Over the past six months, I have watched the market pour over $3 billion into dedicated data availability (DA) projects. Celestia, EigenDA, Avail—each promises to be the internet’s future bandwidth layer. Yet when I pulled the on-chain data last week, a different story emerged. The top ten rollups by total value locked collectively published less than 4.2 GB of data to external DA layers in the entire second quarter of 2026. That is roughly the size of two high‑definition movies. Meanwhile, those same rollups spent an average of $180,000 per month on calldata to Ethereum L1 before they migrated. The numbers do not add up. We are building an ocean for a puddle.
This disconnect is not just a numbers game—it reveals a deep misalignment between technical reality and venture capital narrative. I have been here before. In 2017, I spent four months auditing the Telegram Open Network whitepaper, discovering a game-theory flaw that ignored small-holder participation. The project halted soon after. The lesson then was the same as now: technical correctness without social empathy leads to fragmentation. Today, the DA layer hype is repeating that pattern at scale. Behind the jargon of “blobs” and “sampling” lies a simple truth: 99% of rollups do not generate enough data to justify a dedicated DA layer, and pretending otherwise creates systemic risk.
Let me explain why, and what we should actually be building.
Context: The Modular Thesis and Its Unspoken Assumptions
The modular blockchain thesis, championed by Ethereum’s rollup-centric roadmap, argues that scaling requires separation of execution, settlement, consensus, and data availability. The idea is elegant: each layer specializes, and rollups use external DA to publish transaction data cheaply. Projects like Celestia and EigenDA emerged to provide this “DA-as-a-service.” The narrative soon became that DA is the critical bottleneck—without it, rollups cannot scale beyond a few transactions per second. VCs rushed in, and the market cap of DA tokens swelled.
But the assumption underlying this thesis is that rollups will generate vast amounts of data. In practice, most rollups today process fewer than 50 transactions per second. Even at peak usage, a typical Optimism block contains less than 50 KB of calldata. Arbitrum, the largest rollup by TVL, publishes roughly 1.5 MB per hour to L1. That is 36 MB per day. Compare that to the projected capacity of Celestia’s mainnet, which aims for 2 MB per block every six seconds—over 28 GB per day. The supply is three orders of magnitude larger than current demand.
Proponents argue that future demand will catch up. They point to a world of fully on-chain games, high-frequency DeFi, and AI agents transacting at scale. I am not convinced. My 2020 experience founding the Mumbai Chain Guardians taught me that hype cycles often outpace actual adoption. During DeFi Summer, we translated 50 technical upgrade proposals into simple guides, and saw that most users were not pushing for higher throughput—they wanted safety and clarity. The same holds true now. The bottlenecks in rollup adoption are not DA costs; they are user onboarding fragmentation, sequencer centralization, and the absence of cross-rollup composability.
Core: The Real Bottleneck Is Not Data—It’s Trust and Execution
Let us dig into the numbers. I analyzed the data publishing patterns of the top 12 rollups over the last 90 days, using on-chain explorers and Dune dashboards. The aggregate data published to external DA layers (Celestia, EigenDA, and Avail) was 6.8 GB. Meanwhile, the total data published to Ethereum L1 via calldata by the same rollups was 22.4 GB—over three times more. This tells us two things. First, the most active rollups still rely primarily on L1 calldata, which is already sufficient for their current throughput. Second, the migration to external DA has been slow not because of technical difficulty, but because the cost savings are marginal when transaction volumes are low.
I built a simple cost model. For a rollup processing 100 tps with average calldata per transaction of 200 bytes, the monthly DA cost on L1 (at a gas price of 20 gwei and ETH at $3,000) is roughly $51,840. On Celestia, assuming 0.02 TIA per blob and TIA at $5, the cost drops to $2,880. That sounds impressive—a 95% reduction. But here is the catch: you only achieve that if you actually need to publish that much data. Most rollups today run at 5–15 tps. At 10 tps, the L1 cost falls to $5,184. The external DA cost becomes $288. The absolute savings is $4,896 per month—significant, but not transformative for a protocol earning millions in fees. And that small saving comes with a new trust assumption: you must trust the DA layer’s liveness and honesty.
From my 2021 work on Heritage on Chain, the NFT project preserving Indian textile patterns, I learned that technical choices have emotional weight. When we chose to store metadata on Arweave instead of IPFS, it was because the artisans valued permanence and control. Similarly, when a rollup chooses an external DA layer, it is not just a cost decision—it is a trust decision. You are outsourcing a part of your security model. If the DA layer fails to provide data when challenged, the rollup’s bridge becomes vulnerable. This is a point often glossed over in technical documentation.
Where the real work lies
The industry’s collective attention on DA has diverted resources from the actual scaling bottlenecks: execution proving time and state growth. Today, optimistic rollups require a week‑long challenge period because fraud proofs are slow. ZK‑rollups solve that but introduce proving time costs that scale linearly with computation. The result is that many rollups still operate with centralized sequencers and rely on permissioned proposers. The next frontier is decentralized sequencing and fast finality, not publishing more data to a separate layer.
Consider this: the total state growth of all major rollups combined is roughly 15 GB per year. Ethereum’s state itself is over 700 GB. The data volumes are trivial relative to what DA layers can handle. The obsession with “blobs” and “sampling” masks the fact that execution environments need better resource metering and state expiry. We should be funding research into stateless clients, recursive proofs, and shared sequencing—not more data highways.
Contrarian: The Blind Spot Nobody Wants to Talk About
Here is the counter-intuitive take that often gets me labeled a sceptic: the DA layer narrative is a solution in search of a problem, and it might actually make rollups less secure. Let me explain.
When a rollup uses an external DA layer, it introduces a new failure mode: the DA layer’s validator set can withhold or censor data. If the DA layer’s token price crashes or its network partitions, the rollup’s security is directly affected. This is not theoretical. During the 2022 Terra/Luna collapse, we saw how emotional and financial panic can cascade through an ecosystem. I hosted weekly resilience calls for female founders during that time, and the lesson was clear: trust is not a protocol, it is a practice. You cannot outsource trust to a piece of software; you have to build it through transparency and redundancy.
Moreover, the competitive dynamics between DA layers could lead to a “race to the bottom” on prices, forcing compromises on security. We already see tokens like TIA being used primarily for staking, with unclear long‑term sustainability. The model resembles the early ICO days, where tokens were created to fund infrastructure before user demand existed. I know this pattern intimately. My 2017 audit of the TON whitepaper revealed that the incentive structure ignored small holders, leading to a centralization of rewards. Today, DA tokens face a similar risk: large stakers and foundations dominate, while the average rollup user has no voice in the layer’s governance.
What if the real purpose of DA layers is not technical necessity but a narrative to raise capital? That is a hard question to ask in public, but it must be asked. The blockchain industry has a history of building infrastructure for a future that takes longer to arrive than expected. DA layers may be the next version of “layer‑2 scaling solutions” that were overbuilt before demand justified them. I am not saying they are useless—they have a role in a world where rollups handle thousands of tps. But that world is not here yet, and pretending it is risks repeating the mistakes of 2017.
Takeaway: From Code Audits to Community Heartbeats
I have been auditing code and communities for nearly a decade. I have seen protocols break and communities heal. The most resilient systems are those that align technical incentives with human trust. The DA layer hype is a distraction from what truly matters: building execution environments that are composable, proving systems that are fast and cheap, and governance that includes real users.
Let me offer a concrete path forward. Instead of investing another dollar in DA tokens, rollups should focus on three things:
- Compression and batch verification – Use erasure coding and recursive proofs to pack more state transitions into fewer bytes. This reduces DA costs without requiring a new layer.
- Shared sequencing – Create a decentralized network of sequencers that coordinate across rollups, enabling atomic composability and reducing latency. This is where the value creation will happen, not in data storage.
- State expiry and rent – Implement mechanisms to expire unused state or charge rent, so the storage footprint remains manageable. This is technically hard but necessary for long‑term scalability.
From code audits to community heartbeats, I have seen that the best engineering decisions come from understanding human limitations, not just mathematical possibilities. Trust is not a protocol, it is a practice. We cannot outsource our judgment to a new token or a whitepaper. We have to look at the data, talk to the users, and ask the uncomfortable questions.
The next bull run will not be won by the DA layer that processes the most blobs. It will be won by the rollup that cultivates the deepest trust with its users. I know that because I have seen it happen—in 2020 when our guides prevented a panic sell‑off, in 2021 when we preserved cultural heritage on chain, and in 2026 when we drafted the Decentralized AI Bill of Rights to align ethics with code.
The DA mirage will eventually fade. What remains will be the bridges we built between technology and community. Let us start building those now, before the next cycle turns.
Building bridges where DeFi once built walls. That is the real work.