Ethereum L2 fees plunge 95% after Dencun as sustainability questions mount

The Dencun upgrade activated in March 2024 cut transaction fees on major Ethereum rollups by 90 percent to 95 percent, pushing average costs below $0.10 and triggering a surge in network activity that pushed monthly transaction counts past 50 million.

"The blob market created a separate fee space that never competes with Layer 1 congestion, which is why costs stayed down even as usage exploded," Christine Kim, vice president of research at Galaxy Digital, said.

Layer 2 networks generated roughly $129 million in total revenue during 2025, according to available data. Of that, about $10 million was paid to Ethereum mainnet for settlement and security, while $119 million was retained by L2 operators. Base, the Coinbase-built rollup, alone generated $75.4 million in sequencer revenue in 2025, posting a 98.3 percent profit margin in May 2025 on $5.8 million in revenue. Starknet's average fee dropped from $0.654 to $0.033 by April 2024 and now sits near $0.029.

The sustainability of sub-$0.10 fees depends on blob space demand, which is currently capped at six blobs per block. The upcoming Fusaka upgrade, expected to include PeerDAS, could boost blob capacity and cut L2 fees by an additional 60 percent, but rising demand from new rollups may offset those gains.

How EIP-4844 Reshaped L2 Economics

The technical mechanism behind the fee collapse is EIP-4844, or Proto-Danksharding, which created a dedicated data space called "blobs" for rollups to publish transaction data. Before the upgrade, L2 networks posted data to Ethereum's main layer using calldata, a method that competed for block space with every other transaction on the network. Blob transactions operate in their own fee market, modeled on EIP-1559, where price is determined by supply and demand for that specific space.

The impact cascaded across the ecosystem. Base's transaction volume jumped 224 percent after Dencun. Optimism's calldata usage fell 81 percent. The base fee on Ethereum mainnet stabilized in the 1 to 3 gwei range on most days, reflecting persistent excess capacity on Layer 1. Unique active addresses on Ethereum reached multi-year highs, while the network's share of total DeFi value locked increased to 64 percent, even as the price of ETH declined roughly 10 percent during 2025.

The $119 Million Retention Gap

The redistribution of economic value across layers has become a central tension point. L2 networks retained roughly $119 million in revenue during 2025 while paying only $10 million to Ethereum for settlement and security — a ratio of about 321 to 1. Base's centralized sequencer model, operated by Coinbase, generated $75.4 million in sequencer revenue in 2025 alone, with a profit margin of 98.3 percent in May 2025.

Arbitrum's Timeboost ordering system, which lets users bid for sequencing priority, generated $2 million in fees during its first three months of operation. Aggregated monthly maximal extractable value across all L2 networks was estimated at roughly $8 million for September 2025.

The profitability model for sequencers has shifted toward capturing MEV as transaction fees collapsed. Most major L2s — including Arbitrum, Optimism, Base, zkSync, Starknet, Scroll, and Linea — operate with a single centralized sequencer, a structure that critics say concentrates both revenue and control.

What Comes After Fusaka

The Fusaka upgrade, expected to include PeerDAS, projects a capacity increase for blobs that could further reduce L2 fees by up to 60 percent. This would push transaction costs even lower, potentially accelerating adoption across gaming, DeFi, and payments use cases on Ethereum L2s.

Yet several variables could reverse the trend. Blob space, while currently exhibiting excess capacity, has a hard limit of six blobs per block. A sustained increase in rollup activity or the incorporation of new L2 networks could generate spikes in blob prices, passing those costs to end users. Sequencer decentralization, a stated goal on Ethereum's roadmap, would introduce additional operational costs through consensus mechanisms and multiple validators. Projects such as Espresso Systems and Flashbots are exploring confirmation-layer models that could preserve current operator revenue streams.

Some L2 networks also use subsidies through their native tokens to maintain artificially low fees and attract activity. The reduction or elimination of those subsidies would expose the actual operational cost of each network, potentially pushing fees higher.

The structural change in Ethereum Layer 2 fees constitutes a direct effect of a protocol upgrade with established technical foundations. The ability to maintain fees within the current range depends on the evolution of blob space demand, sequencer decentralization decisions, and the capacity of L2 networks to monetize transaction flow without exclusive reliance on subsidies. Available data indicates a permanent transformation in the network's economics, with implications for all ecosystem participants.

This article is for informational purposes only and does not constitute investment advice.