Research/Ethereum

The Blob Glut: Why Cheap Ethereum Data Is Squeezing L2 Profit Margins

Ethereum's blob space is consistently underutilized, driving L2 data costs near zero. Great for users, but it's reshaping rollup economics.

bcMaoNov 18, 2026

Ethereum’s blob fee market has a surplus problem. Since EIP-4844 launched with the Dencun upgrade in March 2024, rollups have posted data to a dedicated blob space instead of expensive calldata. The upgrade worked: L2 fees dropped by roughly 90%. But the supply of blob space has consistently outpaced demand, pushing the blob base fee to its 1-wei floor for most of 2025 and 2026. For users, this is a windfall. For rollup operators trying to build sustainable businesses, it creates a problem that is only getting worse.

How the Blob Fee Market Works

Blobs use a separate EIP-1559-style fee market. Each blob carries roughly 128 KB of data, and the network sets a target number of blobs per block. When usage exceeds the target, the blob base fee increases exponentially. When usage falls below the target, the fee decreases. The mechanism mirrors execution gas pricing, but with its own independent price discovery.

At the Dencun launch, the target was 3 blobs per block with a maximum of 6. This was deliberately conservative: Ethereum core developers wanted to observe network behavior before scaling up. The result was a fee market that spent most of its early months at or near the minimum price.

The Capacity Ramp-Up

Blob capacity has increased three times since Dencun. The Pectra upgrade in May 2025 implemented EIP-7691, doubling the target to 6 blobs and setting the maximum at 9. It also shifted the target-to-max ratio from 1:2 to 2:3, making fee adjustments less volatile. After Fusaka activated in December 2025, two Blob Parameter Only (BPO) forks raised capacity further: BPO1 moved the target to 10 and maximum to 15 in December 2025, and BPO2 pushed the target to 14 and maximum to 21 in January 2026.

UpgradeDateBlob TargetBlob MaxTarget:Max Ratio
Dencun (EIP-4844)Mar 2024361:2
Pectra (EIP-7691)May 2025692:3
BPO1Dec 202510152:3
BPO2Jan 202614212:3

Each increase outpaced demand growth. By early 2026, average blob utilization had fallen to roughly 20 to 30% of the new 14-blob target. As of late 2026, most blocks carry fewer than half the target, keeping the blob base fee near its floor.

The paradox: Ethereum’s blob scaling roadmap has succeeded at its stated goal of making L2 data cheap. But each capacity increase has driven the blob fee further toward zero, reducing Ethereum’s revenue as a settlement layer without a corresponding increase in demand.

How Cheap Is Cheap?

When blob demand sits below the target, the blob base fee converges to its minimum. For much of 2025 and 2026, that minimum was 1 wei per unit of blob gas, making the cost of posting a full blob effectively zero in dollar terms. Even during moderate demand, blob fees have typically remained orders of magnitude cheaper than the calldata they replaced.

Fusaka introduced EIP-7918, which sets a floor on the blob base fee tied to the execution base fee (roughly one-sixteenth of the L1 gas price). This prevents blobs from being literally free, but the floor is still extremely low: at typical L1 base fees, the minimum blob cost amounts to pennies per blob. The floor is designed to prevent blob spam and maintain a minimum ETH burn from settlement, not to generate meaningful revenue.

What This Means for L2 Data Costs

The impact on rollup economics is dramatic. Before Dencun, L2s paid for data as calldata on Ethereum L1, where costs ranged from thousands to tens of thousands of dollars per day during congested periods. After Dencun, those costs collapsed. By 2026, growthepie data shows all L2s combined paying Ethereum mainnet roughly $1,200 per day for settlement: a figure that once represented a single batch from a busy rollup.

The L2 Revenue Model Under Pressure

Rollups make money the same way most infrastructure businesses do: they charge users more than it costs to serve them. An L2 collects transaction fees from users, then pays Ethereum for data availability and settlement. The difference is the sequencer margin. When data costs were high, L2s could charge a meaningful premium over their L1 posting costs. Users still saved money compared to transacting directly on mainnet, so the arrangement worked for everyone.

With blob fees near zero, the cost side of this equation has effectively vanished. That sounds like good news for margins, but it creates a different problem: pricing power. When your biggest cost disappears, your users notice. L2 transaction fees have entered a race to the bottom, with median fees for a simple USDC transfer falling to roughly $0.02 on Base and even lower on some competitors. The margin may be high as a percentage, but the absolute revenue per transaction has cratered.

Who Is Profitable?

In a low-fee environment, volume is everything. Only a handful of L2s generate enough transaction volume to cover operations, development, and ecosystem incentives. Base, operated by Coinbase, has emerged as the clear leader: it generated roughly $75 to $83 million in sequencer revenue during 2025, with an estimated net profit of approximately $55 million after L1 data costs and the Optimism revenue share. Base processes over 60% of all L2 transactions, giving it the volume needed to sustain margins even at sub-cent fee levels.

Arbitrum, the second-largest rollup by TVL, operates with considerably thinner margins. Daily sequencer revenue averages around $55,000 compared to Base’s approximately $185,000, putting it near breakeven after operational costs. Optimism itself earns indirectly through the Superchain revenue share model, where Base pays the greater of 2.5% of sequencer revenue or 15% of net profit to the Optimism Collective.

L2Est. 2025 Sequencer RevenueL2 Transaction ShareApprox. MarginStatus
Base~$75 to $83M~60%High (~85%)Profitable
Arbitrum~$20M~15%Moderate (~55%)Near breakeven
Optimism (OP Mainnet)Indirect (Superchain share)~5%N/ASubsidized by ecosystem
zkSync EraNot disclosed~3%LowSubsidizing operations
Others (50+ chains)Minimal~17% combinedNegativeZombie or subsidized
The zombie chain problem: Analysts project that more than 40 L2s could become zombie chains by end of 2026: still producing blocks but moving little real value. Kinto has shut down, Loopring closed its wallet, Blast’s TVL dropped 97%, and Mint Blockchain ceased operations. The blob glut accelerated consolidation by eliminating data cost as a competitive moat.

Before and After: How EIP-4844 Changed L2 Economics

The shift from calldata to blob data fundamentally restructured rollup business models. Before Dencun, data posting was the dominant cost for most rollups, often consuming 60 to 80% of collected fees. After Dencun, data costs became negligible, turning execution and proving costs into the primary expenses.

The Margin Inversion

Pre-Dencun, rollups had a natural floor on user fees: they needed to cover calldata costs plus a margin. This created a pricing structure where L2 fees tracked L1 gas prices, usually at a significant discount but still meaningfully above zero. Users accepted this because even discounted L1 rates saved money compared to transacting on mainnet.

Post-Dencun, with data costs near zero, there is no natural floor except the execution cost on the rollup itself. This pushed L2 fees into free-fall. Average L2 fees for the top three rollups (Base, Arbitrum, Optimism) declined 99% from Q1 2024 to Q1 2026, dropping from roughly $0.18 to fractions of a cent. The absolute revenue per transaction shrank even as percentage margins appeared healthy.

The result is a market where only chains with millions of daily transactions can generate sustainable revenue. Base processes over 10 million transactions per day at peak, which makes sub-cent fees profitable in aggregate. A rollup processing 50,000 transactions per day at the same fee level generates roughly $1,000: not enough to pay for a single engineer.

Ethereum’s Revenue Problem

The blob surplus doesn’t just affect L2 operators. It has significant implications for Ethereum L1 itself. Post-Dencun, ETH burn from blob fees is negligible. Daily ETH burned from transaction fees dropped to as low as 53 ETH on some days in 2025, compared to thousands per day during the 2022 to 2023 fee spikes. With issuance running at approximately 800,000 ETH per year, the burn no longer offsets new supply, making ETH effectively inflationary since the upgrade.

Ethereum’s Q1 2026 logged a record 200 million L1 transactions, yet average fees fell to around $0.21. The network is busier than ever and earning less than ever: a direct consequence of successfully offloading activity to L2s while making settlement cheap. EIP-7918’s blob fee floor helps at the margin, but it was designed to prevent spam, not to restore revenue. Some analysts have projected it could supply 30 to 50% of total burn by mid-2026, though this remains speculative and depends on sustained execution layer demand.

What Happens When Blob Demand Exceeds Supply?

The current surplus will not last forever. As L2 adoption grows and new rollups launch, aggregate blob demand will eventually approach and exceed the target. When that happens, the EIP-1559-style pricing mechanism kicks in aggressively: the blob base fee increases by approximately 12.5% per block when usage exceeds the target. Brief demand spikes have already demonstrated this volatility: in late 2024, inscription-related blob activity (sometimes called “blobscriptions”) temporarily pushed blob fees to $27 before they returned close to zero.

The Fee Spike Problem

The exponential fee mechanism means blob prices do not rise gradually. They go from near-zero to expensive within a few blocks. This creates problems for L2 operators who must post data regularly regardless of price. During a spike, an optimistic rollup posting a batch every few minutes might see its data costs increase 100x within an hour. The rollup either absorbs the cost (destroying margins), passes it to users (creating fee unpredictability), or delays posting (weakening security guarantees).

This dynamic is a key reason L2s have explored data compression techniques, alternative data availability layers like Celestia and EigenDA, and batch optimization strategies. If blob space becomes consistently expensive, the economic case for Ethereum-native data availability weakens, and rollups may increasingly post data elsewhere: a scenario the Ethereum community calls “DA fragmentation.”

How L2s Are Adapting

Facing compressed margins and uncertain future data costs, L2 operators are pursuing several strategies to build sustainable businesses.

Volume Aggregation

The Superchain model (Optimism, Base, World Chain, Ink, and others) shares infrastructure costs across multiple chains. By standardizing the OP Stack and sharing sequencing, these chains amortize development and operational costs. Coinbase’s Base benefits from this directly: its parent company drives user acquisition at scale, ensuring the transaction volume that makes low fees work.

Vertical Integration

Some L2s are building revenue streams beyond transaction fees. Arbitrum’s ecosystem fund, powered by the ARB token, finances development without relying solely on sequencer profits. Others are integrating DEXs, lending protocols, and bridge infrastructure to capture value at multiple points in the transaction lifecycle.

Alternative DA Layers

Chains like validiums post data to off-chain layers rather than Ethereum blobs, further reducing costs at the expense of security guarantees. As blob fees oscillate between near-zero and spike-level pricing, the calculus around alternative DA becomes increasingly attractive for cost-sensitive rollups.

The Fee War Dynamics

The blob surplus has intensified the L2 fee war by removing data costs as a differentiator. When every rollup pays essentially nothing for settlement, the remaining competitive dimensions are execution speed, ecosystem quality, distribution channels, and brand. This favors well-funded chains with existing user bases and penalizes smaller rollups that cannot compete on marketing spend.

The consolidation is already visible. Base, Arbitrum, and Optimism together handle nearly 90% of L2 transactions. More than 50 smaller rollups are effectively inactive. The blob glut did not cause this consolidation directly, but it eliminated the possibility that differentiated data management could serve as a competitive advantage.

A Different Model: Bitcoin L2 Cost Structures

The Ethereum blob surplus highlights a broader question about how Layer 2 networks relate to their base layers economically. Bitcoin L2s face a structurally different environment. Bitcoin block space is fixed at roughly 4 MB per block with no mechanism to increase it without a hard fork, and demand frequently exceeds supply. Fees averaged over $3 per transaction during congested periods in 2024 and 2025, with spikes reaching $30 or more during inscription and Runes activity.

This means Bitcoin L2s like Spark operate in an environment where base layer costs are real and persistent. There is no blob-style discount on Bitcoin: every byte of on-chain data competes in the same fee market. This creates different economic pressures. Bitcoin L2s must minimize their on-chain footprint by design, not just for efficiency but for survival.

Spark addresses this through its statechain-based architecture, where transfers happen entirely off-chain via key rotation rather than data posting. Unlike rollups, which must regularly post batches of transaction data to their base layer, Spark only touches Bitcoin L1 for deposits and exits. This means Spark’s operating costs do not scale linearly with transaction volume, and the protocol is not exposed to the same kind of base-layer fee volatility that affects Ethereum rollups.

Different scaling tradeoffs: Ethereum chose to expand its data layer (blobs) to reduce L2 costs, inadvertently creating a surplus that undercuts L1 revenue. Bitcoin’s fixed block space means L2s must be architecturally efficient from day one. Neither approach is strictly better: they reflect different philosophies about the relationship between base layers and scaling networks.

What Comes Next

The blob surplus is temporary in the sense that demand will eventually catch up with supply. But “eventually” could be years, especially as the Ethereum Foundation continues to plan further capacity increases toward the full danksharding roadmap. Each capacity expansion resets the clock, pushing the equilibrium point further into the future.

For rollup operators, the near-term outlook is clear: revenue depends on volume, not margins. The chains that survive will be those with enough users and transactions to generate meaningful aggregate fees at sub-cent price points. For everyone else, the blob glut will continue to squeeze already thin operations.

For Ethereum L1, the challenge is more existential. If blob space remains underpriced, the network risks becoming a low-revenue settlement layer that subsidizes L2 profitability without capturing proportional value. EIP-7918’s fee floor is a first step toward addressing this, but the fundamental tension between cheap data availability and sustainable L1 economics remains unresolved.

Developers building on Ethereum L2s or evaluating scaling architectures should study the blob fee market mechanics and the L2 fee war dynamics to understand where rollup economics are headed. For those exploring Bitcoin-native scaling, the Spark documentation and SDK provide an alternative model where L2 costs are decoupled from base-layer data pricing entirely.

This article is for educational purposes only. It does not constitute financial or investment advice. Bitcoin and Layer 2 protocols involve technical and financial risk. Always do your own research and understand the tradeoffs before using any protocol.