The State of Ethereum Rollups: 2026 Edition

The Ethereum rollup ecosystem has entered a period of accelerating maturity. After three years of infrastructure build-out, the twelve major rollup networks collectively processed 4.8 billion transactions in Q1 2026 — a figure that exceeds Ethereum mainnet's historical cumulative count by a factor of three. This is not a trend. It is a structural shift.

Transaction volumes on leading networks now track upward with near-linear consistency, driven primarily by DeFi protocol migration and the emergence of rollup-native application ecosystems.

Fee compression — long the central promise of the rollup thesis — has materialized. Average transaction costs across optimistic rollups fell to $0.0023 per transaction in March 2026, down from $0.18 in January 2024. For ZK rollups with proof aggregation, the figure stands at $0.00091.

"The economics of Ethereum have been rewritten. L1 is now a settlement layer; L2 is where users live. The question is no longer whether rollups will dominate — it's which architecture wins the long game."

Yet the picture is not uniformly optimistic. Centralization risk in sequencer infrastructure, the opacity of proving systems, and fragmented liquidity across an expanding L2 landscape present structural challenges that the ecosystem has not yet resolved. This report examines each in turn.

ZK rollup proving times have collapsed from days to seconds¹, enabling near-synchronous finality. The performance trajectory continues to outpace even optimistic projections from 2023.

Daily active addresses across all tracked rollups reached 8.2 million in March 2026, representing a +312% increase year-over-year. The distribution of this activity, however, reveals important concentration dynamics that aggregate figures obscure.

Arbitrum One retains the highest absolute TVL at $18.4B, though its market share of the rollup ecosystem has contracted from 41% to 29% as OP Stack chains — Base, Mode, and Zora — absorb significant consumer and social application volume. The trend in Arbitrum's share reflects competitive pressure, not absolute decline.

"Base processed more transactions in February 2026 than all of Ethereum mainnet did in the entirety of 2022. The consumer chain hypothesis has been validated."

ZK rollup adoption is experiencing its inflection point. Combined ZK network TVL crossed $12.1B in Q1 2026, representing 19% of total L2 TVL — up from 8% twelve months prior. The proving time breakthrough by the Groth16 acceleration projects has been the primary catalyst.²

Unique deployer addresses — a proxy for developer activity — totaled 94,000 across tracked rollups in Q1, with zkSync Era leading on this dimension at 31,200 deployers. The maturation of rollup-native tooling, particularly the EVM equivalence improvements in ZK environments, has lowered the barrier for contract deployment substantially.

The rollup fee market has bifurcated. On one side, high-throughput chains operating at near-capacity generate substantial sequencer revenue; on the other, newer chains subsidize user activity aggressively, accepting current-period losses against future market share. Understanding this bifurcation is essential for evaluating which networks will sustain long-term development funding.

Sequencer revenue — the spread between user-paid fees and L1 settlement costs — has been the primary revenue model for optimistic rollups. In Q1 2026, Arbitrum's sequencer generated approximately $41M in net revenue, while Coinbase's Base generated $67M, a figure that reflects both Base's higher transaction volume and its more aggressive fee settings.

The introduction of EIP-4844 blob transactions in 2024 fundamentally altered L1 settlement economics, reducing data availability costs by ~85%. Chains that rapidly migrated to blob storage gained a significant cost advantage, as the per-transaction settlement cost dropped from approximately 2,100 gas to 320 gas equivalents.

"EIP-4844 was the single most important economic event in rollup history. The chains that moved fastest captured the margin advantage, and that advantage compounds over time."

ZK rollup economics present a more complex picture. Proof generation costs — once the dominant expense — have declined dramatically with hardware acceleration and proof aggregation. However, the capital requirements for ZK prover infrastructure remain substantial.³ Several ZK networks are exploring prover marketplace models to commoditize this function.

The architectural divergence between optimistic and ZK rollup designs has narrowed in ways that would have seemed implausible as recently as 2023. ZK rollups have achieved full EVM equivalence — the ability to execute standard Ethereum bytecode without modification — while proving times have compressed to the 200–400ms range for standard transactions. This removes the most significant practical barriers to ZK rollup adoption by established DeFi protocols.

Optimistic rollups retain a meaningful advantage in composability. The 7-day withdrawal period — once considered a serious UX liability — has been largely abstracted away by liquidity bridges and intent-based transaction systems. However, the native finality gap remains a genuine protocol-level distinction with implications for certain high-security use cases.⁴

The most significant architectural shift underway is the emergence of the "shared sequencer" model, in which multiple rollup chains coordinate transaction ordering to enable atomic cross-chain composability. Espresso Systems, Astria, and SUAVE represent distinct approaches to this problem. If successful, shared sequencing would partially dissolve the liquidity fragmentation that currently represents the most significant user-experience friction in the multi-rollup ecosystem.

"The question is not which rollup type wins. The question is whether the rollup interoperability stack develops fast enough to prevent the ecosystem from permanently fragmenting into competing L2 silos."

Decentralized sequencing — independent of the shared sequencer approach — has progressed more slowly. The economic incentive alignment for decentralized sequencers remains contested, and the performance overhead of consensus mechanisms within the sequencer layer is non-trivial at current transaction volumes.

The rollup landscape in 2026 and beyond will be shaped by three dynamics: the resolution of interoperability fragmentation, the commoditization of proving infrastructure, and the regulatory treatment of sequencer operators as financial intermediaries. Each carries substantial uncertainty, but the directional vectors are becoming clearer.

Interoperability protocols — including ERC-7683 (cross-chain intents), the OP Superchain vision, and Polygon's AggLayer — are converging on a design space in which user-facing applications can abstract chain boundaries entirely. If this abstraction succeeds, the "which rollup" question becomes irrelevant to end users, and competition shifts to developer tooling, fee levels, and governance credibility.

Proving commoditization represents a longer-arc opportunity. As ZK proof generation becomes a standardized compute function — purchasable from marketplaces, executed on specialized hardware, priced by proof complexity rather than protocol — the cost advantages of ZK rollups will extend to networks that lack Starknet's or zkSync's proving infrastructure investments.⁵

"By 2028, we expect that the question of 'ZK vs. Optimistic' will be as relevant as 'RISC vs. CISC' — architectural detail that matters enormously to engineers and not at all to users."

Regulatory signals from the United States and European Union remain the most significant wildcard. Classification of sequencers as money transmitters or financial intermediaries would impose compliance requirements that favor permissioned architectures and disadvantage decentralized sequencer initiatives. The outcome of Coinbase's ongoing regulatory posture with Base will be closely watched as a precedent.

Our central scenario for 2026: combined rollup TVL crosses $100B, daily active addresses reach 15M, and ZK rollup market share grows to 28% of total L2 TVL, driven by proving cost parity with optimistic networks and continued improvement in developer tooling.