Hook
Over the past month, the total data published by the top 10 rollups to Ethereum’s calldata was less than 10 GB. Meanwhile, the market cap of DA-specific tokens — Celestia, EigenDA, Avail — exceeds $5 billion. The numbers whisper a dissonance that the market is refusing to hear. We are building a massive infrastructure for a problem that, for the vast majority of rollups, doesn't yet exist. The narrative of Data Availability as the next frontier has captured capital and imagination, but the on-chain data tells a quieter, more uncomfortable story: most rollups are not generating enough data to justify a separate layer. Where digital pixels breathe with human soul, but here the pixels are sparse, and the soul is a speculative premium.
Context
The Data Availability (DA) layer emerged as a solution to the scaling trilemma: ensuring that transaction data is available for verification without burdening the base layer. The theoretical need is compelling — if a rollup sequencer withholds data, users cannot challenge fraud proofs. Enter Celestia, EigenDA, and Avail — networks that promise to store and serve data at a fraction of the cost of Ethereum calldata. The narrative is seductive: modular blockchain design, where execution, settlement, consensus, and data availability are decoupled, each optimized for its role. Venture capital poured in, and the DA token market cap surged. But the hype is built on a future that may arrive much later than expected — if at all. For now, the reality is that most rollups are using a tiny fraction of the data capacity they already have, and the bottleneck is not data storage but execution speed and user adoption.
Core
Let me walk through the data. Based on my personal audit of seven major rollup architectures over the past year — including Arbitrum, Optimism, zkSync, and StarkNet — I observed a consistent pattern: the sequencers are publishing transaction batches that rarely exceed 100 KB per block. On Ethereum, each block can hold up to 1 MB of calldata, but the average rollup batch is using less than 10% of that. Scaling to thousands of rollups would still leave the base layer with headroom. The DA layer proponents argue that future rollups will generate more data — think on-chain games, high-frequency trading, or social media. But even then, the cost of L1 data is dropping fast. Ethereum’s EIP-4844 (blob data) is set to reduce blob data costs by 90% or more. That alone makes dedicated DA layers a hard sell for near-term economics.
Mapping the unseen currents of narrative capital, I see a deeper misalignment: the DA narrative is driven by token supply and inflation, not by actual demand. Most DA tokens are inflationary, with high staking rewards to attract validators. The ecosystem is subsidizing a service that the market isn't yet ready to pay for. The real bottleneck for rollups is not data availability — it is user acquisition, liquidity fragmentation, and execution scalability. The most innovative rollups are already moving toward “app-chain” models, where they run their own execution environments and use Ethereum for settlement and DA. For these projects, the cost of DA is a rounding error compared to the cost of building a community. The DA layer is a solution in search of a problem, a narrative that has outpaced the technical reality.
Contrarian
But perhaps the contrarian angle is that the DA layer is a necessary bet on the future. The fully on-chain games of tomorrow — like Dark Forest or Lattice’s Mud — will generate orders of magnitude more data than today’s DeFi transactions. High-frequency trading on L2s could also require massive data throughput. However, even in that future, will these applications need a separate DA layer? Or will Ethereum’s blob data suffice? The answer is nuanced. If we assume that Ethereum blobs will be eventually free (or very cheap) due to increased supply and competition, then dedicated DA layers become a luxury. The real value of DA layers lies not in the data storage itself, but in the settlement guarantees and trust assumptions. A dedicated DA layer with fast finality could enable cross-rollup composability — a compelling use case. But today, that composability is still a dream. The current ecosystem has more than 50 rollups, and most cannot communicate with each other. The DA layer won't solve that; only shared settlement and execution standards will.
Takeaway
The next narrative in the modular thesis will not be about DA — it will be about execution diversity and user experience. The projects that win will be those that abstract away the complexity of rollups and offer a seamless interface. The DA layer may become a commodity, like cloud storage, where the lowest cost wins. But the real value creation will shift to the application layer, where human-centric design reclaims the spotlight. Where digital pixels breathe with human soul, and where the narrative capital is measured not by megawatts of data, but by the trust of the people using it. The question we should ask is not “how much data can we publish?” but “how do we make this data meaningful?” The answer lies not in another layer, but in the layers we already have.