Un rollup est un Layer 2 (L2) scaling solution that executes transactions outside the main blockchain (Layer 1) but posts transaction data or proofs back to the Layer 1 chain, inheriting its security guarantees while dramatically increasing throughput and reducing costs. Rollups “roll up” hundreds or thousands of transactions into a single batch that is submitted to the base layer, compressing the data footprint and amortizing the cost of on-chain settlement across all transactions in the batch.
The fundamental insight behind rollups is the separation of execution from consensus and data availability. The Layer 1 blockchain, typically Ethereum, handles consensus and data availability, ensuring that all transaction data is published and that state transitions are valid, while the rollup handles execution, processing transactions at a rate far exceeding what the L1 can achieve natively. This architectural separation allows rollups to achieve thousands of transactions per second while preserving the censorship resistance, decentralization, and finality guarantees of Ethereum.
There are two primary categories of rollups: optimistic rollups and zero-knowledge (ZK) rollups. Optimistic rollups (Optimism, Arbitrum, Base) assume transactions are valid by default and use a fraud proof mechanism where anyone can challenge an incorrect state transition within a dispute window, typically seven days. ZK rollups (zkSync Era, StarkNet, Polygon zkEVM, Scroll, Linea) generate cryptographic validity proofs (SNARKs or STARKs) that mathematically guarantee every state transition is correct, providing much faster finality without a challenge period.
As of 2026, rollups collectively process far more daily transactions than Ethereum mainnet, with Arbitrum One and Base leading in TVL and activity, together holding roughly three-quarters of all Layer 2 DeFi liquidity. The rollup-centric roadmap has become Ethereum’s official scaling strategy, with EIP-4844 (Proto-Danksharding, deployed March 2024) reducing rollup data costs by 80 to 99% through the introduction of blob transactions. Ethereum’s Fusaka upgrade in December 2025 then brought genuine Data Availability Sampling to Ethereum blobs for the first time (via PeerDAS) and, through subsequent Blob Parameter Only forks, raised the blob capacity target well beyond its original level, with further expansion planned as part of the path toward full Danksharding.
Origine & Histoire
2014, early concepts: Vitalik Buterin’s original Ethereum whitepaper acknowledges the need for scaling, though the specific concept of rollups does not yet exist. Early research focuses primarily on state channels (such as the Raiden Network) and sidechains.
2018, the rollup breakthrough: Researcher Barry Whitehat publishes an early description of “roll_up,” a concept for aggregating transaction data and posting it to Ethereum via validity proofs. Around the same time, alternative scaling models like Plasma, led by Joseph Poon and Vitalik Buterin, stall due to data availability and complex exit issues.
2020, first implementations: Fuel Labs launches an early optimistic rollup on Ethereum mainnet focused on UTXO-based payments. Loopring deploys a ZK rollup for decentralized exchange trading, and StarkWare introduces StarkEx for application-specific scaling, notably powering dYdX’s original order book.
2021, the rollup-centric pivot: Vitalik Buterin publishes “An Incomplete Guide to Rollups,” cementing them as Ethereum’s primary scaling path over Plasma. Teams like Offchain Labs (Arbitrum One) and Optimism launch their mainnets to the public, quickly becoming dominant Layer 2 networks by total value locked.
2023, EVM equivalence and modular stacks: General-purpose ZK rollups capable of executing complex smart contracts, such as zkSync Era and Polygon zkEVM, go live. Optimism releases the OP Stack framework, enabling Coinbase to launch Base and kicking off the “Superchain” thesis.
2024, the blob era (EIP-4844): Ethereum activates the Dencun upgrade. By introducing blob transactions via EIP-4844, the cost for rollups to post data to Layer 1 drops sharply, often by 90% or more, reducing L2 transaction fees to fractions of a cent in many cases.
2025 to 2026, market maturity and expanding blob capacity: Ethereum’s Fusaka upgrade activates in December 2025, introducing PeerDAS and bringing production-grade Data Availability Sampling to Ethereum blobs for the first time. Subsequent Blob Parameter Only forks raise the blob capacity target well above its original level within weeks of Fusaka’s launch. The L2 ecosystem matures into a genuinely multi-chain market, with 70-plus active rollups collectively securing somewhere in the $45 to 50 billion range in total value locked at various points during 2026, alongside daily transaction counts that dwarf Ethereum mainnet’s own throughput. Based rollups (which use L1 validators for sequencing) and shared sequencing networks continue to develop as attempts to address fragmentation and cross-chain composability, and Ethereum’s forthcoming Glamsterdam upgrade targets further gains in mainnet throughput and settlement capacity for the L2s that depend on it.
“In the long term, rollups will be the dominant scaling model for Ethereum. They give you the same security as L1, with dramatically higher throughput and dramatically lower costs.” Vitalik Buterin, Ethereum co-founder.
En termes simples
The bus analogy (throughput): imagine a busy highway (Ethereum Layer 1) clogged with individual cars. A rollup acts like a shuttle bus service. It picks up hundreds of passengers (transactions), drives them to their destinations via side roads (off-chain execution), and then uses just a single lane on the main highway to report the final seating chart. Instead of hundreds of cars causing traffic, one bus handles the load.
The zip file analogy (data): think of a rollup like compressing a folder of files before emailing it. Instead of sending a thousand individual documents one by one, which would clog your inbox, a rollup “zips” them into a single compressed package (a batch) and sends it all at once. The underlying blockchain only has to store the single attachment.
Caractéristiques techniques clés
Rollup Architecture
Rollups maintain their own execution environment, whether EVM-compatible or a custom virtual machine, where transactions are processed
A sequencer receives user transactions, orders them, and creates batches for submission to L1
Transaction data, or compressed representations of it, is posted to Ethereum as calldata or, more commonly today, as blobs under EIP-4844
A state root, the cryptographic hash of the rollup’s entire state after processing the batch, is posted to L1
Bridge contracts on L1 hold deposited assets and verify state transitions according to the rollup’s proof system
Rollups optimistes
Assume all submitted state transitions are valid by default, hence “optimistic”
Any observer can submit a fraud proof during the challenge period, typically seven days, if they detect an invalid state transition
If a fraud proof succeeds, the invalid batch is reverted and the proposer is slashed, losing staked collateral
The seven-day challenge window means withdrawals from optimistic rollups to L1 require a waiting period; bridging services offer faster exits for a fee by fronting liquidity
Interactive fraud proofs, such as Arbitrum’s, use a bisection protocol to narrow down the exact step where execution diverged
Non-interactive fraud proof systems generate a single proof that identifies the invalid computation
Rollups ZK
Generate cryptographic validity proofs (ZK-SNARKs or ZK-STARKs) for every batch of transactions
The L1 verifier contract checks the proof to confirm that every state transition is mathematically correct
Withdrawals can be processed much faster than optimistic rollups after proof verification, without a multi-day challenge period
Proving is computationally expensive; ZK rollups use specialized prover hardware or distributed proving networks
ZK-SNARKs (used by zkSync, Polygon zkEVM, and Scroll) are smaller but require a trusted setup ceremony
ZK-STARKs (used by StarkNet) have no trusted setup requirement but produce larger proofs
How a Rollup Transaction Works
A user submits a transaction, such as a token swap, to the rollup’s sequencer via RPC
The sequencer includes the transaction in its mempool, orders it, and executes it on the rollup’s state
The user receives a “soft confirmation” within seconds, meaning the sequencer has committed to including the transaction
The sequencer batches hundreds or thousands of transactions together and compresses the data
The batch is posted to Ethereum L1, typically as a blob transaction, containing the transaction data and the new state root
For optimistic rollups, the state root is considered final after the challenge period passes with no successful fraud proof
For ZK rollups, a validity proof is generated and submitted to the L1 verifier contract, which confirms correctness in a single on-chain verification step
Data Availability and EIP-4844
Rollups must post enough data to L1 so that anyone can reconstruct the rollup’s state independently
Before EIP-4844, rollups posted data as calldata, which is permanently stored and comparatively expensive
EIP-4844, activated in March 2024, introduced blobs: large data payloads that are available for roughly 18 days and then pruned
Blob fees are priced in a separate fee market from L1 execution gas, keeping rollup costs predictable and low
Ethereum’s blob capacity has already expanded well beyond its original 2024 launch level, roughly a 3-blob target with a 6-blob maximum per block, through the December 2025 Fusaka upgrade and its Blob Parameter Only forks, which raised the target into the double digits within weeks. Full Danksharding remains the longer-term goal for expanding blob capacity even further, enabling substantially greater rollup throughput over time
Avantages désavantages
Avantages
Désavantages
Ethereum-Grade Security: Rollups inherit L1 security guarantees; funds are secured by Ethereum’s validator set, not the rollup’s own consensus
Sequencer Centralization: Most rollups operate a single centralized sequencer that can censor transactions or capture MEV, though users retain L1 force-inclusion as an escape hatch
Débit massif : Rollups process thousands of TPS, versus roughly 15 to 30 TPS on Ethereum mainnet, enabling high-frequency trading, gaming, and social applications
Délais de retrait : Optimistic rollups require a multi-day challenge period for L1 withdrawals; fast bridges mitigate this but introduce trust assumptions and fees
Frais de transaction peu élevés : Post-EIP-4844, rollup transactions commonly cost fractions of a cent, making micro-transactions and frequent interactions economically viable
Liquidité fragmentée : Dozens of rollups split liquidity and users across isolated ecosystems, creating UX friction for cross-rollup interactions
Compatibilité EVM : Major rollups (Arbitrum, Optimism, Base, zkSync, Polygon zkEVM) support Solidity smart contracts, enabling easy migration of Ethereum dApps
Complexité pour les développeurs : Building cross-rollup applications requires handling bridging, different confirmation times, and varying EVM compatibility levels
Composability with Ethereum: Rollups can interact with L1 contracts through message passing, and rollup tokens can be bridged back to Ethereum for L1 DeFi access
Proving Costs (ZK): Generating ZK proofs is computationally intensive and expensive, requiring specialized hardware that creates centralization pressure on the prover side
Décentralisation progressive : Rollup teams are actively decentralizing sequencers, provers, and governance, moving toward more trustless operation over time
Antécédents en matière de sécurité : ZK rollup codebases are complex and comparatively new; bugs in proving systems could theoretically allow invalid state transitions
Ecosystem Innovation: The rollup-centric roadmap has attracted enormous developer talent and venture funding, accelerating L2 ecosystem growth
Cross-Rollup UX Friction: Moving assets between rollups requires bridging, which can be slow, expensive, and risky; bridge exploits have caused billions of dollars in losses
Gestion du risque
User-Level Risks:
Verify that your rollup of choice has active fraud proof or validity proof systems; some rollups launched with “training wheels,” where an admin multisig can override state
Use canonical bridges for large withdrawals, accepting the delay, rather than third-party bridges, for maximum security
Check the rollup’s stage on L2BEAT: Stage 0 (full training wheels), Stage 1 (permissionless fraud proofs and a trustless exit path), Stage 2 (fully decentralized, with no privileged roles). As of 2026, Arbitrum One is among the few major optimistic rollups to have reached Stage 1, and no major rollup has yet reached Stage 2; prefer Stage 1-plus rollups for significant assets
Keep funds diversified across rollups to mitigate the risk of a single rollup experiencing a critical bug or sequencer downtime
Risques pour les développeurs :
Test contracts thoroughly on rollup testnets; subtle EVM differences, such as precompile availability or opcode gas costs, can cause unexpected behavior
Implement L1 fallback mechanisms for critical operations; if the sequencer goes down, users should be able to interact via L1 force-inclusion
Avoid hard dependencies on sequencer liveness for time-sensitive operations, such as liquidations; design protocols to function under degraded conditions
Monitor rollup upgrade proposals, since admin keys and governance processes can change core contract behavior
Bridge and Interoperability Risks:
Cross-rollup bridges are high-value targets; historical bridge exploits (Ronin, Wormhole, Nomad) collectively account for well over a billion dollars in losses across crypto’s history
Use well-audited bridges with substantial security deposits and avoid unfamiliar bridging protocols
Prefer bridges that rely on Ethereum L1 for settlement rather than external validator sets
Pertinence culturelle
Rollups represent a philosophical compromise in the blockchain scaling debate. Rather than scaling through larger blocks (the “big block” approach associated with chains like Bitcoin Cash and Solana) or sharding the base layer directly, the rollup-centric roadmap preserves L1 decentralization while offloading execution to specialized chains. This approach has become the defining narrative of the Ethereum ecosystem since 2020.
The emergence of “L2 season” in 2023 and 2024, with major airdrops from Arbitrum, Optimism, StarkNet, and zkSync, created new cultural dynamics around “airdrop farming,” where users performed transactions across rollups specifically to qualify for token distributions. This behavior, while sometimes criticized as mercenary, drove genuine adoption and battle-testing of rollup infrastructure.
The proliferation of rollups has also sparked cultural debates about fragmentation. Critics argue that dozens of rollups split Ethereum’s network effects and create a confusing user experience, coining terms like “rollup fatigue.” Proponents counter that shared standards, such as the OP Stack Superchain concept, Polygon’s AggLayer, and zkSync’s Elastic Chain vision, will eventually create smoother interoperability, making the underlying rollup infrastructure largely invisible to end users.
Rollups have fundamentally shifted how new crypto projects launch. Rather than creating a new L1 blockchain, projects increasingly deploy as rollups, using frameworks like Arbitrum Orbit, the OP Stack, or the ZK Stack, inheriting Ethereum security while customizing their execution environment. This “rollup-as-a-service” model has led to application-specific rollups for DeFi, gaming, and social platforms.
Exemples du monde réel
Arbitrum One, DeFi Hub
Scénario: Ethereum’s high gas fees, sometimes tens or hundreds of dollars per swap during peak congestion, were pricing out DeFi users and limiting protocol adoption.
Mise en œuvre: Arbitrum One launched as an optimistic rollup with full EVM compatibility. Major DeFi protocols, including GMX, Aave, Uniswap, and Camelot, deployed on Arbitrum, and native protocols like GMX built perpetual trading platforms that could not have operated economically on L1.
Résultat: Arbitrum grew into the largest L2 by TVL, holding somewhere in the $15 to 18 billion range at various points through 2025 and 2026, and processing large volumes of transactions daily at a small fraction of L1’s cost. The ARB airdrop in March 2023 distributed well over a billion dollars in value to early users at the time and helped establish the Arbitrum DAO as a major governance entity. Arbitrum has also become the only major optimistic rollup to reach L2BEAT’s Stage 1 security rating as of 2026, reflecting its permissionless fraud proof system.
Base, Consumer Crypto Gateway
Scénario: Coinbase wanted to bring its large user base on-chain while maintaining low costs and a smooth experience.
Mise en œuvre: Base launched as an OP Stack rollup in August 2023, using Coinbase’s brand, distribution, and fiat on-ramp integration. Social applications, meme coins, and consumer-facing dApps found a home on Base.
Résultat: Base grew rapidly to become one of the most active L2s by daily transactions and one of the largest by TVL, at times holding more than $10 billion, demonstrating that rollups could serve as a genuine consumer crypto on-ramp. Post-EIP-4844, and further aided by the 2025 Fusaka blob capacity increases, typical Base transactions cost a small fraction of a cent, enabling micro-transaction use cases.
zkSync Era, ZK Rollup for General-Purpose Smart Contracts
Scénario: ZK rollups offered attractive security properties, including much faster withdrawals than optimistic rollups, but achieving full EVM compatibility with ZK proofs was an enormous technical challenge.
Mise en œuvre: zkSync Era by Matter Labs launched a general-purpose ZK rollup with native account abstraction and EVM compatibility, using LLVM-based compilation to convert Solidity code into circuits that can be proven with ZK-SNARKs.
Résultat: zkSync Era demonstrated that ZK rollups could support a broad range of DeFi applications with much lower withdrawal latency than optimistic rollups. Its ZK token airdrop in June 2024 distributed tokens to hundreds of thousands of wallets. Its TVL has fluctuated considerably since then, falling sharply after the airdrop before recovering into the low billions of dollars by 2026, illustrating how much post-airdrop token distributions can affect a rollup’s on-chain activity in the near term.
StarkNet, STARK-Based Execution Environment
Scénario: SNARKs required trusted setups and were limited in certain respects; a rollup using STARKs could offer transparent, post-quantum-resistant proofs instead.
Mise en œuvre: StarkNet by StarkWare uses Cairo, a custom programming language optimized for STARK proof generation, and supports advanced features like native account abstraction and parallel transaction execution.
Résultat: StarkNet established an alternative ZK proving model with no trusted setup requirement. Its ecosystem grew to include DeFi protocols such as Ekubo, JediSwap, and Nostra, and demonstrated that non-EVM execution environments could attract meaningful developer adoption, even as its TVL has remained smaller than the leading EVM-compatible rollups.
Tableau de comparaison
Fonctionnalité
Optimistic Rollup (Arbitrum)
ZK Rollup (zkSync Era)
Sidechain (PoS polygonal)
Alternative L1 (Solana)
Modèle de sécurité
Ethereum L1 plus fraud proofs
Ethereum L1 plus validity proofs
Own validator set (independent security)
Own validator set (independent security)
Withdrawal to Ethereum
About 7 days (challenge period)
Minutes to a few hours (after proof verification)
Roughly 30 minutes to 3 hours (checkpoint)
Requires a bridge (variable)
Finalité de la transaction
Soft: seconds; hard: about 7 days
Soft: seconds; hard: hours (proof time)
Seconds (own consensus)
Sub-second (own consensus)
Coût de transaction
Une fraction de centime à quelques centimes
A fraction of a cent to around a dime
Une fraction de centime
Une fraction de centime
Compatibilité EVM
Full (Nitro engine)
High (LLVM compilation)
Full (EVM chain)
None (Rust/SVM)
Hypothèses de confiance
Requires at least one honest verifier during the challenge period
Requires a correct prover implementation
Requires an honest majority of PoS validators
Requires an honest majority of validators
Note: TPS and exact fee figures vary by network conditions and change as upgrades ship; treat the figures above as directional rather than exact.
Termes connexes
Couche 2 (L2)Layer 2: Any scaling solution that operates on top of a Layer 1 blockchain to increase throughput while inheriting base layer security
Optimistic Rollup: A rollup variant that assumes validity by default and relies on fraud proofs during a challenge period to detect invalid state transitions
ZK Rollup: A rollup variant that uses zero-knowledge cryptographic proofs to mathematically verify the correctness of every batch of transactions
Disponibilité des données: The guarantee that transaction data is published and accessible so anyone can reconstruct the rollup’s state independently
EIP-4844 (Proto-Danksharding): An Ethereum upgrade that introduced blob transactions, dramatically reducing the cost of posting rollup data to L1
séquenceur: The entity responsible for ordering transactions and creating batches in a rollup, currently centralized in most rollups
Preuve de fraude : A mechanism in optimistic rollups that allows anyone to challenge and revert an invalid state transition
Validity Proof: A cryptographic proof (SNARK or STARK) submitted by ZK rollups to verify correct execution on L1
Chaîne latérale: A separate blockchain connected to the main chain via a bridge; sidechains do not inherit L1 security, distinguishing them from rollups
Pont: A protocol that enables asset transfers between L1 and rollups, or between rollups, using lock-and-mint or burn-and-mint mechanisms
Éclat d'eau : The fuller implementation of Ethereum’s data availability scaling roadmap, of which PeerDAS and the 2025-2026 blob capacity increases are an important early step
QFP
Q: What is the difference between an optimistic rollup and a ZK rollup? Optimistic rollups assume transactions are valid and allow a challenge period, typically seven days, where anyone can submit a fraud proof if they detect an error. ZK rollups generate a cryptographic validity proof for every batch that is verified on-chain, providing much faster finality. Optimistic rollups have historically been simpler to build and have offered broader EVM compatibility, while ZK rollups offer faster finality and shorter withdrawal delays, a gap that has been narrowing as ZK tooling matures.
Q: Why do optimistic rollup withdrawals take about 7 days? The window allows anyone to verify the rollup’s state transitions and submit a fraud proof if they find an invalid batch. This delay ensures that even if the rollup operator is malicious, honest verifiers have enough time to detect and challenge incorrect state. Third-party bridges offer faster withdrawals by fronting the liquidity themselves and waiting out the challenge period.
Q: Are rollups as secure as Ethereum mainnet? Rollups are designed to inherit Ethereum’s security, but the degree depends on the rollup’s maturity stage. A fully decentralized rollup, at L2BEAT’s Stage 2, with active fraud or validity proofs, decentralized sequencing, and a strong codebase, would approach L1-level security, though no major rollup has reached that stage as of 2026. Many rollups remain at Stage 0 or Stage 1, with varying degrees of admin override capability. Check L2BEAT for each rollup’s current security assessment.
Q: What did EIP-4844 change for rollups, and has anything changed since? EIP-4844 (Proto-Danksharding), deployed in March 2024, introduced blob transactions, a new, cheaper data type designed specifically for rollup data. Before EIP-4844, rollups posted data as calldata, which was expensive and permanently stored. Blobs are typically 80 to 99% cheaper and are pruned after roughly 18 days. Since then, Ethereum’s December 2025 Fusaka upgrade added real Data Availability Sampling to blobs via PeerDAS and, through follow-on parameter-only forks, meaningfully raised how many blobs each block can carry, generally keeping rollup costs low even as usage has grown.
Q: Which rollup should I use? For DeFi and general use, Arbitrum One and Base remain among the most active ecosystems with the deepest liquidity. For applications that particularly value faster L1 finality, consider ZK rollups like zkSync Era or StarkNet. For specific applications, check which rollup hosts the protocol you want to use, and consider using a bridge aggregator to find efficient routes between rollups.
Q: Will rollups eventually replace Ethereum mainnet? Rollups complement Ethereum mainnet rather than replacing it. Ethereum L1 serves as the settlement and data availability layer, the underlying system that enforces rollup rules. Most user-facing activity has migrated to rollups, but L1 remains essential for high-value settlements, rollup proof verification, and the security guarantees that make rollups possible in the first place.
Q: What is a “based rollup”? A based rollup uses Ethereum L1 validators (proposers) for transaction sequencing instead of running a centralized sequencer. This eliminates sequencer centralization risk and provides atomic composability with L1, but introduces latency roughly equal to Ethereum’s block time for soft confirmations. Based rollups represent an emerging design that maximizes alignment with Ethereum’s decentralization.
La loi sur la clarification du marché des actifs numériques est entrée dans ce qui pourrait être sa phase la plus critique. Après avoir manqué la réunion de la Maison Blanche
Vanguard a franchi une nouvelle étape dans son expansion sur le marché des actifs numériques en publiant une offre d'emploi pour un nouveau poste de direction.
Le PDG de Robinhood, Vlad Tenev, estime que le prochain chapitre de l'industrie des cryptomonnaies sera marqué par la tokenisation des actifs du monde réel.
