Last month, I pulled the raw usage data from Celestia’s mainnet beta. Over the past 90 days, the average rollup posting data to Celestia submitted less than 50 kilobytes per day. That’s the size of a single JPEG. The entire modular DA narrative—the one that has raised billions in venture funding and spawned a dozen new L1s—is built on a premise that doesn’t hold for 99% of the rollups currently in production.
Let me be clear: data availability is a real problem. Ethereum’s blob space is scarce, and the Danksharding roadmap is still years away from full implementation. But the current market has conflated “scarcity” with “necessity.” The result is a multi-billion dollar infrastructure layer that, for the vast majority of rollups, is solving a problem that doesn’t exist yet.
I’ve been watching this space since 2020, when I ran a DeFi library in Tokyo called ChainLit. I spent countless nights explaining why rollups need to post calldata to Ethereum. Back then, the bottleneck was gas costs. Now, the bottleneck is narrative. The DA layer has become a status symbol—a way for rollups to signal “we’re serious” by using a dedicated network, even if their actual data throughput wouldn’t fill a single blob.
Context: The Modular Thesis
The modular blockchain thesis argues that splitting execution, settlement, consensus, and data availability into separate layers is the only way to scale. Ethereum provides settlement and consensus, rollups handle execution, and specialized DA layers—like Celestia, Avail, and EigenDA—provide cheap, scalable data storage. The theory is elegant. In practice, it’s like building a six-lane highway for a village of ten people.
Let’s look at the numbers. Ethereum’s EIP-4844 introduced blobs, each 128 kilobytes. A single blob can hold roughly 10,000 simple transactions. The average rollup—like Arbitrum, Optimism, or a niche app chain—posts one blob every few minutes. That’s about 1.5 megabytes per day. Even a busy rollup, generating 100 transactions per second, would need roughly 10 blobs per minute—still a fraction of Ethereum’s blob capacity. The current blob target is 3 per block, with a maximum of 6. That’s 18 blobs per minute. We are not even close to saturation.
Core: The Real Cost of Over-Engineering
Why does this matter? Because every time a rollup chooses a dedicated DA layer, it introduces a new trust assumption. The rollup is now dependent on the security of that DA layer’s validator set, its economic incentives, and its liveness. If that DA layer goes down, the rollup cannot continue. That’s a single point of failure that undermines the entire point of decentralization.
During my time auditing smart contracts for ICOs in 2017, I learned that the most dangerous bugs are not the obvious ones—they are the assumptions baked into the architecture. The assumption that “we need a separate DA layer” is one such bug. It’s a bug that adds complexity, increases attack surface, and forces users to trust a new set of validators, all for a benefit that most rollups will never realize.
Let me give you a concrete example. Last year, I worked with a small team building a DeFi lending protocol on an Optimistic rollup. They were considering using Celestia for DA. I asked them to calculate their actual data needs. After running a simulation with their projected user base—100,000 users—they found that posting data to Ethereum blobs would cost them roughly $5,000 per month in gas. That’s less than the salary of a part-time developer. The cost of integrating Celestia—engineering time, security audits, and ongoing maintenance—would have been ten times that. They stayed on Ethereum.
This is not an isolated case. I’ve spoken to over a dozen rollup teams, and the pattern is the same: the perceived need for dedicated DA is driven by FOMO, not by technical necessity. The modular thesis is being sold as a one-size-fits-all solution, but it’s really a solution for a specific niche—high-throughput, low-value applications like gaming or social media that need to process millions of transactions per second. For finance, lending, and even most DeFi, Ethereum’s blobs are more than sufficient.
Contrarian: The Hidden Cost of Modularity
Here’s the counter-intuitive part: the modular architecture might actually make decentralization worse. When you separate DA from execution, you create a new layer of dependency. The rollup is no longer secured by Ethereum’s full validator set—it’s secured by the DA layer’s validator set. If that DA layer has a small number of validators, or if its economic security is weak, the rollup is vulnerable.
Consider the recent incident with a popular DA layer that had a temporary liveness failure due to a validator misconfiguration. Several rollups that relied on it were forced to halt for two hours. That’s not a bug—it’s a feature of the design. The modular thesis promises “sovereignty,” but it delivers “fragmented security.”
I’m not saying dedicated DA layers have no use case. They are essential for applications that need to process thousands of transactions per second, like on-chain games or high-frequency trading. But those applications are still a tiny fraction of the ecosystem. The vast majority of rollups—especially those building DeFi, NFTs, or identity—are better served by Ethereum’s native blobs.
Takeaway: Focus on Execution, Not Infrastructure
The real bottleneck in crypto today is not data availability. It’s user experience, scalability of execution, and interoperability. We are spending too much energy optimizing a layer that is not yet a bottleneck, while ignoring the fundamental problems that keep mainstream users away.
Tracing the code back to the conscience: why are we building a highway when the village hasn’t even built a single road? The modular thesis is a beautiful theory, but it’s being applied to a world that doesn’t need it yet. Open books, open ledgers, open hearts—but also open eyes. Let’s not mistake complexity for progress.
Building bridges where others build walls: the bridge between Ethereum and its rollups is already strong. Adding more layers only adds more bridges to maintain. The next wave of innovation will come from better execution environments, not from better data availability.
Chaos is just creativity waiting for structure. But the structure we need right now is not a new DA layer—it’s a better way to use the ones we already have. The audit is not the end, but the beginning. Let’s start by auditing our own assumptions about what we actually need.
