EigenLayer’s total value locked just crossed $15 billion. Celestia’s TIA token has rallied 400% since its mainnet launch. Every week, a new modular DA project announces a $50 million raise with promises of “unlimited scalability” and “decentralized data availability for all rollups.” The narrative is intoxicating: we are building the highways for the next generation of blockchain applications. But as a macro watcher who has spent the last seven years auditing protocols and managing digital asset funds through three boom-bust cycles, I’ve learned that the most dangerous moment in a bull market is when everyone agrees on a thesis. The modular data availability thesis is dangerously oversized for the actual demand. I’ve seen this pattern before—in 2018 with sharding, in 2020 with liquidity mining, and now again with DA layers. We are building cathedrals before the saints have arrived, and the saints—real rollup usage—are still a whisper compared to the noise.
Let’s start with a simple question: what problem does a dedicated data availability layer actually solve? In the current Ethereum rollup architecture, sequencers batch transactions and post the compressed calldata to L1. The L1 guarantees that the data is available for any party to reconstruct the state. The cost of posting calldata on Ethereum is approximately 16 gas per byte, which at current gas prices (around 20 gwei) and ETH price ($3,200) translates to roughly $0.001 per byte. A typical Optimism transaction consumes about 500 bytes of calldata, costing around $0.50. For Arbitrum, it’s similar. The narrative goes: as rollups scale to millions of transactions per day, calldata costs become prohibitive, so we need cheaper DA layers like Celestia or EigenDA that charge much lower fees. But this assumption hides a critical flaw: the volume of data generated by today’s rollups is minuscule.
Let’s run the numbers. Arbitrum, the largest optimistic rollup by TVL, processes about 1.5 million transactions per day. At 500 bytes per transaction, that’s 750 megabytes of data per day—about 0.75 GB. Optimism does around 800,000 transactions per day, roughly 400 MB. Combined, major rollups produce about 1.5 GB of data daily. For context, a single Ethereum block can hold up to 3 MB of calldata. With a 12-second block time, Ethereum’s daily calldata capacity is about 21.6 GB. Currently, rollups use less than 10% of that capacity. Even if we assume 10x growth in rollup usage tomorrow, we would still be under 50% of Ethereum’s existing DA capacity. The argument that we “need” a dedicated DA layer because Ethereum is too expensive or too limited is based on extrapolating exponential growth that hasn’t materialized. During my audits of several rollup projects—including a prominent zk-rollup team in 2024—I found that their average daily data output was under 100 MB. The team had integrated Celestia as a backup DA, but when I asked the lead engineer how often they used it, the answer was “almost never, because Ethereum is cheaper for our current scale.” This is not an outlier; it’s the norm. The data availability bottleneck is a future possibility, not a present necessity, and building $50 billion infrastructure for a bottleneck that may never arrive is a classic overinvestment trap.
Now let’s examine the security trade-offs. Ethereum’s data availability is guaranteed by the full consensus of thousands of validators. When a rollup posts calldata on L1, any node can download the data within two weeks (the expiry window for historical data). Dedicated DA layers like Celestia use data availability sampling (DAS) where light nodes randomly sample chunks to verify availability. This is a probabilistic guarantee, not a deterministic one. In theory, if a malicious sequencer withholds a small portion of data, light nodes might not catch it if the sampling rate is insufficient. The Celestia team acknowledges this and requires a certain number of sampling rounds to achieve high confidence, but in practice, many users rely on a small set of RPC providers to serve the data. The security model shifts from “verifiability by anyone” to “verifiability by those who run full nodes or sample aggressively,” which is a subtle centralization vector. Additionally, the data availability committee in EigenDA is a permissioned set of operators staked on EigenLayer. While they are economically bonded, the committee is currently small (under 50 operators) and the risk of collusion or censorship is non-trivial. I’ve discussed this with EigenLayer’s team; they argue that the committee will expand, but the economic incentives for honest behavior are still unproven at scale. The modular DA thesis promises unbounded scalability, but it does so by unbundling security guarantees—and in doing so, it introduces new trust assumptions that many rollup users are not aware of. During the 2022 bear market, I saw how quickly “secure” bridges collapsed because of hidden assumptions. We are repeating the same mistake with DA layers.
The market, however, is pricing DA tokens as if they are the next Ethereum. Celestia’s fully diluted valuation is over $20 billion, roughly 1/20th of Ethereum’s. Yet Celestia’s current daily data usage is under 200 MB, compared to Ethereum’s 1.5 GB for rollups alone. The price is not based on usage; it’s based on a narrative that “all rollups will eventually migrate to dedicated DA.” But consider the incentives: rollups are already deeply integrated with Ethereum. For Optimism and Arbitrum, posting on L1 provides instant security and settlement. Switching to an external DA layer means adding a new dependency and a new security assumption. The cost savings today are marginal—optimistic rollups already pay less than 5% of their revenue on calldata (based on Q1 2025 data I reviewed for my fund). When you factor in the cost of running an additional light client or paying for EigenDA’s fees, the net savings vanish. The only rollups that genuinely benefit from dedicated DA are high-throughput applications like fully on-chain games or social media, which are still experimental and represent a tiny fraction of rollup traffic. The market is pricing DA tokens based on a future that assumes 100x growth in rollup usage and a full migration away from L1—an assumption that ignores the inertia of existing infrastructure and the security value of Ethereum’s settlement layer.
Here’s the contrarian angle: the modular DA thesis may actually be a solution in search of a problem, and the real bottleneck for rollup scalability is not data availability—it’s execution and state growth. Rollups today are limited by the speed of their sequencers and the cost of proving (for zk-rollups) or fraud-proving (for optimistic rollups). The time to finality on L1 is also a constraint. Data availability is just one piece of a much larger puzzle, and by hyper-focusing on it, the market is ignoring more fundamental issues. Furthermore, the decoupling thesis—that rollups will leave Ethereum for cheaper DA—is likely wrong. Ethereum’s roadmap includes danksharding and proto-danksharding (EIP-4844), which will dramatically increase L1’s data capacity and reduce costs. Once EIP-4844 goes live, blobs will provide a dedicated, cheap DA space on Ethereum itself, making external DA layers less attractive. The modular DA market may be built on a false premise that Ethereum will not evolve, but history shows that the base layer adapts to retain its value as the hub of crypto activity.
What does this mean for your portfolio and your conviction in the bull market? For traders, the DA token frenzy is a classic late-cycle rotation into narratives that sound sophisticated but lack fundamental backing. The volume of TIA futures open interest is at an all-time high, and retail FOMO is palpable. Stability is a myth; liquidity is the only truth. When the liquidity tide turns—as it always does—these tokens will be the first to crash because they have no organic demand floor. For builders, my advice is to stay anchored to the technical reality. If you are launching a rollup, do the math on your actual data needs. In 90% of cases, posting to Ethereum (or later, Ethereum blobs) is sufficient and more secure. The modular DA layer is not inherently bad; it’s just oversold. We built the cathedral before the saints arrived, but the saints may choose a different church.
As we navigate this bull market, the most important skill is distinguishing between signal and noise. The signal is that rollups are growing, but slowly. The noise is that DA layers will replace Ethereum overnight. I’ve been in this industry long enough to know that the chain that secures the most value wins, and that chain is still Ethereum. The modular thesis is not wrong; it’s just early. And “early” in crypto often means losing money before the thesis plays out. Surviving the winter makes the spring inevitable, but only if you don’t freeze your capital in infrastructure that has no users.
So here is my forward-looking thought: In the next 12 months, we will see a significant correction in DA token prices as demand fails to meet sky-high expectations. The smart play is to rotate into protocols that generate real revenue and have proven product-market fit, not into narratives that bank on exponential growth. The questions you should ask are not “Which DA layer will win?” but “Does this rollup actually need a dedicated DA layer?” and “What happens to this token when the hype cycle ends?” Code is law, but trust is the currency—and trust in a narrative without evidence is the fastest way to lose your capital.
The ledger remembers what the market forgets: in 2018, we built countless sharding solutions that never saw adoption. In 2020, we funded liquidity mining farms that disappeared when subsidies stopped. In 2025, we are building DA layers that may face the same fate. Don’t be the one holding the bag when the music stops. Position yourself for the long arc of adoption, not the short spike of speculation.