The DA Layer Mirage: Why 99% of Rollups Don't Need Dedicated Data Availability

For the past eighteen months, the crypto narrative has been obsessed with a single bottleneck: data availability. Projects like Celestia, Avail, and EigenDA have raised hundreds of millions of dollars on the premise that rollups are drowning in data and need a dedicated highway to offload their blobs. The pitch is seductive: modular blockchains, specialized layers, scalability without compromise. But after auditing the on-chain activity of over forty rollups and talking to a dozen operators, I’ve come to a quieter conclusion. The narrative is running ahead of the reality. The data availability layer is not the bottleneck the market believes it to be. It’s a solution in search of a problem, and most rollups are generating so little data that a dedicated DA layer is both an over-engineering and a distraction. Listening for the quiet hum of the second layer. Let me rewind. The thesis for modular blockchains emerged from a valid observation: monolithic chains like Ethereum face an inherent trade-off between security, decentralization, and scalability. Rollups solve execution by moving computation off-chain, but they still need to post compressed transaction data on-chain to inherit security. As rollup usage grows, the argument goes, the calldata or blob space on Ethereum will become congested and expensive. Therefore, we need alternative DA layers that can cheaply store and verify this data. The logic seems airtight, but it relies on an implicit assumption that rollup data generation will explode exponentially. In practice, that assumption has not held. Mapping the ghosts in the machine of trust. I spent the first quarter of 2026 pulling data from the leading rollups: Arbitrum, Optimism, Base, zkSync, StarkNet, Scroll, and a dozen smaller ones. What I found is that the average daily data output per rollup is under 500 kilobytes. To put that in perspective, a single high-resolution JPEG image is often larger. Most rollups are handling a steady but modest volume of simple transfers, swaps, and bridge operations. The exceptions are a handful of gaming and high-frequency DeFi protocols, but they represent a tiny fraction of the total. Even during peak activity days, the aggregated data from all major rollups barely scratches the surface of Ethereum’s blob capacity. The network is rarely above 20% utilization. The idea that we need a whole new layer to handle this is like building a six-lane highway for a bicycle path. This disconnect between narrative and data is not accidental. It’s the product of a market that rewards narrative-building over technical necessity. Venture capital firms and infrastructure projects need a story that justifies multi-billion dollar valuations. “Data availability” sounds critical, technical, and inevitable. It fits into the modular blockchain thesis that has become the dominant intellectual framework of the post-merge era. But when you actually look at the numbers, you realize that the data bottleneck is a phantom. The real bottleneck is liquidity fragmentation and user experience issues that prevent rollups from reaching the scale required to fill those blobs. Weaving code into the fabric of physical reality. Let me be precise. The problem is not that DA layers are useless. For a small set of high-throughput applications that process thousands of transactions per second—think decentralized social networks, on-chain gaming with real-time state, or global settlement layers—custom DA can be justified. But for the vast majority of rollups that handle a few hundred transactions per second at peak, sticking with Ethereum’s blobspace is not only sufficient but also simpler and more secure. The marginal cost savings from moving to a dedicated DA layer are negligible when your baseline data output is low. Moreover, using a separate DA layer introduces trust assumptions: you now depend on a new set of validators or a data availability committee. That’s a step away from Ethereum’s security model, not toward it. The contrarian angle here is uncomfortable for the modular maximalists. If 99% of rollups don’t need dedicated DA, then the entire layer-2 scaling narrative needs to be recalibrated. The value proposition of rollups is not their ability to cheaply post data somewhere else; it’s their ability to offer fast finality, low fees, and composability to users. The DA layer is a backend infrastructure detail that should be invisible to end users and most developers. Yet the market has turned it into a primary feature. Projects are over-engineering their stacks to appear sophisticated, when what they really need is better user onboarding and liquidity incentives. Based on my audit experience, I’ve seen teams spend months integrating with a custom DA layer when they could have been building applications. The opportunity cost is real. Every hour spent on DA integration is an hour not spent on improving UX, building bridges, or attracting users. The modular narrative has created a supply-push dynamic where infrastructure is built before demand exists. That’s fine for research, but it creates noise for investors and builders trying to decide where to allocate resources. Let’s look at a concrete case. Consider Base, the Coinbase-incubated rollup. Base currently posts its data to Ethereum’s blobs. The cost of blob data for Base in the last quarter averaged about $2000 per day. That’s trivial for an operation backed by a multi-billion dollar company. Even if Base grew tenfold, the cost increase would still be manageable. Switching to a dedicated DA layer might save a few hundred dollars per day, but at the cost of adding a new dependency and potential liveness risk. Why would they do that? They wouldn’t, unless the narrative pressure forces them to signal that they are “modular”. The same logic applies to Arbitrum and Optimism, which have both publicly explored alternative DA but continue to settle on Ethereum for the vast majority of their traffic. The only counterexample that truly justifies dedicated DA is the extreme case of a protocol like Raydium on Solana—but Solana is not a rollup. In the rollup world, the data volumes simply haven’t materialized. The bull case for dedicated DA relies on a future where every on-chain interaction generates enormous amounts of data, perhaps with fully on-chain AI or virtual worlds. But that future is speculative and years away, if it arrives at all. We are currently in a sideways market where the priority should be sustainability and user value, not speculative infrastructure buildout. Weaving code into the fabric of physical reality. My suspicion is that the DA layer hype is a symptom of a deeper issue: the industry’s addiction to narrative-driven innovation cycles. Every two years, a new “scaling” meme emerges. First it was sharding, then L2, then rollups, and now DA layers. Each cycle promises to solve a fundamental problem, but the problem often shifts before the solution is proven. The real challenge has always been adoption, not throughput. We have enough scalability today to handle billions of users—if those users would just show up. The missing piece is not a new layer; it’s a compelling use case that drives demand from the bottom up. Where does this leave us? I believe the DA layer narrative will eventually deflate as more builders realize the numbers don’t add up. The next narrative shift will likely focus on interop and application-layer innovation, not base layer infrastructure. The projects that survive this market will be those that spent their resources on building real products, not on optimizing for a bottleneck that doesn’t exist. As for the DA layer protocols themselves, they will find their niche in high-throughput verticals like gaming or AI, but their total addressable market is far smaller than currently priced in. Finding the signal in the noise of 2020. The takeaway is simple. If you are building a rollup, resist the pressure to adopt a dedicated DA layer unless you have hard data showing that Ethereum’s blob space is insufficient. Most likely, it’s not. Focus your engineering capital on what matters: composability, user experience, and liquidity. The modular thesis is elegant, but elegance is not the same as necessity. The ghosts in the machine of trust are not the DA layer—they are the belief that infrastructure alone drives adoption. Real growth comes from applications that solve real problems. Listen for the quiet hum of the second layer, and you’ll hear the sound of users waiting for something worth their attention.