Base, Coinbase's Ethereum Layer 2 network, activates its Azul upgrade on mainnet May 13, 2026. The upgrade introduces a multiproof architecture that combines trusted execution environment (TEE) proofs with zero-knowledge (ZK) proofs, compressing withdrawal finality from the standard 7-day optimis...
"L2s' progress to stage 2 (and, secondarily, on interop) has been far slower and more difficult than originally expected." — Vitalik Buterin, Co-founder, Ethereum
Base, Coinbase's Ethereum Layer 2 network, activates its Azul upgrade on mainnet May 13, 2026. The upgrade introduces a multiproof architecture that combines trusted execution environment (TEE) proofs with zero-knowledge (ZK) proofs, compressing withdrawal finality from the standard 7-day optimistic rollup challenge window to approximately one day when both proof systems agree. It is the first production deployment of a dual-prover system on an optimistic rollup at scale.
The upgrade arrives as Base processes 12.89 million daily transactions and holds $11.2 billion in TVL — second among all Ethereum L2s and commanding approximately 28–33% of total L2 market share. Coinbase reported $185,291 in average daily sequencer revenue from Base in Q1 2026, with priority fees constituting 86.1% of that figure. Azul does not alter Base's fee economics but restructures its proof infrastructure and client architecture.
No Ethereum L2 has reached Stage 2 decentralization as classified by L2BEAT. Azul positions Base to move from Stage 1 — achieved in late 2025 with the introduction of permissionless fault proofs — toward Stage 2, where no single entity can override the proof system except in defined bug-resolution scenarios.
Azul's core technical contribution is a multiproof framework that runs two independent proof systems in parallel. The first is a TEE-based prover, which generates attestations inside a hardware-isolated execution environment. The second is an SP1 zero-knowledge prover, developed in collaboration with Succinct Labs, which produces cryptographic validity proofs.
Either proof type can finalize a state proposal independently. This redundancy eliminates single-point-of-failure risk in the validation pipeline. An attacker would need to simultaneously compromise both the TEE hardware enclave and the ZK proof circuit — two fundamentally different cryptographic systems — to manipulate withdrawal operations.
When the two systems disagree, the framework defaults to the permissionless ZK proof over the permissioned TEE proof. This hierarchy ensures that the system cannot be subverted by a compromised or manipulated hardware enclave alone. The on-chain dispute resolution occurs automatically through smart contract logic, with no manual intervention required from a security council.
The multiproof approach differs from single-prover optimistic rollups (Arbitrum One, OP Mainnet) which rely solely on fraud proofs submitted by third-party watchers. It also differs from pure ZK-rollups (zkSync Era, Starknet) which rely exclusively on validity proofs. Base's hybrid model uses ZK proofs not to replace the optimistic mechanism, but to provide a second, independent confirmation layer.
The withdrawal time improvement is the most user-facing consequence of Azul. Standard optimistic rollups impose a 7-day challenge period on native withdrawals to Ethereum L1. This window exists to give honest watchers time to detect and submit fraud proofs against invalid state transitions. On Arbitrum, the effective period is approximately 6.4 days.
Users have historically circumvented this delay through third-party liquidity bridges — services like Across Protocol, Stargate, or Hop Protocol that front funds in exchange for a fee, typically completing transfers in under 5 minutes. These bridges charge spreads ranging from 0.05% to 0.30% per transaction.
Azul compresses native withdrawal finality to approximately one day when both the TEE and ZK provers agree on the same state root. The mechanism is straightforward: if two independent proof systems attest to the same state, the confidence threshold for finality is substantially higher, and the required challenge window shrinks accordingly. This reduces reliance on third-party bridges for users willing to wait 24 hours rather than pay bridge fees.
For context, ZK-rollups like Starknet and zkSync Era achieve finality in approximately 1–12 hours depending on proof generation time and L1 posting frequency. Azul narrows the gap between optimistic and ZK-rollup withdrawal experiences without requiring Base to abandon the OP Stack's optimistic architecture entirely.
Azul introduces several backend infrastructure changes alongside the multiproof system.
Execution client consolidation. Base-reth-node becomes the sole execution client, replacing the previous multi-client configuration. A new consensus client, base-consensus, derived from the Kona project, handles block ordering and validation. The Base team has indicated plans to merge both into a single binary in a future release.
Empty block reduction. Over the two months preceding the upgrade, empty blocks on Base fell 99%, from approximately 200 per day to around two. This reflects improved block production efficiency and transaction packing.
Throughput capacity. Base has sustained multiple transaction bursts of up to 5,000 transactions per second during testing. The network currently averages 12.89 million daily transactions and 382,500 daily active users.
Ethereum compatibility. Azul aligns Base with Ethereum's Osaka execution-layer specifications, adopting six EIPs including transaction gas caps of approximately 17 million, a new CLZ opcode, and MODEXP adjustments. Most existing decentralized applications require no modifications.
Security audit. An Immunefi audit competition ran from April 21 through May 4, 2026, with a maximum reward of $250,000 for critical vulnerabilities.
The L2BEAT framework classifies rollup maturity across three stages. Stage 0: the chain depends on a centralized operator and users cannot exit without sequencer cooperation. Stage 1: permissionless fraud or validity proofs are live and users can independently exit to Ethereum. Stage 2: the system is fully autonomous — no group of actors can post an invalid state root except under defined bug-resolution procedures.
As of May 2026, four Ethereum L2s have achieved Stage 1: Arbitrum One, Base, OP Mainnet, and Starknet. Zero have reached Stage 2.
Vitalik Buterin acknowledged this deficit in February 2026, stating that L2 decentralization progress had been "far slower and more difficult than originally expected." He proposed that native rollup precompiles — where Ethereum itself verifies ZK-EVM proofs as part of the protocol — could accelerate the path. Under this model, if the precompile contains a bug, Ethereum would hard-fork to fix it, removing the need for L2s to maintain their own security councils.
Base's current governance structure includes a 10-member security council drawn from independent entities. Any software upgrade requires 75% consensus among the council, Base, and Optimism. Azul does not remove the security council, but it provides the on-chain infrastructure — multiproofs — that could eventually allow the council's role to be reduced or eliminated entirely.
The Base team has stated that Azul enables the security council to "provably detect and adjudicate bugs via the multi-proof system," shifting the council's function from a trust-based override to a verification-based mechanism.
Base's economic significance extends beyond its technical architecture. Coinbase reported Q1 2026 revenue of $1.41 billion, missing analyst estimates of $1.52 billion, with a net loss of $394.1 million. Transaction revenue fell to $755.8 million versus $805.2 million expected. Total revenue declined 31% year-over-year.
Against this backdrop, Base remained a relative bright spot. The network averaged $185,291 per day in sequencer revenue during Q1, with priority (tip) fees representing 86.1% of that total. On an annualized basis, this equates to approximately $67.6 million in sequencer revenue.
In 2025, Base captured 62% of total L2 revenue — $75.4 million out of $120.7 million across all Ethereum L2s. Applications deployed on Base generated $369.9 million in protocol revenue during 2025. The network benefits from direct distribution access to Coinbase's 9.3 million monthly active users, a channel no competing L2 can replicate.
Average transaction fees on Base stand at approximately $0.05, with a median of $0.02 — the lowest among the top eight L2s. Arbitrum and OP Mainnet average $0.08–$0.09 per transaction. These fee levels were enabled by EIP-4844 (the Dencun upgrade, March 2024), which reduced L2 data-posting costs to Ethereum by 80–90%.
From an economic-value perspective, Base is among a small number of L2s that extracts operational revenue exceeding its on-chain costs — a distinction relevant in an ecosystem where 85–90% of value flows remain subsidy-driven.
The Ethereum L2 market has consolidated around three networks. Base, Arbitrum, and OP Mainnet collectively account for approximately 90% of all L2 transactions. TVL distribution as of late April 2026:
| Network | TVL | Market Share | Stage | |---------|-----|-------------|-------| | Arbitrum One | $14.9–16.9B | 40–44% | Stage 1 | | Base | $10.7–11.2B | 28–33% | Stage 1 | | OP Mainnet | $5.6B | ~15% | Stage 1 | | Starknet | — | — | Stage 1 | | zkSync Era | — | — | Stage 0 |
Azul's multiproof system gives Base a structural differentiation. No other optimistic rollup runs dual provers in production. However, this advantage is time-limited: Optimism's OP Stack, which Base itself uses, will likely adopt similar multiproof capabilities for the broader Superchain ecosystem. Arbitrum has signaled a timeboost MEV auction for sequencer slots and plans a multi-party sequencer for late 2026.
The competitive dynamic is not purely technical. Base's roadmap includes three more upgrades in 2026: a performance-focused release by end of June, a UX-focused release by end of August with native account abstraction, and the launch of Base Vibenet (a public devnet) in mid-May. The cadence is aggressive relative to competitors.
ZK-rollups maintain a structural advantage in finality speed. Starknet and zkSync Era achieve proof-based finality in 1–12 hours. Azul narrows this gap but does not close it. The trade-off: optimistic rollups retain broader EVM compatibility and lower proof generation costs.
Azul is an infrastructure upgrade, not a product launch. It restructures how Base validates state transitions, adding a second proof system that increases security redundancy and compresses withdrawal timelines. The multiproof architecture is technically sound: requiring simultaneous compromise of two independent cryptographic systems raises the attack surface substantially.
The upgrade does not solve the L2 decentralization problem. No Ethereum rollup operates at Stage 2, and the path from Stage 1 to Stage 2 remains undefined for most networks. Azul provides necessary infrastructure — on-chain proof verification that could eventually replace trust-based security councils — but that transition requires further governance changes that are not part of this release.
Economically, Base operates as one of the few L2s generating net-positive sequencer revenue, a distinction that matters in an ecosystem where the vast majority of blockchain networks remain subsidy-dependent. Whether that revenue scales sufficiently to justify Coinbase's ongoing investment in Base infrastructure — particularly during a quarter when the parent company posted a $394 million loss — remains an open question.
The multiproof model will be tested under adversarial conditions once mainnet goes live. Until then, Azul's impact is theoretical. The reduction from 7-day to 1-day withdrawals is meaningful for capital efficiency. Whether it changes user behavior at scale depends on bridge economics and user price sensitivity to the remaining 24-hour delay.