Ethereum L2s: Optimistic vs. ZK Rollups — Where Are We in 2025?

# Ethereum L2s: Optimistic vs. ZK Rollups — Where Are We in 2025?

The L2 narrative has shifted faster than most people expected. Two years ago, ZK rollup advocates were confidently predicting that fraud proof windows and 7-day withdrawal delays would make optimistic rollups obsolete within a short time. In practice, the TVL distribution in 2025 looks quite different from that prediction, and it's worth being precise about why.

## How the Two Approaches Work

*Optimistic rollups* — Arbitrum One, OP Mainnet, Base — execute transactions off-chain and post compressed transaction data to Ethereum. The "optimistic" refers to the core assumption: transactions are assumed valid unless challenged. A 7-day challenge window exists during which anyone can submit a fraud proof demonstrating an invalid state transition.

*ZK rollups* — zkSync Era, StarkNet, Polygon zkEVM, Linea — execute transactions off-chain and post a cryptographic validity proof with each batch. The proof mathematically guarantees correctness; no challenge period is required. Finality is achieved when the proof verifies on-chain.

The theoretical security argument for ZK is cleaner. You're not trusting that someone will challenge invalid proofs; you're relying on mathematics. This is genuinely real and it matters, particularly for applications where fast, trust-minimized finality has direct value.

## The Fraud Proof Problem

Optimistic rollup fraud proofs have had a complicated history worth acknowledging. Arbitrum's fraud proof system has been live and functional since 2022. Optimism's — in the form of Cannon — became functional in 2023 but wasn't fully enabled on OP Mainnet until 2024. Before that, OP Mainnet technically ran with a training-wheel security model: the fraud proof mechanism existed in the codebase but wasn't permissionlessly executable by anyone.

This is often glossed over in L2 comparisons. "Fraud proof" as an active, operational, trust-minimized security mechanism is different from "fraud proof exists in the design." Arbitrum had the former considerably longer than Optimism's OP Stack chains.

The 7-day withdrawal window is a genuine UX problem for anyone wanting to move capital from L2 to L1 without using a third-party bridge. Seven days is a long time. Third-party bridges solve the UX problem by providing liquidity but introduce their own trust assumptions and bridge-specific risks — which is a separate problem the space is still working through.

## ZK Proof Generation: Real Costs and Real Progress

ZK proof generation is computationally expensive, and the hardware requirements for generating validity proofs at production scale are substantial. In 2023 and early 2024, zkSync Era and StarkNet both experienced periods where proof generation lagged, creating delays in finality that weren't supposed to happen in the theoretical model.

This has improved substantially. Hardware acceleration — GPUs, purpose-built ZK chips from vendors including Ingonyama and others — has made proof generation cheaper and faster. The trajectory is clearly in the right direction. But the infrastructure cost is real and shows up in the economic model of ZK rollup operators in ways that affect fees and throughput decisions.

## Where the TVL Actually Is

Despite the cleaner theoretical security model of ZK rollups, the 2025 TVL distribution heavily favors optimistic rollups. Arbitrum One holds the largest L2 TVL by a significant margin. Base — Coinbase's OP Stack chain — grew faster in 2024 than any other L2 and now hosts substantial DeFi activity.

The explanation isn't that users understand or prefer fraud proofs over validity proofs. Most DeFi users don't think about the underlying proof system when choosing where to trade. The explanation is that optimistic rollups got there first, have mature developer tooling, have established liquidity pools and ecosystem depth, and carry lower fees due to simpler proof infrastructure.

zkSync Era has meaningful TVL. StarkNet has a committed developer community building native Cairo applications. Polygon zkEVM has institutional backing and enterprise relationships. None has closed the gap with Arbitrum or Base in terms of raw activity, and the gap widened rather than narrowed in 2024.

## The Convergence Scenario

The endgame scenario that Ethereum researchers have discussed involves both approaches eventually converging: optimistic rollups replacing their fraud proof mechanisms with ZK validity proofs. Optimistic rollup teams aren't philosophically opposed to this; they're following a practical roadmap where working products ship first and security upgrades follow.

If ZK proof generation continues getting cheaper and faster — and the evidence suggests it will — the main remaining disadvantage of ZK rollups diminishes. At that point, the question becomes whether existing optimistic rollup ecosystems adopt ZK validity proofs, or whether ZK-native chains grow to displace them.

I think the former is considerably more likely. Arbitrum's ecosystem didn't develop because of Arbitrum's proof system. Migrating to a different proof mechanism is technically feasible without disrupting the applications built on top. The transition would be mostly invisible to end users.

But "more likely" and "imminent" are different things.

> **Key Takeaway:** ZK rollups have the cleaner security story, but optimistic rollups won the TVL race by building first and iterating faster. The gap is likely to persist in the medium term. Convergence — optimistic chains adopting ZK validity proofs — is plausible as a long-run outcome but isn't happening on a near-term timeline.

Ethereum L2s: Optimistic vs. ZK Rollups — Where Are We in 2025?

// COMMENTS

ON THIS PAGE