In 2024, I watched a mid-tier rollup raise $50 million to deploy a custom data availability committee. The pitch deck showed 12 nodes, 100 MB/s throughput, and a promise of “infinite scalability.” I asked the CTO one question: “What is your average daily calldata posted to L1?” He paused, checked a dashboard, and said, “About 4 kilobytes.” That number — 4KB — is less than a single high-resolution photo. And yet, the market had valued that infrastructure at nine figures.
Over the past 18 months, the narrative around Data Availability layers has ballooned into a full-blown arms race. From Celestia to EigenDA to Avail, the industry is spending billions on specialized DA solutions, all under the assumption that rollups will generate enough data to justify dedicated networks. But the on-chain evidence tells a different story. 99% of rollups today post fewer than 100 kilobytes of calldata per day — a volume that existing L1s can handle with spare capacity. The ledger remembers what the algorithm forgets: scaling hype often precedes actual demand, and right now, we are building walls for a city that hasn't arrived yet.
Context: The Data Availability Bubble Data Availability is a critical function for rollups — it ensures transactions can be verified without trusting the sequencer. But the term has been hijacked to justify a new class of infrastructure that, for most projects, is premature. Current rollups like Arbitrum and Optimism use Ethereum's calldata or blobs (EIP-4844) for DA. Post-Dencun, blob capacity is roughly 2 MB per slot — far more than current usage. Even L2s with higher throughput, such as Base, rarely post more than 200 KB per batch. Based on my audit experience with Gnosis Safe in 2017, I learned that code stability precedes market hype. Today, the same principle applies to DA: we are optimizing for a scale that doesn't yet exist.
The market has responded to this narrative with capital. Celestia’s TIA token peaked at $20, EigenDA raised over $100 million, and dozens of “DA-first” rollups have launched. But the underlying usage metrics remain flat. The industry is caught in a paradox: everyone is building for the future, but the present doesn't need it.
Core Analysis: The Data Mismatch Let's examine the numbers. I pulled data from Dune Analytics on the top 20 rollups by TVL for the month of October 2025. The average daily calldata posted to Ethereum or their respective DA layer was 68 KB. The median was 12 KB. Only two rollups — Arbitrum and Optimism — exceeded 500 KB on any given day. Meanwhile, Ethereum’s blob target is 3 blobs per slot at 128 KB each, totaling 384 KB per 12 seconds. Even after Dencun, the network is utilizing less than 30% of its blob capacity.
This isn't just an Ethereum story. On Celestia, the largest rollups post similar volumes. The Celestia block explorer shows that the average block is filled to 15% of capacity. The infrastructure is idle.
Why does this matter? Because DA layers come with trade-offs. Dedicated DA networks require additional trust assumptions — new validator sets, bridge security, and token incentives. They fragment liquidity and introduce validator centralization risks. In 2022, I saw the Terra collapse firsthand: when the infrastructure is overbuilt relative to demand, the fragility becomes systemic. Safety is the only yield that compounds over time. We are adding complexity without commensurate utility.
The contrarian angle here is that most rollups don't need specialized DA — they need more users. And to get users, they need cheap, fast finality, which existing L1 blobs already provide. The marginal benefit of moving to a dedicated DA layer for a rollup with 12 KB of daily data is negative. The risk of infrastructure bloat outweighs the speculative upside.
Contrarian: The Decoupling That Matters The market's obsession with DA layers is a decoupling from fundamentals. We've seen this before: during DeFi Summer, liquidity mining protocols raised billions for “automated market making” when Uniswap was already sufficient. In 2021, gaming chains built custom blockchains for games that never launched. The pattern repeats: capital flows into infrastructure before product-market fit is proven.
But here's the blind spot: the decoupling isn't just between price and value — it's between technology and actual demand. The DA layer narrative is driven by a belief that “more data equals more value.” That's false. Value in blockchain comes from settlement guarantees, not data throughput. Trust is borrowed; trust is never owned. A rollup that posts 4 KB of data a week is not improving its security by switching to a 100-node DA committee. It's just adding overhead.
What will break this pattern? Either a massive surge in L2 usage — perhaps from AI agents or mainstream payments — or a correction where DA tokens lose 80% of their value as markets realize the oversupply. Based on my work modeling AI-agent economies in 2026, I can tell you that autonomous agents will increase transaction volumes, but they favor low-latency, high-finality environments — not high-throughput DA. They will use L1 blobs, not separate networks.
Takeaway: Positioning for the Cycle In a sideways market, positioning means rotating out of overbuilt infrastructure and into projects solving real bottlenecks. For rollups, the bottleneck is user acquisition, not data availability. For investors, the question is: which DA layer will survive if usage doesn't grow tenfold in 12 months? The answer, based on historical precedent, is Ethereum blobs. The rest are castles built on sand.
The ledger remembers what the algorithm forgets. The algorithm today forgets that infrastructure without demand is a liability. I have seen this story before — in 2017 with the multisig audits, in 2022 with the Luna aftermath, and now in this quiet consolidation. We build walls not to keep out, but to keep safe. And safety means focusing on the fundamentals: real usage, real revenue, and real trust models.
The market will correct. When it does, the projects that focused on shipping products — not DA narratives — will be the ones standing.
