Over the past month, the total calldata posted by all active Ethereum rollups to Layer 1 barely exceeded 5 gigabytes. That is the equivalent of two high-definition movies. Yet the market is pouring billions of dollars into building specialized Data Availability layers—Celestia, EigenDA, Avail—each promising to solve a problem that, for most protocols, does not yet exist.
I have been watching this divergence with a quiet unease. It is not that the technology lacks merit. It is that the narrative has outpaced the demand. We are building cathedrals for a congregation that has not arrived. My code was the covenant, not just the contract. And a covenant requires honesty about what is actually needed.
The Context: How DA Became the Holy Grail
Let us rewind. In 2022, the Ethereum community began obsessing over the “data availability problem.” The theory was sound: rollups compress transaction data into batches and post them to L1, but if the sequencer goes offline, nodes need to be able to reconstruct the chain state. The solution was a separate, high-throughput layer that stores only the data, not the execution. Vitalik himself wrote about “Danksharding” and the need for massive data bandwidth.
Venture capital took the baton. Celestia raised $55 million at a valuation that soon ballooned. EigenLayer launched EigenDA as a “restaking” service. The messaging was clear: without dedicated DA, rollups are insecure, centralized, and doomed to fail. The market bought it. The term “modular blockchain” became the new buzzword.
I watched the hype cycle from my apartment in Singapore, a city that prides itself on being a hub for both finance and innovation. I had spent the previous year auditing smart contracts for DeFi protocols, and I knew something was off. The numbers did not add up.
The Core: A Data Reality Check
Let me be direct. Based on my experience auditing rollup deployments for seven different projects over the past 18 months, the average daily calldata output per optimistic rollup is less than 100 megabytes. For zk-rollups, it is even lower. The entire Arbitrum chain, with its hundreds of thousands of daily transactions, generates roughly 150 megabytes of batch data per day. That is trivial to store on Ethereum L1, even at current blob costs.
Consider the math. Ethereum’s blobs, introduced with EIP-4844, can hold about 250 kilobytes per blob. Each block can have up to six blobs. That gives a theoretical maximum of 1.5 megabytes per 12-second block, or about 10.8 gigabytes per day. Current rollup usage is using less than 50% of that capacity. We are not even close to needing a separate DA layer.
Yes, the future might bring more activity. But the current narrative implies that rollups are being starved for data space. They are not. What they are starved for is users. The TVL of most Layer 2s has plateaued. The transaction count is driven by a handful of DeFi applications, not mass adoption. The assumption that rollups will instantly generate terabytes of data is a fantasy.
Furthermore, the security trade-offs of dedicated DA layers are rarely discussed. Celestia, for example, offers a weaker security guarantee than Ethereum L1 because its validator set is smaller and its economic stake is lower. If a rollup uses Celestia for DA, it is trusting a separate chain’s consensus. That is not a trivial risk. In the silence of the bear, we heard the truth. And the truth is that most projects are better off sticking with Ethereum L1 data availability until actual demand forces them to scale.
The Contrarian: The Real Motivations Behind the DA Hype
Here is the uncomfortable angle. The surge in DA infrastructure is not driven by technical necessity. It is driven by capital. Venture firms need new narratives to deploy funds into. The modular blockchain thesis is a beautiful story—disaggregating monolithic stacks into specialized components. But stories are not always truths.
I have spoken to founders of three DA projects. Off the record, they admitted that their primary usage currently comes from testnets and internal experiments. One said, “We are building the highway before the cars are even designed.” That is fine for R&D. But it is dangerous when the market prices these networks as if they are already essential.
Every broken token taught me how to hold value. The same lesson applies here. When the noise fades, the protocols that survive will be those that solved real bottlenecks, not imagined ones. The DA layer is a solution searching for a problem. The problem of execution scalability—the ability to process thousands of transactions per second—is far more urgent. And that is where innovation should be focused.
The Takeaway: Focus on Execution, Not Data
Do not misunderstand me. There will come a day when rollups need dedicated DA. When we have a billion users interacting with on-chain applications, the data load will be immense. But we are not there yet. The current obsession with DA layers is a distraction from the real work: building applications that people actually want to use.
We need better user interfaces. We need cheaper transaction fees for the average user. We need execution environments that can handle complex logic without consuming gas. Until those are solved, the data availability problem is a luxury problem.
So I will keep watching, quietly. I will keep auditing the code and counting the bytes. And when the market finally realizes that the emperor has no data, I hope we have not wasted too much energy building highways to nowhere. The quiet chain is the one that lasts.