear to be mechanically looping calls to a single DEX contract without any meaningful slippage or outcome. I traced these wallets back to a single deployer address funded by the project’s own treasury. The 'volume' is fake. The 'users' are ghosts.
Volume without intent is just digital noise.
Project B, a ZK-optimistic hybrid, boasts a '600,000 TPS' on its landing page. On-chain, it processed an average of 2.4 actual user transactions per second over the past week. The rest? Proof generation and verification overhead. The chain’s native token has appreciated 400% in three months based on this narrative, but if you strip out the automated traffic, the fundamental activity is lower than a dead Ethereum testnet. This is not scaling; it’s staged performance.
But here is where it gets even more deceptive. The projects are burning significant ETH on L1 for data availability. Project A spent 1,200 ETH last month just to post its batch data to Ethereum mainnet. At current gas prices, that’s roughly $2.4 million. Meanwhile, the fees it collected from users (excluding itself) were just $18,000. That is a 133x subsidy. This is not a sustainable business model; it is a money-losing operation propped up by venture capital.
I shared these findings with a few analyst friends at major funds. Some pushed back, saying I was ignoring the 'future potential' of ZK and that 'user growth takes time.' That is the same argument I heard from the Terra/Luna team in early 2022. Circular liquidity does not become sustainable just because you believe in the tech. The on-chain data should be the anchor, not the narrative.
So what is really happening? These projects are gaming the metrics that VCs and retail investors celebrate. TVL is inflated by their own treasury deposits. Transaction count is bloated by bot activity. Unique wallet count includes dust accounts created solely to claim airdrops. The signal-to-noise ratio is abysmal. And the industry is buying it—hook, line, and sinker.
The contrarian angle is this: maybe the ZK Rollup thesis is not flawed, but its current implementation is. The proving costs are too high for real-world adoption. The latency is too long for DeFi to feel like CeFi. The developer tooling is still clunky. But the biggest blind spot is that we are measuring the wrong metric. TVL and TPS are vanity. The only metric that matters is sustainable, non-incentivized user intent. And on that front, every project I looked at is failing.
I have been through two cycles now. In 2017, we measured GitHub commits. In 2020, we measured yield. In 2025, we measure 'killer dApps' that are just wrapped V3 Uniswap forks with a new paint job. The pattern repeats because the incentives are always to sell a story, not to build utility.
What should you do? Look at the contract interactions. Use tools like Tenderly to replay transactions and see if they are real user paths or spam. Check the gas consumption per user. If a project claims 1 million users but its total gas spend is less than that of a cheap ENS mint campaign, you know something is off. Don't let the narrative blind you to the data.
Six months from now, I will rerun this analysis. If the signals change—if I see organic NFT mints, real derivative volume, or cross-chain stablecoin transfers originating from new wallets—I will update my thesis. But until then, I am shorting the hype and waiting for the washout.
The market will eventually punish these mirages. It always does.