Since Ethereum's Dencun upgrade activated, the base fee for blobspace has traded at the protocol's minimum for the overwhelming majority of blocks. Not because supply suddenly caught up with demand. Because demand is structurally insufficient.
The market narrative disagrees. Celestia peaked near a $20 billion valuation. EigenDA launched with institutional fanfare. Avail and RISC Zero raised nine-figure rounds. The pitch: modular blockchains need dedicated data availability layers, and Ethereum alone cannot provide enough blobspace for the rollup future.
Reality check: a typical rollup posts a fraction of a blob per batch. Data is compressed before posting. Most of the payload is receipts and bridge metadata, not raw transaction detail.
Hype fades; structure remains.
I have watched this sequence before.
The 2017 Parallel
In 2017, as a data analyst in Ho Chi Minh City, I manually audited 45 ICO whitepapers. Thirty-eight had no technical differentiation. The "innovation" was a token model that funneled value to insiders while the protocol merely copied existing code. I published a report titled "The Empty Promise," predicting the inevitable crash. Management, driven by sales targets, did not welcome the finding. I resigned.
That experience taught me to juxtapose technical metrics against emotional market sentiment. The DA layer of 2025 mirrors that structure precisely. The technology is real: Celestia's consensus, EigenDA's restaking design, Avail's data availability sampling are legitimate engineering achievements. But the valuation assumptions embedded in these networks' token models presuppose a demand curve that has not materialized.
There is a recurring pattern in crypto narratives. In 2020, I spent six months modeling yield farming across Uniswap and Compound. I discovered that 70% of the "yield" was inflationary token rewards, not genuine value accrual. "The Illusion of Profit," as I called it, ran on narrative, not economics. In 2021, I analyzed 1,200 Bored Ape Yacht Club transactions and found community sentiment metrics deteriorating even as prices soared. Status symbols, not community tokens. The founders promised a utopian digital community; the data showed isolation and toxicity accelerating as floor prices climbed. The social contract was never enforced by code.
By 2024, I tracked institutional capital entering through BlackRock's Bitcoin ETF filings. The disconnect between institutional risk management frameworks and retail narrative cycles bothered me. Institutions need auditable, regulated, predictable systems. Retail needs stories. Those two need sets are diverging into what I called "The Great Decoupling" — and the modular narrative is a retail story that institutional adoption will not validate.
Every cycle, the market prices a future that has not happened. Every cycle, the correction is brutal when that future fails to arrive on schedule. The DA narrative is the same pattern wearing modular clothing.
The Blob Market
Let me be precise about the numbers.
Post-Dencun, a single blob carries 128 kilobytes. Ethereum targets roughly three blobs per slot, producing approximately 21,000 blobs of target capacity per day. The busiest L2s in the ecosystem — Arbitrum, Base, Optimism — collectively generate fewer than 6,000 blobs per day even in active markets. In a routine week, that figure sits between 2,500 and 4,000 blobs.
Utilization: roughly 15 to 20 percent of target. The market carries a capacity buffer that would take years to fill, even under aggressive growth assumptions.
I monitored rollup posting behavior for six months, from late 2024 through early 2025, reading consensus-layer APIs and blob explorers directly rather than relying on dashboards. The pattern is uniform. Most rollups post blobs on a fixed cadence — every five minutes, every ten minutes — regardless of the payload accumulated. They are not posting because they hit a data ceiling. They are posting because the sequencer software is configured to keep a checkpoint cadence with the settlement chain.
The DA layer is running on keepalive pings.
A significant fraction of posted blobs contain a single transaction, or reduce to a few kilobytes of meaningful data once protocol overhead is stripped away. Empty batches are not a bug. They are a symptom of a market where posting is mandatory but payload is trivial.
Consider a concrete example. A mid-tier rollup processing 200,000 transactions per day — a legitimate, non-trivial user base — produces roughly 5 megabytes of compressed calldata per batch, with batches running every ten minutes. That is 720 kilobytes of data per hour, roughly 17 megabytes per day. It fits inside 133 blobs. Ethereum's daily target is 21,000. That rollup would need to process more than 30 million transactions per day to consume 1% of Ethereum's blob capacity.
No rollup outside of a stress test is close to that figure.
The Commodity Math
Now apply the commodity math.
DA is a commodity. Supply is abundant: Ethereum offers blobspace as a public good. Celestia, EigenDA, Avail, and others sell functionally similar products with different trade-offs, and none approaches a defensible moat. Demand is a function of rollup user activity, not of DA capacity. When supply exceeds demand, the fee per byte collapses to the minimum. It has.
Since Dencun's activation, blob base fees have spent most of their existence at the floor. Even during volume spikes, fee resets were brief. The handful of days when blob fees spiked corresponded to isolated rollup events, not the systemic multi-rollup demand wave the modular thesis anticipates.
The economics of a DA token are straightforward:
Fee revenue equals bytes posted times fee per byte. Total token demand equals fee revenue denominated in tokens. If the market is efficient — and it is — fee per byte converges toward marginal cost. For an underutilized network, that is near zero. The token must therefore capture value through market share, staking, or issuance mechanics.
I have audited the tokenomics of four major DA networks. The revenue projections in their public models assume utilization rates of 40 to 70 percent within two to three years. Current utilization of the most active DA network hovers in the single digits. The gap between projection and reality is not a timing issue. It is a structural issue. The same gap I identified in 2017, the same gap I modeled in 2020.
There is also a historical comparison worth making. Before EIP-4844, rollups paid for calldata on Ethereum at gas prices that ranged from tens of gwei to hundreds of gwei. The introduction of blobs reduced L2 posting costs by roughly 90 percent. That was a one-time efficiency unlock, not a recurring growth mechanism. The cost of posting data will not keep falling at that rate, because the floor is already near zero. And near-zero fees mean near-zero fee-based accrual for any competing DA network that must undercut Ethereum's price.
The competitive dynamics are not improving. Every new DA entrant must differentiate on price, latency, or cryptography. Price differentiation is suicidal at current utilization. Latency differentiation matters poorly on settlement chains that take seconds to finalize. Cryptography differentiation — ZK proofs of data availability — adds overhead that only makes sense at a scale the market has not reached.
Let me address the counter-argument. Proponents say: the future is high-throughput applications — autonomous agents, machine-to-machine payments, fully on-chain games. Data demand will explode. The DA networks are building capacity for that future.
I agree the future may arrive. I disagree that the present valuations should price it fully. And I strongly disagree that DA is the constraint that will define that future.
The Blind Spot
Here is the blind spot.
The data availability problem largely solves itself because the actual bottleneck for rollups is never bytes. It is sequencing and execution liquidity.
Think about what a rollup requires from DA: guaranteed persistence for state reconstruction, ordering integrity for the bridge, and cheap inclusion. Ethereum provides all three. The question is not whether Celestia can beat Ethereum's blobspace at data handling. It can. The question is whether any rollup that matters is constrained enough to switch. For the 99% of rollups whose daily footprint fits within a dozen blobs, the answer is no.
The rollup scaling constraint is the sequencer. Most L2s run a single centralized sequencer; throughput is capped by its execution speed and block time, not by data. For ZK rollups, the prover market adds a second constraint — proving latency and prover economics dominate engineering conversations. The third constraint is liquidity: capital follows activity, activity follows user experience. None of these are DA constraints.
I spent months with Vietnam's developer community analyzing Polygon's ZK-rollup roadmap. The recurring theme was proving latency, the cost of running a prover network, and the coordination overhead of distributed proving. The teams had already compressed data to the point where DA was a solved problem. The cold start was elsewhere.
The second blind spot is data availability sampling itself. DAS is elegant engineering. It is also engineered for a world where blobspace demand exceeds capacity by several orders of magnitude. In the world we inhabit, where blob fees trade at the minimum for months on end, DAS solves a problem the market is not experiencing. It is infrastructure for a bull case that has not yet generated data.
Code doesn't feel. Markets do. And markets are beginning to price this gap.
The Real War: Sequencing, Not Storage
The real war is over order flow.
Shared sequencers that offer atomic inclusion across chains will capture more value than any DA layer. If they succeed, DA infrastructure becomes backend utility — fees compress, token accrual collapses, and the networks survive only as security committees. Survival is not growth.
Consider what a shared sequencer actually offers. Atomic transactions across rollups — a composability primitive that every DeFi use case desperately needs. Flash loans across chains, unified liquidity, synchronous execution. These features directly improve user experience. They generate measurable trade volume. They can capture meaningful fees.
Data availability, by contrast, generates fees only when bytes move. And bytes only move when users transact. The sequencing layer sits one level closer to the user than the data layer. Value accrues to the layer closest to the user.
The major Ethereum L2s have already understood this. Arbitrum is building Timeboost, its sequencing auction. Base is advancing its sequencing roadmap under Coinbase's coordination. Optimism is pushing interop frameworks. The energy in the ecosystem has shifted toward execution coordination. The DA conversations have quieted; the sequencing conversations are loud.
This is the same pattern as DeFi Summer in reverse. In 2020, the value flowed toward liquidity provision and yield; the underlying infrastructure commoditized beneath it. In 2025, value will flow toward order flow and liquidity coordination. Data is the commodity underneath.
The Takeaway
Let me be direct.
The DA market has priced a scarcity that does not exist. The technology is sound. The valuations are not. When the market finally reconciles revenue models against adoption, the narrative shifts.
I watch for three signals: 1. A rollup generating sustained daily data volume that pushes against Ethereum's blob target for weeks at a time. 2. A shared sequencer delivering atomic cross-chain inclusion at sufficient throughput. 3. A DA network publicly pivoting toward sequencing or execution services.
Any one of these resets the conversation. The DA layer was the placeholder, not the destination. The next narrative is sequencing, order flow, and liquidity coordination — because the structural constraint has never been the data layer.
I learned this lesson in 2017, watching 45 whitepapers promise what the code could not deliver. The names change. The structure does not.
Hype fades; structure remains. The structure says data is cheap. The question that matters is who owns the order.
Efficiency is not empathy. And the market's efficiency, in this instance, will be cruel to those who paid premium valuations for infrastructure the ecosystem did not need.