The widening gap between Ethereum blob burn and network costs

The direct blob base-fee burn reached only 0.293 ETH per week during the snapshot period ending September 3, 2026. At a price of $2,350 per ETH, this equates to an annualized burn of $37,500. This amount fails to cover the annual costs for a fleet of 12,000 replicated nodes. I calculate the central estimate for the cost of such a fleet at $0.5 million per year. This means the current direct blob burn covers only about 7.5% of the network cost. The network requires significantly more to compensate for bandwidth, electricity, and hardware headroom. I observe that the direct burn does not compensate operators for the work they perform. The discrepancy between the $37,500 burn and the $0.3 to $1.0 million annual network cost remains stark. In a strict zero-blob counterfactual where no blob storage or processing exists, the long-run avoidable network cost is approximately $350,000. This figure accounts for the existing machines and connectivity that nodes already maintain. The central gap between direct burn and network cost reaches roughly $462,500 per year. Because the direct burn is so small, the network cost is about 13 times larger than the amount burned. For 12,000 nodes, the annual SSD and power floor sits between $32,400 and $64,700. This gap remains because the direct blob base fee only covers the data itself and not the physical resources required to keep that data available.

Rollups like Base and Linea saw improved profit margins following the Pectra upgrade in May 2025. Base earned $1.22 million in revenue after onchain costs. Since Pectra went live, rollups increased their daily blob purchases by 20.8%. This increase brought daily blob purchases from 21,200 to 25,600. This represents a difference of 2.7 gigabytes of data capacity daily pre-upgrade against 3.3GB today. However, the average number of blobs per block remains 33% below the updated target rate. This surplus capacity makes blobs virtually free. The median cost per blob object since Pectra went live is just $0.00000000035. This represents a near 100% decline in blob object fees compared to the 60 days before the upgrade. In those 60 days, rollups paid a daily average of $16,250 in blob fees, totaling $1.095 million. Rollups have paid a daily average of $11,015 in total costs since Pectra, which is a 51% decline from the $20,660 average seen previously. Linea and Base maintain the strongest percent margins after onchain costs at 98.86% and 98.54%, respectively. While rollups purchase more data, they only use 40% of the maximum amount of data capacity available each day. Blobs have been 86% full on average post-Pectra against 82% full in the 60 days before the upgrade. The reduction in blob costs has improved the profit margins of rollups in both relative and absolute terms.

The Pectra upgrade increased the target blob count from 3 to 6. It also raised the maximum from 6 to 9. This change expanded the data capacity of the network. Users see this through lower L2 transaction costs. The fee market uses a specific mechanism to prevent total fee collapse. EIP-7918 provides a reserve-price relationship tied to execution gas. This prevents the fee from falling too low when usage is low.

The technical foundation relies on KZG polynomial commitments to verify data without every node downloading everything. Each blob carries up to 128 kilobytes of data. I find the current setup efficient for rollups but insufficient for nodes. The network requires 2,000 supernodes with 128 groups and 500 large nodes averaging 96 groups. Another 9,500 nodes carry 4 to 8 groups each. These nodes must hold data for 4,096 epochs. With the active target of 14 blobs per block and the maximum of 21, the current average usage of 5.70 blobs per block suggests that the market remains in a reserve-price regime where the network does not encounter actual scarcity. I note that the 12,000 nodes include 2,000 supernodes and 500 nodes custodying 64 to 127 columns. EIP-7691 increases blob throughput to improve availability, while EIP-7840 standardizes blob scheduling across clients to support scaling.

The decision to reduce the target blob count to manufacture scarcity remains a point of contention. Some argue a target of 5 would create upward fee pressure. I believe that lowering the target does not directly finance the infrastructure that carries blobs. It also does not increase proposer income. The current usage of 5.70 blobs per block is well below the target of 14. This capacity leaves plenty of room for growth. You probably already know that Pectra increased the target blob count. I find the argument that reducing the target does not help operators to be sound because burn does not pay electricity bills. The current target of 14 is well above the usage of 5.70. If demand remains at the current 5.70 level, a target of 5 would create upward fee pressure, but a target of 6 or 7 might not. Will the current pricing mechanism ever satisfy the needs of the 12,000 nodes?