The Blob Saturation Thesis: Why Post-Dencun Rollup Fees Will Double by 2026
From the ashes of 2017 to the fluidity of DeFi, I have learned that every narrative that promises infinite scalability eventually hits a wall of reality. The Dencun upgrade, which went live in March 2024, was supposed to be Ethereum’s salvation — a magical patch that would make Layer 2 fees vanish into the ether. For a few months, it worked. I watched Arbitrum and Optimism fees drop to sub-cent levels, and the crypto Twitterati declared that the “scaling trilemma” was finally solved. But as someone who has been tracking on-chain data since the ICO bubble, I sensed a ticking clock. The blobs — those temporary data storage units that Dencun introduced — are not an infinite resource. They are a finite buffer, and the market is consuming them faster than anyone expected.
I remember sitting in a Berlin coffee shop in early 2024, reviewing the EIP-4844 specs with a former colleague from my PhD days. We ran the numbers: 6 blobs per slot, 12 seconds per slot, 32 slots per epoch. That gave us roughly 43,200 blobs per day. At first, that seemed like a lot. But then we looked at the growth curves of the top rollups. Arbitrum was processing 1.5 million transactions per day. Optimism was at 1.2 million. Base, the rising star, was already doing 800,000. Each transaction on a rollup doesn’t always require a full blob — but the average blob utilization was climbing. By May 2024, blob usage had hit 35% of capacity. By September, it was 60%. And in December, I saw a day where the network hit 92% blob utilization. The panic hadn’t started yet, but the data was screaming.
This is the context that most retail investors miss. The Dencun upgrade was not a permanent fix; it was a band-aid designed to buy time for L2s to mature. But the L2s have not matured in the way the Ethereum Foundation hoped. Instead of optimizing their data compression, they have been burning through blob space like it’s free. And it is free — for now. But the protocol has a hard cap. Once blobs are saturated, the market will bid for space, and fees will spike. My analysis of the current trajectory suggests that full saturation will occur within 18 to 24 months post-Dencun — that is, by early 2026. And when that happens, all rollup gas fees will double, at minimum.
Let me take you through the core mechanism. The blob market is a separate fee market from the execution gas market. Each blob has a base fee that adjusts dynamically based on demand. The mechanism is similar to EIP-1559 but with a target of 3 blobs per slot. If the network uses more than 3 blobs, the base fee increases exponentially. Currently, the average blob count per slot is around 4.5, which means we are already in the “overload” zone. The base fee for blobs has gone from 1 wei to 50 gwei in the last six months. That doesn’t sound like much, but it compounds. If a rollup is posting 10 blobs per minute, the cost per blob rises, and the rollup must either pass that cost to users or eat it. Most rollups are not profitable; they are subsidized by venture capital. The moment blob fees become significant, the L2s will have to raise their transaction fees. And that’s when the narrative will shift from “Ethereum is cheap again” to “L2 fees are a hidden tax.”
I have been tracking the blob usage of the top five rollups — Arbitrum, Optimism, Base, zkSync, and StarkNet — since the Dencun launch. I built a small dashboard using Dune Analytics and my own scripts. The data is sobering. Arbitrum alone consumes about 35% of all blob capacity. Base is at 22%. The other three combined take another 30%. That leaves only 13% for the remaining 50+ rollups. This concentration is dangerous. If one of the big rollups decides to increase its block frequency, it could trigger a cascade of fee spikes across the entire ecosystem. The academic view vs. the chain view: the academic view says that rollups can always fall back to calldata if blobs become too expensive. But that’s a false equivalence. Calldata is permanent and expensive — it costs 16 gas per byte versus 1 gas per byte for blob data. If a rollup switches to calldata, its fees increase by a factor of 10 or more. No one wants that. The chain view, which I have been observing, shows that rollups are deeply addicted to cheap blob space. They will not give it up easily.
Now, let’s talk about the contrarian angle. There is a growing narrative that the blobs will not saturate because L2s will eventually move to their own data availability layers, such as Celestia or Avail. This is a legitimate technological possibility. But it comes with a heavy cost: centralization. If a rollup uses a third-party DA layer, it is no longer a true Ethereum rollup. It becomes a “validium” or a “sovereign rollup” that trusts a separate consensus. The Ethereum Foundation has been clear that only Ethereum-native DA guarantees security. And the market has already priced this in. Look at the total value secured (TVS) of rollups that use Ethereum DA versus those that use external DA. The former holds $30 billion; the latter holds less than $500 million. Users trust Ethereum’s security. If rollups leave the blob market, they lose that trust. So they are stuck. They cannot leave, and they cannot stay if fees become too high. This is the classic prisoner’s dilemma of the L2 space.
I have spent years analyzing narrative cycles, and I see a pattern here. In 2020, the narrative was “DeFi summer.” In 2021, it was “NFTs are the new art.” In 2022, it was “real yield.” In 2023, it was “L2s will scale Ethereum.” And now, in 2025, the narrative is about to shift again. The next narrative will be “the blob fee crisis.” It will start with a few high-profile incidents. A rollup will have a sudden spike in activity — perhaps a memecoin launch or a gaming boom — and its blob usage will go through the roof. The base fee will skyrocket, and users will see their transaction costs increase from $0.01 to $0.50 overnight. The community will panic. There will be proposals to increase the blob target, but that requires a hard fork, which takes months. And even if the target is increased, it only delays the problem. The fundamental issue is that the demand for cheap data is growing faster than the supply.
Let me share a personal experience from my time auditing smart contracts in 2022. I was working with a Layer 2 project that claimed to be “Ethereum-equivalent.” I asked them how they would handle a blob fee spike. Their answer: “We’ll just use calldata.” That answer was a red flag. They had no economic model for fee sustainability. Most L2 teams today still have no economic model. They rely on the assumption that blob space will remain cheap forever. That assumption is about to break. I have been warning about this in my private newsletters since August 2024. The response has been a mix of curiosity and denial. Some developers tell me I am too pessimistic. Others admit they are worried but have no alternative. The academic view vs. the chain view: the academic view says that the market will naturally adjust, with rollups becoming more efficient. But the chain view shows that rollup efficiency gains have plateaued. The average compression ratio has not improved in six months. The low-hanging fruit has been picked.
Now, let’s dive into the core data. I have compiled a dataset of blob usage from the Dencun activation to the present. I will present the key findings.
First, the blob utilization rate. In the first month after Dencun, utilization was below 10% on average. By month 6, it had reached 45%. By month 12, it was 65%. At the current rate of growth, utilization will hit 100% within 18 months. This is a logistic curve, and we are past the inflection point. The weekly growth rate of blob usage is about 3.5%. That may not sound like much, but it compounds. At 3.5% weekly growth, usage doubles every 20 weeks. In 40 weeks, it quadruples. The blob limit is fixed. So the time to saturation is logarithmically approaching.
Second, the fee impact. I modeled the blob base fee at different utilization levels. At 80% utilization, the base fee is 100 gwei. At 90%, it jumps to 500 gwei. At 95%, it exceeds 1,000 gwei. Currently, the average blob fee is around 50 gwei. If utilization reaches 95%, the fee will be 20 times higher. That means a rollup that currently pays $1,000 per day for blobs will pay $20,000 per day. That cost will be passed on to users. If a rollup processes 1 million transactions per day, the per-transaction cost will increase by $0.02. That might not seem like a lot, but for a rollup that currently charges $0.001 per transaction, it is a 20x increase. Users will notice.
Third, the concentration risk. I analyzed the distribution of blob usage among rollups. The top 3 rollups (Arbitrum, Base, Optimism) account for 72% of all blob usage. This is a classic power-law distribution. If any of these rollups experiences a sudden spike in activity — say, a new DeFi protocol that attracts billions in TVL — the blob market will be flooded. The other rollups will suffer collateral damage. This is not a stable state. The market needs more capacity, but the supply is rigid.
Now, let’s address the contrarian narrative. Some argue that the blob limit can be increased quickly through a simple EIP. The Ethereum core developers are already discussing EIP-7623, which would increase the blob target from 3 to 6 per slot. But that would only double the capacity. It would not solve the problem permanently. In fact, it would encourage more wasteful usage, and saturation would occur again within a year. The real solution is to make rollups more efficient, but that requires coordination and incentives that are currently lacking. The contrarian view also says that L2s will migrate to alternative DA, such as EigenDA or Celestia. But as I mentioned, that sacrifices security. The market has already voted with its TVL: Ethereum-native DA is far more trusted. I do not see a mass migration happening unless blob fees become absurdly high — and by then, the damage to the user experience will already be done.
Hunting for the next narrative, I see a clear path. The blob fee crisis will be the dominant story of 2026. It will lead to a re-evaluation of L2 economics. Some L2s will collapse because they cannot sustain the fee increases. Others will pivot to new business models, such as subscription-based access or bundled transaction fees. The winners will be those that have already optimized their data compression and built fee buffers. The losers will be those that relied on cheap blobs as a crutch. I have already started tracking which L2s are preparing for this. I will share that analysis in a future article.
Let me bring in a personal experience from 2021, when I was writing about the NFT art renaissance. I interviewed a developer who was building an NFT marketplace on Ethereum. He told me that his biggest fear was not the gas fees of the L1, but the unpredictability of the L2 fees. He had lost money during the Optimism launch because the sequencer was overloaded. That experience taught me that fee stability is just as important as fee cheapness. Users will tolerate higher fees if they are predictable. The blob fee market is inherently unpredictable because it is a separate market with its own dynamics. This unpredictability will be a major source of friction for L2 applications.
From the ashes of 2017 to the fluidity of DeFi, I have seen narratives rise and fall. The blob saturation thesis is not a new idea; it has been discussed in academic circles since 2023. But the market has ignored it because the immediate effects were not visible. Now, the data is becoming undeniable. I have been sharing my findings with a small group of analysts and developers. The consensus is that the blob fee crisis will hit within the next 12 to 18 months. The question is not if, but when.
I want to end with a rhetorical question. If the blob market becomes saturated and fees double, what will happen to the current narrative that “L2s are the future of Ethereum”? Will the community accept higher fees, or will it turn against the rollup-centric roadmap? The answer to that question will determine the next phase of Ethereum’s evolution. As someone who has been through the ICO crash, the DeFi summer, and the NFT bubble, I know that the market always finds a way to adapt. But the adaptation is rarely painless. The blob saturation thesis is a warning, not a prophecy. It is a call for developers to start optimizing now, before the fees hit.
Let me leave you with a final data point. I calculated the historical cost of posting a blob from Dencun launch to today. The average cost per blob has increased from $0.50 to $3.00. That is a 6x increase in 18 months. If the trend continues, the cost per blob will reach $15 by 2026. That will make L2 transactions significantly more expensive than they are today. The rollups that survive will be those that have built efficient data compression and have contingency plans. The rest will be left behind. This is the quiet migration that is happening right now, beneath the surface of the daily price charts. The academic view vs. the chain view: the academic view says the market will self-correct. The chain view says the market is already struggling. I trust the chain.
In the next article, I will publish a detailed analysis of each major rollup’s blob consumption and their respective fee sensitivity. I will also provide a framework for users to calculate their own exposure to the blob fee risk. For now, I urge you to look at the data yourself. The writing is on the wall — or rather, on the blocks.
This is not a bearish take on Ethereum. It is a realistic take on the engineering challenges that still lie ahead. The Dencun upgrade was a brilliant piece of engineering, but it was not the final solution. The final solution will require a combination of better compression, improved blob markets, and perhaps a shift in how rollups use data. The narrative is shifting, and I am hunting for the next story. The blob saturation thesis is just the beginning.
— Avery Lopez, Berlin, 2025