Explainers
What Base is, and what its on-chain data shows
Base is a rollup on Ethereum, incubated at Coinbase: it executes transactions itself, posts the data to Ethereum, and lets anyone dispute an invalid claim about its state there. It launched in 2023 on Optimism’s OP Stack and in 2026 moved to a stack of its own. A single sequencer orders its transactions, upgrades need the approval of two multisigs, and when checked on 22 November 2026, ETHBEAT, the site l2beat.com now redirects to, rated it Stage 1 under the stages framework L2BEAT introduced.
A rollup incubated at Coinbase
Base’s blog dates the testnet to February 2023, the opening to builders to 13 July 2023 and the public launch to 9 August 2023. A post later that month described Base as an open, permissionless chain incubated at Coinbase and built on the open-source OP Stack in collaboration with the Optimism Collective. What a rollup is, and how the designs differ, is set out in layer 2 explained.
The stack has changed since. In February 2026 Base announced a unified, Base-operated stack: it said it was moving away from the OP Stack, would keep working with Optimism as a client of its support service, and that node operators would follow releases from Base’s own repository rather than Optimism’s. The upgrades page lists the upstream OP Stack hard forks Base adopted, the most recent being Jovian on 2 December 2025, and calls Azul, activated on mainnet on 28 May 2026, Base’s first independent network upgrade. The protocol specification now says Base was originally built on the OP Stack and follows Base’s own specification after Jovian. Much of the inherited design is still visible, down to contract names such as OptimismPortal.
Base is an EVM chain, so its addresses, signatures and account model are Ethereum’s, and an address on Base is the same as on Ethereum (why wallets are chain-specific).
Sequencing: one operator, and a route around it
Base’s protocol overview names three kinds of participant: users, sequencers and validators. The sequencer accepts users’ transactions and the deposits arriving from Ethereum, orders them into blocks and posts the data needed to rebuild those blocks to Ethereum. Validators execute the chain independently from that data, and some act as proposers or challengers of claims about its state.
Base runs a single active sequencer, according to its reliability page, as a high-availability cluster that passes leadership to a follower if the current leader stops producing blocks. It builds a block every 200 milliseconds and, the ordering page says, orders the transactions in each block by priority fee.
A single sequencer can refuse a transaction or stop. The documented route around it runs through Ethereum: a user can send a transaction as a deposit through the OptimismPortal contract, and the derivation rules make a Base block that omits a required deposit invalid. If the sequencer posts nothing at all, nodes wait one sequencing window, 3,600 Ethereum blocks or about twelve hours, and then derive the missing blocks themselves, pending deposits included.
Finality on Base comes in three stages: about 200 milliseconds to inclusion in a sequencer block, about two minutes until the batch containing it is posted to Ethereum, and about twenty minutes until that batch is final on Ethereum.
Fees: execution plus Ethereum’s share
Every Base transaction pays two fees: an execution fee for running it on Base and an L1 fee for the estimated cost of publishing its data to Ethereum, which the documentation says is usually the larger of the two. The execution fee follows Base’s own implementation of EIP-1559, with a base fee that rises and falls with demand and, since the Jovian upgrade, a floor of 0.005 gwei on mainnet. The L1 fee moves with activity on Ethereum.
Gas is paid in ETH, though the fees page also points to a way of paying it in other tokens. Base launched without a network token of its own, and its July 2023 post said it had no plans to issue one; in September 2025 it said it was beginning to explore one. DefiLlama’s snapshot records no token symbol for Base.
How withdrawals are proved
A withdrawal from Base to Ethereum through the canonical bridge takes three steps. It is initiated on Base, proved on Ethereum once the relevant Base state has been posted there, and finalised after a window during which the claim about that state can be disputed. Since the Beryl upgrade the window is five days when one proof backs the claim, or one day when both a TEE proof, produced inside a hardware enclave, and a zero-knowledge proof back the same claim. Waiting for the state to be posted adds about 20 to 60 minutes. Third-party bridges pay out sooner from their own liquidity and do not shorten the canonical window, and ordinary transactions within Base wait for none of this.
Base made fault proofs permissionless in October 2024, so that anyone could take part in disputing state claims, and Azul introduced a multi-proof system in which the same proposal can be backed by one or two proofs before withdrawals rely on it.
Changes to the contracts need two approvals. Base’s Security Council page describes a 2-of-2 arrangement between a 3-of-6 multisig operated by Coinbase and a Security Council, an 8-of-11 multisig of independent entities and individuals; an upgrade takes effect only when both reach their thresholds. The same page dates permissionless fault proofs to October 2024 and the Security Council’s control of upgrades to April 2025, and says Base has since reached Stage 1.
Where it sits in the stages framework
The stages framework grades how far a rollup has moved from control by a few entities, Stage 0, to control by code, Stage 2, with Stage 1 in between. Its page says the idea was proposed by Vitalik Buterin in November 2022 and made precise by the L2BEAT team, which introduced the framework in June 2023; l2beat.com now redirects to ETHBEAT, where the project pages continue. When checked on 22 November 2026, the ETHBEAT page for Base called it a Stage 1 optimistic rollup that passes its walkaway test, meaning users can exit even if operators act maliciously and the Security Council disappears.
The page listed two items missing for Stage 2. Upgrades unrelated to bugs that can be proved on-chain give users less than thirty days to exit, and the Security Council’s actions are not limited to such bugs. It marked the exit window as none, since the contracts can be upgraded at once when both parties approve, and rated sequencer failure as self-sequencing with up to a twelve-hour delay, matching Base’s own figure. The page also says stages are an opinionated assessment of maturity and do not reflect a project’s security.
Documented incidents
Base publishes post-mortems for major incidents on its engineering blog and keeps an incident history on status.base.org.
One came within a month of the public launch. On 5 September 2023 block production stopped for 29 minutes when the Ethereum nodes the sequencer relied on all ran out of disk space (post-mortem). On 21 September 2024 a misconfigured sequencer cluster caused a 17-minute block-building outage: when the active instance became unhealthy, another could not start building (post-mortem). The 5 August 2025 halt lasted 33 minutes, after an automatic handover passed leadership to a sequencer that was still being provisioned and could not hand over again (status page).
Two came in 2026, after Azul’s activation. From 29 to 31 May, withdrawals were delayed because an issue with a TEE enclave halted proposals (status page). On 25 and 26 June, a bug in the sequencer’s block building produced a block no other node would accept, causing outages of 116 and 20 minutes; Base said the chain’s integrity was not compromised and funds were safe (post-mortem).
Each halt in block production in that list traces to the sequencer or the infrastructure around it, which is the dependence the deposit route exists for. At up to twelve hours, though, that route is a backstop against censorship or a long stall, not a way around a short one.
What the on-chain data shows
As of 22 November 2026, per DefiLlama, Base had $4.7B of total value locked, 23.6% lower than thirty days earlier, with 910 protocols listed and $4.8B of stablecoin supply on the chain (data page). The total is the value of assets deposited in DeFi protocols on the chain at the snapshot’s prices, leaving out liquid staking, bridges and deposits DefiLlama flags as double-counted. Price moves in the deposited tokens change it with nobody depositing or withdrawing (what TVL measures).
One protocol leads the list of largest protocols by a wide margin. Morpho Blue, filed under Lending, stands at $3.4B on Base, a figure equal to 72.4% of the chain total (computed), although the list and the total are on different bases. Next come two entries DefiLlama files as risk curators, Steakhouse Financial at $1.2B and Gauntlet at $589.8M; Aave V3 at $428.5M; Grove Finance at $400.4M, filed as an on-chain capital allocator, which DefiLlama describes as deploying capital into, among other things, Aave lending and Morpho vaults; and Uniswap V3 at $256.3M, the largest exchange entry. When one protocol places its deposits in another, the same funds can appear in both entries, one reason the list and the total are not meant to match.
Stablecoin supply, at $4.8B, comes from a separate DefiLlama feed and counts what circulates on the chain whether or not it is deposited anywhere (where stablecoin supply lives).
Base is one of the chains with a hack page. DefiLlama’s record lists 69 incidents that touched Base, 37 of them on Base alone (computed). An incident on several chains counts in full on each chain it touched, so the page’s totals include multi-chain incidents such as exchange hot-wallet thefts; how crypto gets hacked explains the overlap.