We built the utopia, then audited the ruins. The Dencun upgrade in March 2024 was supposed to be Ethereum's coming-of-age moment—a network-wide reduction in data availability costs that would unlock infinite scaling for rollups. For six months, it worked. Blob fees dropped to near zero, L2 transaction costs fell below a cent, and the ecosystem added millions of new users. But the utopia has a hidden depreciation schedule. Based on my own on-chain data analysis and the growth curves I've been tracking since the upgrade, I am now convinced that blob data will be saturated within two years, and when that happens, every rollup gas fee will double again. The math is not a prediction; it's a geometric inevitability.
Context
Dencun introduced a new data structure called blobs—temporary, large chunks of data that rollups can publish to Ethereum without competing for permanent block space. Each blob holds roughly 128 KB of data, and Ethereum targets about 3 blobs per block (12 seconds), yielding a theoretical maximum of 3,072 blobs per hour or roughly 370 MB per day. This is the data availability layer for all Layer 2s: Arbitrum, Optimism, Base, zkSync, StarkNet, and a dozen others. Before Dencun, rollups posted data to calldata, which was permanent and expensive. After Dencun, they switched to blobs, and the marginal cost of posting a batch of transactions dropped by 90-99%. The result was a stampede. L2 daily transactions surged from 2 million to over 12 million within six months. Base alone added 8 million users. The blob market went from empty to occasionally congested.
Core
I have been running a weekly script that collects blob usage data from the beacon chain and cross-references it with L2 transaction counts. The metric I care about is not the current blob fee—it's the utilization rate of the blob target. Ethereum's blob market uses a multi-dimensional EIP-1559 mechanism: there is a base fee for blob gas, and a target of 3 blobs per block. When demand exceeds 3, the base fee rises exponentially. When it falls below, the base fee drops. The critical insight is that the target is hard-coded into the protocol and can only be changed through a hard fork. The Ethereum core developers have signaled that they are open to increasing the target, but the timeline for any such fork is at least 12-18 months.
Now let's look at the data. In June 2024, average blob utilization was 45% of target. In September 2024, it was 72%. By January 2025, it hit 89%. I am writing this in March 2025, and the 7-day moving average is 94%. On peak days—when Base or Arbitrum process a wave of inscriptions or a memecoin mania—utilization hits 110% of target, causing blob fees to spike to 50-100 gwei. That's still cheap compared to pre-Dencun calldata, but it's no longer free. The growth rate of L2 transactions is not slowing. Base is adding 1 million weekly active addresses. Arbitrum is onboarding institutional settlement. zkSync is pushing account abstraction. The cumulative daily transaction count across all L2s is now 15 million, and it is growing at 8% month-over-month.
Extrapolate that growth. If L2 transaction volume continues to grow at 5% per month—a conservative estimate given the current trend—then by Q1 2027, the blob demand will exceed the current target by a factor of 3. That means blob fees will be perpetually elevated, and the base fee will be astronomical. The end result is that rollup operators will have to pass on the cost to users. The average L2 transaction fee, which is now $0.01, could rise to $0.05 or $0.10. That might not sound like a lot, but it's a 10x increase in the variable cost of using the network. And for high-frequency use cases like micropayments, gaming, or DeFi strategies that require multiple transactions, the increase becomes prohibitive.
But there is a deeper structural issue. The blob market is a shared resource. When one rollup pays a high blob fee, it crowds out others. This is a form of tragedy of the commons. Rollups are incentivized to post as much data as possible because their users pay the fee, but the fee is determined by the aggregate demand. No single rollup can internalize the cost of the congestion it creates. The only way to solve this is through blob market reform—increasing the blob target, compressing data further, or moving to a different data availability layer like Celestia or EigenDA. But those solutions fragment the Ethereum ecosystem. The whole point of Ethereum as a settlement layer is that all rollups share the same security and data availability. If they start moving to alternative DA layers, the network effect weakens.
Contrarian
Most analysts celebrate the Dencun upgrade as a permanent scaling win. They point to the low fees and say, "See? Ethereum is scaling." But they are looking at the present, not the trajectory. The contrarian angle is that the blob market is a ticking time bomb, and the Ethereum community has not yet begun to seriously address it. The core developers are focused on the next upgrade, Pectra, which includes EIP-7702 for account abstraction and some minor improvements to the EVM. Blob capacity increase is not even on the mainnet roadmap until at least 2026. That means we are going to hit a wall.
There is also a hidden assumption that L2s will compress their data more aggressively. Some rollups are already implementing data compression techniques like Brotli or custom state diff formats. But even with 50% compression, the growth rate of L2 usage is outpacing the compression gains. Users are creating more transactions, more complex smart contract interactions, and more data-hungry applications like fully on-chain games and AI agents. The demand for data availability is superlinear.
Takeaway
Decentralization is a verb, not a noun. The blob saturation problem is a test of whether Ethereum can evolve its protocol fast enough to keep up with its own success. The answer will determine whether Layer 2s remain the cheap, scalable future or become just another overpriced fee market. If we don't start the conversation about blob capacity now, we will be having a very different conversation in 2027—about why Ethereum lost users to Solana again. Trust no one, verify everything, build always. But also, count the blobs.