With Fees Below $0.10, L2 Security Stage Now Decides

The $48B L2 Ecosystem: What Shifted Between 2025 and 2026

The Ethereum Layer 2 landscape entered 2026 as a fundamentally different market than it was twelve months earlier. As of May 2026, 73 active rollups [1] collectively secure more than $48 billion in total value locked — roughly double the combined L2 TVL recorded in early 2025 [1]. This growth did not happen in a vacuum. The Dencun upgrade (EIP-4844), activated on Ethereum mainnet in March 2024 [3], introduced blob transactions — a dedicated data-availability lane that cut L2 data-posting costs to Ethereum by 80–90% [1]. The immediate result: sub-$0.10 fees became universal across every major network, stripping cost leadership as a viable differentiator among competing chains. What replaced fee competition as the defining metric are three pillars — security stage, raw throughput, and ecosystem depth — that now determine how capital and users allocate across the L2 field.

Quick Answer: As of May 2026, 73 Ethereum Layer 2 rollups collectively hold over $48B in TVL — double their early-2025 combined total. The Dencun upgrade cut fees 80–90%, making security stage, throughput, and ecosystem fit the dominant factors when selecting an L2 for any DeFi strategy.

The impact on Ethereum mainnet fees has been equally dramatic. According to data tracked by BlockEden, average mainnet gas dropped from 7.14 gwei in January 2025 [3] to approximately 0.50 gwei by January 2026 [3] — a 93% decrease that benefited all rollups simultaneously by lowering their settlement overhead. When every major L2 operates below a tenth of a dollar per transaction, competing on price alone becomes structurally unviable.

The most consequential behavioral shift is the bifurcation of the user base. Base dominates retail transaction volume, processing 12.89 million daily transactions in February 2026 [1] through Coinbase's consumer distribution pipeline. Arbitrum One, by contrast, concentrates institutional-scale and large-wallet DeFi — a pattern visible in its $13.8–16.9B TVL despite lower daily user counts. This divergence shapes every use-case recommendation in this guide. Knowing which chain fits your strategy requires understanding how each network built its specific structural advantage.

The networks that have built genuine advantages in at least two of the three key dimensions — security stage, throughput, and ecosystem depth — are pulling ahead of the field. The rest are competing for a shrinking slice of a market that can no longer differentiate on fees alone. This guide maps the data to concrete strategy decisions for each major DeFi use case.

TVL Rankings: Where DeFi Liquidity Concentrates in 2026

Total value locked is the most reliable proxy for where serious capital has chosen to deploy, and in 2026 the distribution across Ethereum L2s is sharply uneven. Arbitrum One leads all L2 networks with approximately $13.8 to $16.9 billion in TVL [1][2], representing 40–44% of all L2 DeFi liquidity. Base holds second position at $10.74 to $11.2 billion [2], a figure that becomes remarkable in context: at the start of 2025, Base held just $2.1 billion in TVL [2]. Together, Arbitrum and Base account for roughly 77% of all L2 DeFi liquidity as of April 2026 [1] — a duopoly that defines the competitive landscape for this strategy cycle.

Base's 18-month TVL trajectory — from $2.1 billion to over $10.7 billion — represents the fastest expansion in L2 history, according to analysis by Eco. The primary driver is Coinbase's direct consumer distribution: an estimated base of over 100 million Coinbase account holders provides a built-in onboarding funnel that no other L2 operator can replicate organically. The Aerodrome DEX and Morpho lending protocol anchor Base's DeFi stack, supported by $3.9 billion in stablecoin float [2].

Arbitrum's TVL dominance reflects a different origin story: four years of iterative protocol deployment, deep integrations with Aave v3, Uniswap v3, GMX, Curve Finance, and Pendle, and $4.2 billion in stablecoin liquidity [1] that gives large-position traders the slippage characteristics they require. The 2024 launch of Arbitrum Stylus extended developer support to Rust, C++, and WebAssembly, attracting 147 hackathon submissions with 17 selected projects [3] — broadening the developer base well beyond EVM-native teams.

The third tier — OP Mainnet, zkSync Era, and Linea — holds meaningful TVL but occupies a structurally different competitive position. OP Mainnet's $5.6 billion [2] is backstopped by its role as the OP Stack reference chain and its Superchain relationship with Base and World Chain. zkSync Era at $4.1 billion [2] and Linea at $3.4 billion [2] serve use cases where ZK architecture provides finality advantages over optimistic competitors. Data from DefiLlama confirms the tiered structure across the full rankings.

Sources: Coin Bureau, Eco, DefiLlama (April 2026). Market share figures are approximate and reflect a $48B total L2 TVL baseline.

Security Stages Explained: Why L2BEAT Scores Drive Capital Decisions

The L2BEAT security stage framework has become the de facto standard for evaluating trust assumptions on Ethereum rollups — and in 2026 it is the metric that most directly determines how institutional capital and large-wallet DeFi participants make deployment decisions. Stage 0 designates networks where trusted operators retain control of upgrade keys, meaning the system's security ultimately depends on the honesty and competence of a centralized team rather than cryptographic or economic enforcement. Stage 1 confirms that permissionless fraud proofs (for optimistic rollups) or validity proofs (for ZK rollups) are live in production, so any participant can verify and challenge the network's integrity without trusting any single operator. For DeFi users allocating six or seven figures, this distinction determines whether protocol risk is cryptographically bounded or operationally contingent on the operator team acting in good faith.

As of 2026, the Stage 1 cohort among major L2s includes Arbitrum One, Base, OP Mainnet, Starknet, and Scroll [1]. This classification confirms that each of these networks has live, permissionless mechanisms to verify transaction validity without relying on operator honesty. zkSync Era and Linea remain at Stage 0 [1], meaning their governance and upgrade controls still rely on centralized assumptions — a meaningful risk variable for large positions, regardless of their ZK architecture's mathematical properties.

Arbitrum One occupies a distinct position even within the Stage 1 cohort. It is currently the only major L2 offering fully permissionless fraud proofs in production [2] — meaning any participant can submit a fraud challenge without requiring permission from the Arbitrum DAO or Offchain Labs. This represents the most decentralized trust model available among active L2s, and it partially explains why large-position DeFi capital concentrates on Arbitrum despite Base's higher user activity numbers.

"Stage 1 is the minimum bar for a network where you're comfortable holding significant capital without an emergency exit plan. Stage 0 is appropriate for experimentation, but the upgrade-key risk is real and quantifiable — it does not disappear because the underlying math is sound." — Coin Bureau L2 Security Analysis, April 2026

The practical implication for DeFi strategy is direct: for positions above a threshold where smart-contract risk becomes material — typically $50,000 and above in institutional-grade risk frameworks — deploying on a Stage 0 network means accepting that the protocol's security guarantees can be altered by a centralized party without on-chain challenge. This is not a hypothetical risk; it is the explicit design of Stage 0 systems. zkSync Era and Linea are both targeting Stage 1 upgrades, but governance decentralization is measured in months, not days, and timelines remain unconfirmed as of May 2026.

For retail traders operating at smaller scale, Stage 0 networks are operationally viable. The actual probability of a malicious upgrade is low, and the fee and throughput characteristics may justify the tradeoff in many contexts. The key is making that tradeoff consciously, with accurate understanding of what Stage 0 actually means rather than treating all L2 security claims as equivalent.

Fees and Throughput: The Post-Dencun Comparison

Fee comparison across Ethereum L2s in 2026 is a more nuanced exercise than it was in 2023 or 2024, precisely because the gaps have narrowed to near-irrelevance for most DeFi participants. Following the Dencun upgrade's blob transactions, every major network now operates below $0.10 per transaction, and the median figures are compressed enough that fee differences between chains rarely justify switching networks for a given strategy. Median per-transaction fees in Q1–Q2 2026 are: Base at $0.02, OP Mainnet at $0.03, Arbitrum One at $0.04, zkSync Era at $0.05, and Scroll at $0.06 [2]. For high-frequency traders executing dozens of transactions daily, even a $0.04 difference compounds meaningfully — but for most DeFi participants, throughput ceiling is the binding constraint, not unit cost.

Sources: Eco, Coin Bureau (Q1–Q2 2026). Daily active user figures from February 2026 data snapshots.

On throughput, Base leads all L2s at 89 TPS peak capacity [2], followed by Arbitrum One at 40–62 TPS across its Nitro stack architecture [1] and World Chain at 48 TPS. OP Mainnet sustains 34 TPS. The ZK rollup cohort trails significantly: zkSync Era achieves 28 TPS, Linea 22 TPS, Starknet 19 TPS, and Scroll 14 TPS [1] — each constrained by the computational overhead of generating validity proofs at scale. This is an engineering tradeoff, not a fundamental architectural flaw. ZK proof generation is computationally intensive, and the hardware acceleration required to match optimistic rollup TPS is an active research and development priority across the ZK ecosystem.

For DeFi traders, the throughput gap matters primarily in two scenarios: high-frequency perpetuals trading (where queue latency during peak demand periods affects execution quality) and protocol launches that generate sudden transaction spikes. In both cases, Base's 89 TPS headroom provides meaningful buffer over OP Mainnet's 34 TPS. For lending, yield farming, and standard DEX trading, the practical difference between 34 and 89 TPS is rarely felt at the individual user level — network demand relative to capacity, not absolute capacity in isolation, determines the experience.

Optimistic vs ZK Architecture: What It Means for DeFi Withdrawals

Architecture choice between optimistic and ZK rollups does not meaningfully affect day-to-day DeFi operations for most users — but it becomes highly relevant the moment you need to move large capital back to Ethereum mainnet. Optimistic rollups, including Arbitrum One, Base, and OP Mainnet, assume transaction validity by default and apply a 7-day fraud-proof challenge window before native bridge withdrawals settle to Ethereum L1 [1][4]. ZK rollups — including zkSync Era, Starknet, Linea, and Scroll — generate cryptographic validity proofs that mathematically confirm each transaction batch, enabling withdrawal finality on L1 in under one hour [1]. The 7-day window is not daily friction for most traders, but it creates a structural liquidity management consideration for anyone running large positions or needing rapid capital reallocation.

Third-party liquidity bridges — primarily Across Protocol and Stargate Finance — effectively solve the 7-day window for the majority of retail use cases. These bridges pre-fund withdrawals from their own liquidity pools and settle with the network afterward, completing cross-chain transfers in under 5 minutes for a fee of 0.05–0.20% [1][4]. The practical constraint is bridge liquidity depth: for transfers significantly above $5–10 million in a single transaction, bridge liquidity pools may be insufficient to complete the transfer at the quoted fee, forcing either partial execution or use of the native bridge with the full 7-day delay. This threshold effectively differentiates institutional from retail bridge accessibility.

"Deutsche Bank's Project Dama 2 leverages zkSync infrastructure with participation from 24 financial institutions [3] — specifically because ZK proof architecture provides cryptographic transaction finality that aligns with regulatory proof-of-settlement requirements that optimistic rollups cannot satisfy within the same timeframe." — BlockEden Ethereum L2 Analysis, February 2026

The Deutsche Bank signal is meaningful precisely because it validates the institutional thesis around ZK architecture: when settlement finality must be provable to regulators or counterparties in under an hour, the 7-day optimistic challenge window is structurally incompatible with traditional finance compliance frameworks. For retail DeFi users, this distinction rarely surfaces — but it explains why institutional TVL on ZK networks is growing faster than user activity metrics would suggest, and why zkSync Era maintains $4.1 billion in TVL [2] despite its Stage 0 security classification and lower daily user counts.

The architecture choice also carries implications for smart contract compatibility. Optimistic rollup ecosystems offer broad EVM compatibility, reducing porting friction for protocols migrating from Ethereum mainnet. ZK rollups using bespoke virtual machines — particularly Starknet's Cairo VM — may require contract rewrites, limiting available protocol selection for DeFi users. EVM-compatible ZK rollups (zkSync Era, Scroll, Linea) have largely closed this compatibility gap by 2026, but Starknet's native DeFi ecosystem remains smaller partly as a result of this historical friction. For most traders, protocol availability on the target chain matters more than the underlying VM architecture.

Matching Layer 2 to Your DeFi Strategy

The fundamental question in L2 selection is not which network posts the lowest fees — it is which network's specific combination of security stage, throughput, ecosystem depth, and architectural properties best matches your actual use case. In 2026, with fees commoditized and security stages increasingly differentiated, this matching exercise has become more consequential than at any prior point in L2 development. The following framework maps four distinct DeFi profiles to their optimal network choices based on the data examined in this guide, covering the largest capital pools and highest-frequency use cases in the current market.

High-frequency trading and perpetuals: Base or Arbitrum One. For traders executing multiple transactions per session in perpetuals or on-chain order books, Base's 89 TPS ceiling [2] and $0.02 median fees [2] minimize per-trade friction costs, while 382,500 daily active users [1] generate the order flow density needed for tight spreads. For perpetuals specifically, Arbitrum One hosts GMX — the leading perpetuals DEX by open interest — alongside Camelot and Uniswap v3, providing deeper liquidity pools than any competing L2. The choice between Base and Arbitrum for this profile depends primarily on protocol selection: Base for emerging derivatives infrastructure and retail-driven spread, Arbitrum for established institutional-depth perpetuals where open interest and liquidity depth are the primary execution variables.

Large-position DeFi and lending: Arbitrum One, without close competition for positions at scale. Deploying six-figure or larger allocations into lending protocols (Aave v3) or concentrated liquidity (Uniswap v3) requires both deep liquidity — Arbitrum's $4.2 billion in stablecoin float [1] provides the slippage characteristics large trades demand — and the lowest achievable protocol-layer risk. Arbitrum's Stage 1 status and fully permissionless fraud proofs mean that at the cryptographic level, no centralized party can alter the protocol's behavior without triggering an on-chain challenge. For large-capital allocators, this matters significantly more than a $0.02 fee differential per transaction.

Yield farming and new protocol exposure: Base. The Coinbase distribution flywheel continuously onboards new users and capital, creating the user growth trajectory that generates elevated APYs on newly launched protocols in their bootstrapping phase. Aerodrome and Morpho have demonstrated this dynamic: both achieved substantial TVL within months of launch on Base. The tradeoff is that newer protocols carry higher smart-contract risk relative to battle-tested Arbitrum protocols — yield farming on Base inherently trades the security confidence of established infrastructure for early-access growth premium on emerging strategies.

Compliance-sensitive or institutional DeFi: zkSync Era or Starknet. Despite zkSync Era's Stage 0 governance status, its ZK proof architecture provides cryptographic transaction finality in under one hour [1] — a property that aligns with regulatory proof-of-settlement requirements that optimistic rollups cannot satisfy within the same window. The Project Dama 2 participation of 24 financial institutions [3] validates this use case at institutional scale. For compliance-driven allocation, the governance risk of Stage 0 is weighed against the structural impossibility of Stage 1 optimistic networks providing sub-hour settlement finality — a tradeoff that tilts toward ZK architecture when finality proof is the regulatory requirement.

2026 Outlook: Security Upgrades and the Next Inflection Points

The L2 ecosystem's most significant structural development in the next 12–18 months will be the Stage 0-to-Stage 1 transition for zkSync Era and Linea. Both networks have announced governance decentralization roadmaps, but the critical milestone — removing centralized upgrade-key control and making fraud or validity proofs fully permissionless — had not been completed as of May 2026. When these upgrades execute, the security stage playing field will equalize across the major ZK rollups, and ZK networks' finality advantages will carry more weight in capital allocation decisions with governance risk removed. Monitoring the on-chain deployment of permissionless provers on both networks is the single most consequential governance event for DeFi capital allocation in H2 2026.

The OP Stack Superchain architecture presents a second major inflection point. Base and OP Mainnet currently share sequencer infrastructure under the Optimism Superchain framework [5], with World Chain and Zora also building on the same modular stack. A shared sequencer upgrade — enabling atomic cross-chain composability within the Superchain — could reshape how liquidity routes across these networks, potentially creating a unified liquidity layer that competes with the depth Arbitrum One has built independently. The timeline for this upgrade is a key variable to monitor in the second half of 2026 and into 2027.

Scroll, World Chain, and Starknet are each gaining traction in specific verticals that broader-market networks have not fully addressed. Scroll's ZK-native architecture targets privacy-adjacent and high-verification-demand applications; World Chain's identity-gated DeFi model, built on Worldcoin's biometric verification infrastructure, targets Sybil-resistant protocol design; Starknet's Cairo VM provides a performance ceiling for computationally intensive DeFi strategies that EVM-based chains cannot match at equivalent scale. Each represents a potential capital migration inflection point if its target use case achieves broader adoption in 2026.

The primary risk vector across the 2026 landscape remains Stage 0 governance. Networks where centralized teams retain upgrade-key control are exposed to a specific tail risk: a governance failure, regulatory action, or security breach that forces an emergency upgrade could alter protocol behavior in ways users cannot predict or challenge. This risk is operator-trust-dependent regardless of underlying ZK proof quality — a mathematically sound validity proof does not prevent a team from deploying a malicious contract upgrade if governance controls permit it. Treating Stage 0 and Stage 1 networks as equivalently secure based on ZK architecture alone is the most common analytical error in current L2 evaluation.

Frequently Asked Questions

What is an L2 security stage and why does it matter for DeFi?

The L2 security stage is a classification defined by the L2BEAT framework that describes how much a rollup network relies on trusted operators versus cryptographic enforcement. Stage 0 means a centralized team controls upgrade keys — the network's security guarantees can be altered by the operator without users being able to challenge the change on-chain. Stage 1 means permissionless fraud proofs (optimistic rollups) or validity proofs (ZK rollups) are live in production, so anyone can verify and challenge the network's behavior without trusting the operator. For DeFi users, this distinction determines whether capital safety is cryptographically enforced or operationally contingent on the team not acting in bad faith. For large positions — typically above $50,000 — Stage 0 means accepting operator-dependent protocol risk rather than bounded cryptographic risk. In 2026, Arbitrum One, Base, OP Mainnet, Scroll, and Starknet have achieved Stage 1; zkSync Era and Linea remain at Stage 0 [1].

Why does Arbitrum have far more TVL than Base despite Base having more daily active users?

TVL measures capital deployed, not transaction count or user volume — and these metrics reflect fundamentally different user profiles. Arbitrum One's $13.8–16.9 billion TVL [1] reflects institutional-scale and large-wallet DeFi activity: protocols like Aave v3, GMX, Curve Finance, and Pendle attract multi-million-dollar position deployments from funds and sophisticated traders. Base's 382,500 daily active users [1] primarily represent retail-scale Coinbase consumer transactions — many of which are small, frequent, and individually low-value. A single large lending position on Arbitrum represents more TVL than thousands of retail swaps on Base. The two metrics are not in conflict; they describe different capital profiles on different chains, each dominant in its own segment and serving different strategic purposes.

How does the 7-day withdrawal window affect DeFi on optimistic rollups?

Native bridge withdrawals from optimistic rollups (Arbitrum, Base, OP Mainnet) to Ethereum mainnet require a 7-day challenge window before settlement — this is a fundamental property of the fraud-proof mechanism, not a temporary technical limitation. For most retail DeFi users, this is rarely a practical constraint because third-party liquidity bridges (Across Protocol, Stargate Finance) complete cross-chain transfers in under 5 minutes for a fee of 0.05–0.20% [1], pre-funding withdrawals from their own liquidity pools. The 7-day window becomes a real operational constraint only for very large single withdrawals — typically above $5–10 million — where bridge liquidity pools may be insufficient to fully cover the transfer, forcing use of the native bridge and accepting the delay. For institutional players managing large exits or requiring rapid capital reallocation between chains, this is a genuine structural consideration that ZK rollups (sub-one-hour finality) address by design.

Are ZK rollups better than optimistic rollups for DeFi in 2026?

Neither architecture is universally superior for DeFi in 2026 — each has structural advantages that map to specific use cases. ZK rollups offer faster L1 finality (under one hour versus 7 days natively) and cryptographic validity proofs enabling trustless settlement verification, making them structurally better for institutional DeFi with compliance requirements, frequent large withdrawals, and applications requiring sub-hour settlement proof. However, optimistic rollups currently lead on every practical metric relevant to most DeFi users: TVL depth ($28+ billion combined for Arbitrum and Base alone), ecosystem protocol diversity, TPS (Base at 89 TPS versus Scroll's 14 TPS [2]), daily active users, and security stage parity — Arbitrum and Base are both Stage 1, equivalent to ZK networks Scroll and Starknet. The decision should be based on withdrawal frequency, position size, compliance requirements, and protocol availability on the target chain, not architecture as an abstract preference.

What did the Ethereum Dencun upgrade actually change for Layer 2 users?

The Dencun upgrade, activated on Ethereum mainnet in March 2024 [3], implemented EIP-4844, which introduced a new transaction type called blob transactions. Blobs provide a dedicated data-availability lane for L2 networks to post transaction data to Ethereum at a fraction of the previous cost — reducing L2 data-posting expenses by 80–90% [1]. The immediate effect was that per-transaction fees on every major L2 fell below $0.10, with most major networks now operating between $0.02 and $0.06. The strategic consequence is that fee competition as a differentiator between L2s was effectively eliminated — every major network is now cheap enough that marginal cost differences rarely determine where capital deploys. The decision framework shifted entirely to security stage, throughput capacity, and ecosystem depth as the primary L2 selection criteria. This shift is the core reason this 2026 guide leads with security stage rather than fee benchmarks.

Where the L2 Ecosystem Heads from Here

The 2026 Ethereum L2 landscape is defined by a productive tension: fees have been commoditized by Dencun, but security, throughput, and ecosystem depth are differentiating faster than most market participants anticipated. The result is a market where the right selection framework — matching network properties to use-case requirements — delivers materially better risk-adjusted outcomes than defaulting to the largest TVL or lowest-fee chain. Arbitrum One and Base have built durable advantages in different dimensions — decentralization depth versus distribution scale, respectively — and neither is likely to cede ground in its primary segment within this strategy cycle.

For active DeFi participants, the actionable sequence is straightforward: first, identify your security stage requirement based on position size; second, confirm the protocols you need are deployed and liquid on the target chain; third, validate throughput headroom against your transaction frequency. Fee optimization comes last, not first. This order of operations represents the primary strategic shift from the 2023–2024 era when fee minimization drove most chain selection decisions. The $48 billion TVL distributed across 73 active rollups reflects a market that has already made this transition — the question for any individual allocator is whether their strategy reflects the same logic.

zkSync Era and Linea's pending Stage 1 upgrades remain the single most important milestone to track for capital reallocation decisions in H2 2026. When those upgrades complete, the ZK architecture's finality advantages will combine with Stage 1 security guarantees — creating a compelling framework for rebalancing toward ZK networks for compliance-sensitive and high-withdrawal-frequency strategies. Until then, the Stage 1 combined with deep liquidity that Arbitrum One represents remains the conservative anchor for large-position DeFi allocation in this environment.

Last updated: 2026-05-16. This article reflects on-chain data, L2BEAT security classifications, and fee/throughput figures current as of April–May 2026. L2 security stages are subject to change as governance upgrades are implemented; verify current stage designations directly at L2BEAT before making capital allocation decisions based on security stage.

Related Articles

• Best Ethereum L2 in 2026: Arbitrum vs Base vs Optimism vs zkSync
• Optimistic vs ZK Rollups 2026: Choosing the Right L2 for DeFi
• Layer 2 DeFi Showdown 2026: Security, TVL & Ecosystem Fit
• Ethereum L2 Guide 2026: TVL, Fees & Security Across Major Rollups
• Arbitrum vs Base vs zkSync vs Optimism: Best L2 for DeFi 2026