Skip to content

Ethereum Mainnet vs Rollups: Rollups Win on Cost, Not on Exits

Arbitrum, Base, OP Mainnet, zkSync, Starknet, Linea, Scroll — 7 networks compared by value secured, fees, security stage, and real use-case fit.

Updated

Ethereum Mainnet vs Rollups: Rollups Win on Cost, Not on Exits

The Layer 2 Ecosystem After EIP-4844

The Ethereum Layer 2 ecosystem is a collection of scaling networks that execute transactions off Ethereum mainnet, compress and post transaction data or cryptographic proofs to Ethereum for final settlement, and inherit its security guarantees at a fraction of the base-layer cost. On November 9, 2026, ETHBEAT (formerly L2BEAT) showed about $36.4 billion in total value secured across the Layer 2 projects it tracks, with Base and Arbitrum One together holding about 63% of it. The structural catalyst for Layer 2 adoption was EIP-4844, the Dencun upgrade deployed on Ethereum mainnet in March 2024, which introduced "blob" transactions and cut L2 data-posting costs to Ethereum by 50–90% [1]; in CoinBureau's February 20, 2026 comparison table, average fees on all seven networks it compared were under 4 cents [3]. The seven networks compared in this guide averaged about 10.5 million transactions a day combined in the week to November 8, 2026 (ETHBEAT), and BlockEden reported combined L2 volume exceeding Ethereum mainnet's [1]. Post-Dencun fee parity means cost alone no longer drives chain selection; architecture, security posture, and ecosystem depth are now the defining criteria.

Quick Answer: As of November 9, 2026, ETHBEAT (formerly L2BEAT) showed about $36.4B in total value secured across the Layer 2 projects it tracks. EIP-4844 (March 2024) cut data-posting costs 50–90% (BlockEden), and CoinBureau's February 2026 table put average fees on the seven networks it compared under 4 cents. Base now leads Arbitrum One on value secured ($13.8B vs $9.1B; L2BEAT had Arbitrum One slightly ahead at the end of May 2026) and on activity (about 8.0M vs 1.1M daily transactions, 7-day average). Architecture and security stage — not fees — now determine network selection.

Arbitrum One and Base dominate the capital picture. In CoinBureau's February 20, 2026 snapshot, Arbitrum One held $16.88 billion in total value secured (TVS) and Base $10.74 billion [3]; OP Mainnet added $1.91 billion to the optimistic rollup total [3]. The ZK rollups covered here are far smaller: on November 9, 2026, L2BEAT showed about $0.32 billion secured on Starknet, $0.30 billion on Linea, $0.21 billion on zkSync Era and $0.03 billion on Scroll. These concentration patterns are self-reinforcing: deep liquidity attracts more protocols, which attract more users, which attract more liquidity — making value secured a lagging indicator of prior ecosystem investment rather than a real-time measure of network quality.

The real decision-making challenge for traders and developers is not whether to use a Layer 2 — the economics make that a clear choice — but which L2 to use and why. With fees now indistinguishable in practical terms, the meaningful distinctions involve withdrawal finality mechanics, smart contract language compatibility, DeFi TVL depth by protocol category, and the governance maturity underpinning actual security guarantees. Those distinctions are the subject of this guide.

Optimistic Rollups: How Arbitrum, Base & OP Mainnet Work

Optimistic rollups are Layer 2 networks that assume all submitted transactions are valid by default, execute them off-chain, and post transaction data to Ethereum mainnet without a validity proof attached. Instead of proving correctness upfront, they rely on a fraud-proof challenge window — about 6.4 days on Arbitrum One and 7 days on OP Mainnet, and, per Base's documentation, 5 days on Base since its Beryl upgrade (1 day when both a TEE proof and a ZK proof back the same proposal) — a period during which any network participant can submit cryptographic evidence that a transaction was invalid, triggering a dispute resolution process [1]. Only once that window expires without a successful challenge do native withdrawals finalize on Ethereum mainnet. The architectural trade-off is direct: lower computational overhead on the proving side enables full EVM compatibility, which means existing Solidity smart contracts port to Arbitrum One, Base, or OP Mainnet with near-zero code changes. That compatibility helps explain why capital concentrates on optimistic rollups: established DeFi protocols deployed on them without rewriting a line of code. CoinBureau's February 20, 2026 snapshot put Arbitrum One alone at $16.88 billion in value secured [3]; on November 9, 2026, L2BEAT showed Base ($13.8 billion) and Arbitrum One ($9.1 billion) together holding about 63% of the roughly $36.4 billion secured across Layer 2s. For retail traders, this native withdrawal window (up to about a week) is the most immediately practical constraint to internalize.

A multi-day native withdrawal window (about 6.4 days on Arbitrum One, 7 on OP Mainnet, 5 on Base) does not mean funds are locked for that long. Third-party liquidity bridges such as Across Protocol and Stargate Finance provide faster withdrawals from optimistic rollups by fronting liquidity on Ethereum mainnet and recouping funds once the challenge window expires. Their fee and settlement time vary with network congestion and bridge utilization, so check the quote the bridge shows before confirming. For smaller positions, compare that fee with the cost of waiting out the native window. The challenge window becomes most relevant when moving very large positions where bridge liquidity depth or fee sensitivity becomes a material constraint.

According to BlockEden Research (February 2026): Arbitrum's implementation of optimistic rollups delivers 40–60 transactions per second with full EVM compatibility. GMX, Uniswap, Aave, and Curve all have major deployments on Arbitrum.

Arbitrum's development roadmap reflects a deliberate strategy to expand beyond EVM-native Solidity developers. Arbitrum Stylus, now live in production, allows developers to write smart contracts in Rust, C, and C++ — languages compiled to WebAssembly and executed alongside the existing EVM environment [1]. The Stylus Sprint developer program attracted 147 submissions, a meaningful signal that the systems-programming community — historically excluded from Ethereum L2 by the Solidity requirement — is beginning to engage with the network [1]. WebAssembly contracts also offer performance characteristics suited to computationally intensive applications — high-frequency on-chain derivatives, complex pricing engines, and AI inference at the contract layer — that would be prohibitively expensive in pure Solidity.

Base, built on the OP Stack by Coinbase, has taken a different approach to growth: rather than competing for DeFi capital depth, it has focused on consumer applications and retail onboarding via Coinbase's existing user base. In CoinBureau's February 20, 2026 snapshot, with 12.89 million daily transactions and 382,500 daily users, Base led the networks it compared on raw activity despite holding substantially less value secured than Arbitrum One [3]. On L2BEAT's value-secured measure the order has since flipped: Arbitrum One led Base $13.6 billion to $13.1 billion at the end of May 2026, while on November 9, 2026, Base led $13.8 billion to $9.1 billion. This reflects a structural distribution advantage: direct Coinbase Wallet integration and consumer-facing application primitives generate high transaction counts with smaller per-user capital commitments than DeFi protocols — a different growth vector, not a lesser one.

ZK Rollups: zkSync, Starknet, Linea & Scroll Side by Side

ZK rollups are Layer 2 networks that generate a cryptographic validity proof — a mathematical certificate of correctness — for every batch of transactions, then post that proof alongside compressed transaction data to Ethereum mainnet for on-chain verification. Because the proof mathematically guarantees the validity of every included transaction, no challenge window is required: withdrawals finalize once the batch's proof is verified on Ethereum, so timing depends on each network's proof-submission frequency and any security delay it adds (zkSync Era, for example, documents a minimum 3-hour delay before batches are executed on Ethereum). This finality advantage has meaningful implications for institutional users and high-frequency traders who need cryptographic certainty of settlement rather than economic incentivization via fraud challenges. On November 9, 2026, the four ZK rollups covered here — Starknet, Linea, zkSync Era and Scroll — together secured about $0.87 billion on L2BEAT, less than a tenth of Arbitrum One's $9.1 billion.

Within the ZK rollup category, the critical technical distinction is between SNARK-based and STARK-based proof systems. SNARK-based networks — zkSync Era, Linea, and Scroll (Polygon zkEVM, once part of this group, was shut down in July 2026) — produce compact proofs with a fixed on-chain verification cost (zkSync Era's Boojum prover wraps a STARK proof inside a SNARK, per Nethermind), while supporting Solidity development. The trade-off is a requirement for a trusted setup ceremony: a coordinated multi-party computation event that initializes the cryptographic parameters and must be trusted to have been conducted honestly. Scroll's documentation lists only a short set of EVM differences and says they won't affect the average Solidity developer, making it a low-friction migration target for teams porting existing Solidity contracts.

According to Nethermind Engineering's comparative analysis: SNARK-based systems are expected to have constant proof verification costs, zkSync Era's Boojum proof system wraps a STARK proof inside a SNARK, and Starknet uses a STARK-based proving system expected to have polylogarithmic verification cost.

Starknet is the principal STARK-based rollup in active production. Its proof system requires no trusted setup, derives security from hash functions rather than elliptic curve assumptions, and, per StarkWare, relies on cryptographic assumptions that resist quantum attacks. The practical consequence is larger proof sizes and higher on-chain verification costs. More significantly for adoption, Starknet uses Cairo — a non-EVM smart contract language built for STARK provability — rather than Solidity. Starknet's native account abstraction can be used for features such as social recovery, according to its documentation, but the language requirement limits dApp portability and constrains the developer pool. On L2BEAT's measure, value secured on Starknet slipped from about $0.53 billion at the end of May 2026 to $0.32 billion on November 9 — its technical advantages have yet to translate into deep capital.

Linea, maintained by ConsenSys, and zkSync Era, by Matter Labs, both occupy the SNARK space with a critical governance distinction from the Stage 1 networks: both were Stage 0 in CoinBureau's February 20, 2026 table [3] and still were on L2BEAT on November 9, 2026, meaning upgrade authority is not yet constrained the way L2BEAT requires for Stage 1. This governance gap is addressed in the security section below. Neither has drawn deep capital — despite Linea's MetaMask ecosystem integration — and both have slipped on L2BEAT's value-secured measure: Linea from about $0.43 billion at the end of May 2026 to $0.30 billion on November 9, and zkSync Era from $0.28 billion to $0.21 billion.

Fees & Throughput: 2026 Network Performance Data

Transaction fees across Ethereum's major Layer 2 networks converged dramatically following the Dencun upgrade in March 2024. CoinBureau's February 20, 2026 comparison table put average fees on the seven networks it compared between $0.0007 (OP Mainnet) and $0.0348 (Linea) per transaction [3]. On the L2Fees board CoinBureau cites, a simple ETH transfer cost about $0.07 on zkSync Era and $0.09 on Arbitrum One and OP Mainnet (Base was not listed) [3]. The spread between those networks is a few cents per transaction — economically marginal for all but the highest-frequency trading strategies. The more operationally significant metrics are daily transaction throughput, active address density, and the nature of applications generating that volume, since transaction-type distribution directly affects congestion dynamics and MEV exposure for DeFi participants.

Network Type Value secured (ETHBEAT, Nov 9, 2026) Daily transactions (ETHBEAT, 7-day avg., Nov 2–8, 2026) Stage (ETHBEAT, Nov 9, 2026)
Arbitrum One Optimistic $9.1B 1.13M Stage 1
Base Optimistic $13.8B 8.02M Stage 1
OP Mainnet Optimistic $1.4B 1.18M Stage 1
zkSync Era ZK (SNARK) $0.21B 15.1K Stage 0
Linea ZK (SNARK) $0.30B 14.3K Stage 0
Scroll ZK (SNARK) $0.03B 15.1K Stage 0
Starknet ZK (STARK) $0.32B 81.7K Stage 1

Source: ETHBEAT (formerly L2BEAT) — value secured and stage as of November 9, 2026; daily transactions are the 7-day average for November 2–8, 2026 (activity). Fee figures are discussed in the text above.

Base's throughput in CoinBureau's February 20, 2026 snapshot — 12.89 million daily transactions and 382,500 daily users — requires context to interpret correctly [3]. High transaction count does not imply high value secured. Consumer and social applications — on-chain social graphs, creator monetization primitives, lightweight gaming interactions — generate large volumes of low-value transactions. Each individual transaction involves a small capital commitment, but scales across a broad user base. Arbitrum One's 4.30 million daily transactions in the same snapshot [3], by contrast, ran on a network BlockEden describes as DeFi-led, with the deepest liquidity for automated market makers, lending protocols, and derivatives platforms [1]. In CoinBureau's February 20, 2026 figures, Arbitrum One held roughly 1.6× Base's value secured despite processing only one-third of its daily transactions [3]. By November 9, 2026, L2BEAT showed Base ahead in value secured ($13.8 billion vs Arbitrum One's $9.1 billion) while still processing about seven times as many daily transactions (about 8.0 million vs 1.1 million, 7-day average).

BlockEden puts OP Mainnet at approximately 130 TPS against 40–60 TPS for Arbitrum [1], but its 2.35 million daily transactions in CoinBureau's February 20, 2026 snapshot suggest throughput is not the current bottleneck [3]. ZK networks also carry far less traffic today — on L2BEAT, zkSync Era averaged about 15,000 and Starknet about 82,000 daily transactions in the week to November 8, 2026, under 1 TPS each — and ZK proof generation adds computational overhead that limits batch submission frequency. As proof generation hardware and algorithm optimizations mature through 2026 and 2027, ZK throughput ceilings could rise, narrowing the performance gap with optimistic rollups.

Security & Decentralization: L2BEAT Stage Rankings Explained

L2BEAT's "training wheels" framework is the standard methodology for evaluating how decentralized and trustless an Ethereum Layer 2 network is in practice — distinct from its theoretical architecture. The framework assigns one of three maturity stages. Stage 0 networks are operational but retain centralized multi-sig upgrade authority: a sufficiently compromised key set could alter the protocol unilaterally without on-chain governance. Stage 1 networks have deployed active fraud or validity proofs with meaningful on-chain governance constraints limiting upgrade key authority. Stage 2 is the state CoinBureau describes as closer to being controlled by code rather than humans [3]. In CoinBureau's February 20, 2026 table, five of these networks were Stage 1 — Arbitrum One, Base, OP Mainnet, Starknet, and Scroll — and zkSync Era and Linea were Stage 0 [3]; by November 9, 2026, L2BEAT listed Scroll at Stage 0 too. None of the networks covered here has reached Stage 2, making the decentralization race the defining governance story of 2026.

Network L2BEAT Stage (Feb 20, 2026, via CoinBureau) Proofs Active Upgrade Authority Primary Stage 2 Blocker
Arbitrum One Stage 1 Yes — fraud proofs Security Council (multisig with governance constraints) Admin override capability retained
Base Stage 1 Yes — fault proofs OP Stack Security Council Admin override capability retained
OP Mainnet Stage 1 Yes — fault proofs OP Stack Security Council Admin override capability retained
Starknet Stage 1 Yes — STARK validity proofs StarkWare governance-constrained multisig Admin override capability retained
Scroll Stage 1 Yes — ZK validity proofs Governance-constrained multisig Admin override capability retained
zkSync Era Stage 0 Yes — ZK validity proofs Centralized multisig (Matter Labs) Stage 1 not yet reached; upgrade authority not yet constrained
Linea Stage 0 Yes — ZK validity proofs Centralized multisig (ConsenSys) Stage 1 not yet reached; upgrade authority not yet constrained

Sources: CoinBureau [3] — L2BEAT classifications as of February 20, 2026.

According to CoinBureau's 2026 Layer 2 security analysis: L2BEAT's stages are a “training wheels” maturity framing — Stage 0 means heavy dependence on centralized controls, Stage 1 meaningful progress but still not fully code-controlled, and Stage 2 closer to being controlled by code rather than humans. The simplest user rule is to watch decentralization progress, upgrade controls, and who can change the system.

The practical implication for capital allocation: Stage 0 is not a disqualifier — zkSync Era and Linea have operated under Stage 0 conditions with hundreds of millions in value secured (about $0.21 billion and $0.30 billion on L2BEAT on November 9, 2026) — but it is an explicit risk factor that should be consciously evaluated. A Stage 0 protocol sits closer to a trusted service than a trustless infrastructure layer in its actual security model. For institutional capital with compliance requirements, this distinction is increasingly material, particularly as on-chain governance standards mature and counterparty risk assessments of DeFi infrastructure become more rigorous.

The absence of any Stage 2 network among the major L2s covered here deserves emphasis as a systemic observation. Every major L2 in production today retains some form of administrative capability over the protocol — a security council with upgrade keys, time-locked admin overrides, or governance systems with emergency powers. Achieving Stage 2 requires eliminating all such override mechanisms while maintaining enough protocol agility to address critical vulnerabilities — an unsolved governance engineering problem. The first major L2 to credibly reach Stage 2 will set a new trust standard that the rest of the ecosystem will face sustained pressure to match.

Ecosystem Depth: Superchain, Stylus & Institutional ZK

Ecosystem depth — the breadth of protocols, developer tooling, user primitives, and institutional integrations available on a network — is increasingly the variable that distinguishes networks with similar fee profiles and security stages. In 2026, three distinct ecosystem development strategies stand out. The OP Stack Superchain model connects Base, Worldcoin's network, and other OP Stack chains that share bridging protocols, governance systems, and development tooling, so liquidity and users can move across participating chains [1]. BlockEden cites a Messari projection that Superchain integration could lift Optimism's TVL by 40–60% during 2026 [1]. Arbitrum pursues a different angle — expanding the developer surface via Stylus to attract systems programmers — while zkSync targets institutional and regulated DeFi deployments where ZK proof finality satisfies compliance settlement requirements.

Arbitrum Stylus is a major expansion of the developer surface on an L2. By enabling Rust, C, and C++ smart contracts compiled to WebAssembly and executed alongside the existing EVM runtime, Stylus opens Ethereum L2 development to a population of systems engineers who previously had no viable deployment path to on-chain infrastructure [1]. The 147 submissions to the Stylus Sprint developer program signal early-stage momentum in that developer community [1]. WebAssembly contracts also unlock performance characteristics suited for computationally intensive applications — sophisticated derivatives pricing, real-time on-chain gaming physics, and AI model inference — that would be economically unviable in Solidity.

On the institutional ZK front, Matter Labs' zkSync Era was selected for Deutsche Bank's Project Dama 2, a consortium of 24 financial institutions testing regulated asset tokenization on ZK rollup infrastructure [1]. This deployment pattern — using ZK proof finality to satisfy proof-of-settlement requirements for regulated securities — is a preview of where institutional DeFi infrastructure is heading. ZK rollups' ability to generate verifiable mathematical settlement proof within hours aligns naturally with traditional finance clearing and settlement timelines. BlockEden presents Project Dama 2 as evidence that when compliance, auditability, and privacy must coexist, zero-knowledge proofs aren't optional [1].

Starknet's Cairo-native account abstraction is a user-experience feature in its own right. Starknet's documentation says it supports account abstraction natively, and names social recovery — retrieving a lost wallet through a selected set of contacts — as one example of what account contracts can program. For consumer-oriented applications targeting non-technical users, this is a genuine competitive advantage. The trade-off — Cairo's non-EVM language limiting dApp portability and constraining the addressable developer pool — remains the primary ceiling on Starknet's broader adoption trajectory against EVM-compatible alternatives.

Which Layer 2 Fits Your Strategy? A Decision Framework

Selecting an Ethereum Layer 2 in 2026 is primarily an ecosystem alignment and risk calibration decision, not a fee or throughput decision. Post-Dencun, fee differences between major networks are measured in cents, making them irrelevant for most trading strategies. The meaningful selection criteria are: liquidity depth in the specific protocol category you intend to use, withdrawal finality speed relative to your capital management requirements, smart contract language compatibility for developers, and L2BEAT security stage relative to your risk tolerance for governance-level protocol exposure. For a retail trader deploying capital in established DeFi markets — perpetuals, structured lending, yield optimization, or spot liquidity provision — Arbitrum One remains a strong option for its long-established DeFi deployments such as GMX, Uniswap and Aave [1], but it no longer holds the most capital: on November 9, 2026, Base led on both L2BEAT's value-secured measure ($13.8 billion vs $9.1 billion) and DefiLlama's DeFi TVL ($5.8 billion vs $1.4 billion), so compare liquidity for the specific protocol and market you plan to use. The concentration of blue-chip DeFi protocols on Arbitrum One is not accidental — it is the compounding result of first-mover scale, which attracted the first wave of DeFi migrations, which attracted liquidity and users, which attracted further protocol deployments in a self-reinforcing cycle.

For users prioritizing activity-rich consumer applications or Coinbase ecosystem integration, Base is the clear answer. Its 12.89 million daily transactions and 382,500 daily users were the highest of the networks in CoinBureau's February 20, 2026 snapshot, and Coinbase's distribution infrastructure remains a structural advantage for onboarding non-crypto-native users to on-chain activity [3]. Base's lower value secured relative to Arbitrum One in CoinBureau's February 2026 data reflected its use-case mix — consumer applications generate volume with smaller per-user capital commitment than DeFi protocols — not inferior infrastructure quality, and by November 9, 2026, Base led Arbitrum One in value secured on L2BEAT ($13.8 billion vs $9.1 billion). Both networks are Stage 1, and both averaged under 2 cents per transaction in CoinBureau's February 2026 table [3]; the differentiator is protocol ecosystem, not network fundamentals.

According to CoinBureau's Layer 2 analysis for 2026: once most rollups become “cheap enough,” fees stop being the deciding factor for everyday users. The differentiators move up the stack.

For ZK-focused use cases, the selection fork divides on EVM compatibility versus proof-system properties. zkSync Era and Scroll both support Solidity — Scroll documents only a short list of EVM differences — enabling Solidity teams to migrate with limited friction. zkSync Era has institutional validation via Project Dama 2 [1]. Starknet is the rational choice when post-quantum security and native account abstraction are prioritized over ecosystem breadth, accepting Cairo language proficiency as a prerequisite. For traders or protocol teams who require ZK proof finality certainty without moving to a non-EVM smart contract language, zkSync Era and Scroll remain the Solidity-friendly ZK options, but as of November 9, 2026, L2BEAT rates both Stage 0 (Scroll was Stage 1 in CoinBureau's February 2026 table [3]), so neither is yet a mature, low-trust choice; among the ZK rollups covered here, only Starknet is at Stage 1.

Frequently Asked Questions

What is the core difference between optimistic and ZK rollups?

Optimistic rollups assume all submitted transactions are valid by default and rely on a fraud-proof challenge window before native withdrawals finalize on Ethereum mainnet — about 6.4 days on Arbitrum One, 7 days on OP Mainnet and 5 days on Base (1 day when both TEE and ZK proofs back the same proposal). Any network participant can dispute an invalid transaction during that window; if no successful challenge occurs, the state is accepted as final. ZK rollups generate a cryptographic validity proof for every transaction batch — a mathematical certificate proving the batch is correct — and submit it to Ethereum for immediate on-chain verification. Withdrawals finalize once that proof is verified, with no challenge period; how long that takes depends on each network's proof cadence and security delay. Both architectures cut fees well below Ethereum mainnet's; CoinBureau's February 2026 table put average fees on all seven networks it compared under 4 cents [3]. The core practical trade-off is EVM compatibility (straightforward for optimistic rollups, historically constrained for ZK) versus finality speed and cryptographic settlement certainty (inherent in ZK rollups). For Solidity-based ZK rollups the compatibility gap has narrowed — Scroll, for example, documents only a short list of EVM differences — making the finality advantage of ZK networks more relevant for capital allocation decisions.

Which Ethereum Layer 2 has the lowest fees in 2026?

It depends on the measure. In CoinBureau's February 20, 2026 comparison table, OP Mainnet had the lowest average fee ($0.0007), followed by Arbitrum One ($0.0044) and Scroll ($0.0068); Base averaged $0.0161 [3]. For a simple ETH transfer on the L2Fees board that CoinBureau cites, zkSync Era was the cheapest of the networks CoinBureau quotes from it, at about $0.07, against $0.09 on Arbitrum One and OP Mainnet and $0.19 on Starknet; Base was not listed [3]. In practical terms, the spread among the networks CoinBureau quotes from that board is about 12 cents at most. For most retail trading strategies, this difference is negligible. Fees were a meaningful differentiator before EIP-4844; post-Dencun, ecosystem depth, security stage, and withdrawal mechanics are more consequential selection variables for most users than the marginal cent-level fee differences between networks.

What does L2BEAT Stage 1 vs Stage 0 actually mean for my funds?

A Stage 0 classification means the network's smart contracts can be upgraded by a centralized set of private key holders — typically a multi-sig wallet controlled by the core team — without requiring on-chain governance approval. If that key set were compromised or coerced, the protocol's rules could be altered unilaterally, potentially affecting user funds with no recourse mechanism. As of November 2026, L2BEAT lists zkSync Era, Linea, and Scroll at Stage 0. Stage 1 means that fraud or validity proofs are active on-chain and upgrade keys are subject to meaningful governance constraints — such as a security council with time-locked overrides or a multi-sig requiring supermajority approval with a defined delay period. Arbitrum One, Base, OP Mainnet, and Starknet are at Stage 1 (Scroll was Stage 1 in CoinBureau's February 2026 table [3]). Stage 2 — closer to being controlled by code rather than humans [3] — has not been reached by any network covered here. The practical framing: Stage 0 requires trusting the team; Stage 1 requires trusting the governance system; Stage 2 would require trusting only the code.

Why does Base now hold more value than Arbitrum One?

TVL — Total Value Locked — measures the capital deposited in DeFi protocols: lending pools, automated market maker liquidity positions, derivatives vaults, and yield strategies. L2BEAT's total value secured (TVS) is broader: it counts all assets bridged to or minted on the chain. CoinBureau's February 20, 2026 snapshot had Arbitrum One leading with $16.88 billion in TVS [3], reflecting the depth of its established DeFi ecosystem — GMX, Uniswap, and Aave [1] — where a single perpetual position or lending supply might represent tens of thousands of dollars, while Base, with 12.89 million daily transactions, was listed in the same table as best for social and consumer apps [3]. On L2BEAT, Arbitrum One was still slightly ahead at the end of May 2026 ($13.6 billion vs $13.1 billion), but by November 9 Base led $13.8 billion to $9.1 billion. DefiLlama, which counts only DeFi deposits, has shown Base ahead since at least May 2026 ($4.4 billion vs $1.5 billion on May 31; $5.8 billion vs $1.4 billion on November 9). Base also still processes far more transactions — about 8.0 million a day against Arbitrum One's 1.1 million (7-day average, L2BEAT). Transaction volume and TVL measure different things — frequency of activity versus concentration of capital — and a network can lead on both, as Base now does.

How do I move funds off an optimistic rollup without waiting out the challenge window?

Third-party liquidity bridges bypass the native challenge window (about 6.4 days on Arbitrum One, 7 days on OP Mainnet, 5 days on Base) by fronting liquidity on Ethereum mainnet immediately, then recouping the bridged funds from the rollup once the challenge window expires. Across Protocol and Stargate Finance are two such bridges for Arbitrum One, Base, and OP Mainnet withdrawals. The fee and settlement time vary by current bridge utilization, network congestion, and the token being bridged, so check the quote before confirming. Compare that fee with the cost of waiting out the native challenge window. Key risk factors to evaluate before using a bridge: the bridge's smart contract audit history and security track record, liquidity depth for the specific token and destination (large withdrawals may face slippage), and whether the bridge is routing through additional networks that introduce additional smart contract exposure.

What's Next for Ethereum Layer 2

The 2026 Layer 2 landscape reflects a market that has passed through its fee-competition phase and entered an infrastructure maturity phase. EIP-4844 resolved the cost problem; the remaining competitive dimensions — decentralization stage, ecosystem depth, ZK proof system quality, and cross-chain interoperability — are more complex and slower to resolve than a single protocol upgrade. The race to Stage 2 governance is the defining technical-political narrative: every major network must eventually remove its administrative training wheels without triggering a security incident in the process. The first major network to credibly achieve Stage 2 while maintaining production stability will set a new trust standard that the rest of the ecosystem will face pressure to match.

For retail traders, the actionable framework is straightforward: identify the protocols you intend to use, confirm they are deployed on a specific network, verify that network's current L2BEAT security stage, and understand its withdrawal mechanics before committing capital. Arbitrum One for long-established DeFi protocols (though Base now holds more DeFi capital, so check pool depth); Base for consumer-scale activity and Coinbase-native integration; Scroll or zkSync Era for ZK finality without non-EVM language overhead, keeping in mind both are Stage 0 on L2BEAT as of November 9, 2026; Starknet for post-quantum proof assurance and account abstraction capabilities. As ZK proof generation costs continue to fall and batch frequencies increase, the throughput gap between ZK and optimistic rollups could narrow, shifting the selection criteria further toward ecosystem composition and governance maturity rather than underlying architecture type.

The institutional ZK trajectory — exemplified by Deutsche Bank's Project Dama 2 with 24 participating financial institutions [1] — signals the next growth vector beyond retail DeFi. Regulated asset tokenization at scale requires the cryptographic settlement finality that ZK rollups uniquely provide. As compliance frameworks for on-chain tokenized assets mature across major jurisdictions through 2026 and 2027, the value secured on ZK rollups could grow faster, narrowing the gap with optimistic rollups.

Last updated: 2026-11-09. Unless otherwise dated, value-secured, fee, and throughput figures are February 2026 data from CoinBureau and BlockEden Research; those figures and the February 2026 stage table have not been refreshed, while the comparison table in the fees section uses dated ETHBEAT (formerly L2BEAT) figures from November 2026. On 2026-11-09, L2BEAT showed about $36.4 billion in total value secured across the Layer 2 projects it tracks, with Base ($13.8 billion) ahead of Arbitrum One ($9.1 billion) — verify current figures at ETHBEAT (formerly L2BEAT, ethbeat.org) before making capital deployment decisions.