L2 bridging cost is the friction users feel most directly when moving assets between Ethereum L1 and rollups (covered in the Ethereum L2 comparison),, or between two L2s. The choice between native bridges (operated by each rollup) and third-party bridges (Across, Hop, Hyperlane, LayerZero) shapes both cost and confirmation time. As of April 2026, third-party bridges handle roughly 78% of cross-rollup volume per DefiLlama bridge data, despite native bridges existing for every major L2.
​

This guide compares the two paths across cost, latency, security model, and asset support. The headline: native bridges from L1 to L2 are cheaper for one-way deposits but slow for L2-to-L1 withdrawals (7 days for optimistic rollups). Third-party bridges complete in seconds for a 0.05–0.20% fee. CCTP, Circle's stablecoin-native rail, sits in a separate category — cheaper than third-party for native USDC, slower than third-party but faster than canonical.
​

A bridge is a protocol that moves assets between two blockchains. The underlying mechanism varies. Lock-and-mint bridges hold assets in a contract on the source chain and mint a wrapped representation on the destination. Burn-and-mint bridges (CCTP, native stablecoin rails) destroy the asset on the source and issue native tokens on the destination. Liquidity-pool bridges (Across, Hop, Stargate) maintain pools on each chain and pay relayers from those pools.
​

Each model has trade-offs. Lock-and-mint creates wrapped tokens that are not fungible with native versions on the destination. Burn-and-mint preserves fungibility but requires issuer support — only a few stablecoins (USDC via CCTP, USDT via LayerZero's OFT, PYUSD) operate this way today. Liquidity-pool bridges trade speed and fungibility for paying liquidity providers a fee.
​

Every L2 ships with a canonical bridge that moves assets between Ethereum L1 and the L2. Examples: Arbitrum's Arbitrum Bridge, Optimism's Standard Bridge, Base's Base Bridge, zkSync's Portal, Linea's bridge, Scroll's bridge. These are operated by the rollup teams (or governance) and are part of the chain's security model.
​

How they work. Depositing from L1 to L2: the user sends ETH or an ERC-20 to the bridge contract on L1; the rollup's sequencer credits an equivalent amount to the user's address on L2 within 1–15 minutes. Withdrawing from L2 to L1: the user submits a withdrawal transaction on L2; the rollup posts the state root including this withdrawal to L1; for optimistic rollups, the user must wait the 7-day fraud-proof challenge window before claiming on L1; for ZK rollups, the user waits 30–90 minutes for the validity proof to verify, then claims.
​

Cost. The L1 deposit transaction costs $4–18 in mainnet gas at typical conditions. The L2 deposit credit is essentially free. The L2-to-L1 withdrawal transaction costs $0.10–$2 on the L2; the L1 claim transaction (after the wait period) costs $4–25 depending on L1 conditions. Total round-trip for a deposit-and-withdraw can run $8–43 in gas before any fees.
​

Trust model. Native bridges inherit the rollup's security. For Stage 0 chains (most rollups today), the rollup's security council can pause the bridge or override withdrawals via emergency upgrade. For Arbitrum One (Stage 1), withdrawals are permissionless once finalized.
​

Third-party bridges connect any-to-any chains via liquidity pools or messaging-layer protocols. The dominant systems by 2026 volume:

The comparison below shows the cost of a representative $1,000 USDC bridge between Ethereum L1 and Arbitrum, as of April 2026.

For one-way L1-to-L2 deposits, the native bridge is competitive — the protocol fee on third-party bridges adds $0.30–$1.20 on top of the same L1 gas cost. For L2-to-L1 withdrawals, third-party bridges win on time (seconds to minutes versus 7 days for optimistic rollups). The 7-day delay is the single largest reason third-party bridges dominate withdrawal volume.
​

For moving between two L2s (Arbitrum to Base, Optimism to Linea, etc.), native bridges are not a direct option. The canonical path is L2 → L1 → L2, which incurs two sets of L1 gas costs and the 7-day delay for optimistic rollups. Third-party bridges or stablecoin-native rails are essentially mandatory.
​

For a $1,000 USDC transfer from Arbitrum to Base in April 2026:

The economics make canonical L2-to-L2 bridging impractical for any user-facing flow. For protocols moving stablecoins as part of treasury operations or trading, the choice is between speed (Across, Stargate) and cost (CCTP for USDC).
​

Burn-and-mint stablecoin protocols are a separate category. They preserve native stablecoin issuance — no wrapped tokens, full fungibility on the destination chain.
​

Circle's CCTP burns USDC on the source chain and mints fresh USDC on the destination. Supported on 12 chains as of April 2026: Ethereum, Arbitrum, Base, OP Mainnet, Polygon PoS, Avalanche C-Chain, Solana, Noble (Cosmos), Sui, Aptos, zkSync Era, with Linea and Scroll integrations announced. CCTP fee is paid in destination-chain gas only — no protocol surcharge from Circle. Total cost runs $0.30–$1 plus origin and destination gas.
​

Tether's USDT0 uses LayerZero's OFT standard for omnichain USDT issuance. Supported on 30+ chains. Fee is paid via LayerZero's messaging layer ($0.50–$3 base) but no surcharge from Tether.
​

PYUSD via LayerZero covers PayPal's USD stablecoin across 8+ chains using the same OFT mechanism.
​

For applications transferring native stablecoins specifically (the stablecoin liquidity guide ranks chains by depth),, these rails are usually cheapest and preserve issuer-issued tokens on every chain. The trade-off is they only work for the specific stablecoins integrated.
​

Bridge aggregators sit on top of multiple bridges and route to the cheapest or fastest option per transaction. The major ones in 2026 are LI.FI, Socket, Bungee, and Rango. Aggregators query bridge APIs in real time and quote the best price for a given source-destination-amount combination.
​

Most aggregators charge no protocol fee — they earn through bridge integration kickbacks. The user sees the underlying bridge fee plus a small platform margin (typically 0.05–0.10%). For applications that need to support many bridge paths without integrating each separately, aggregator SDKs are the common solution.
​

The trade-off with aggregators: they add an extra latency hop for routing decisions (50–200ms typically) and the bridge selection logic is opaque. Some aggregators have been criticized for routing to bridges that pay higher kickbacks rather than the best user-side price.
​

Bridge security is where the comparison gets technical. The three primary models:
​

Lock-and-mint with rollup security (native bridges): Inherits the rollup's security model. Stage 1 chains like Arbitrum One have permissionless withdrawals; Stage 0 chains have multisig override. The trust assumption is the rollup's proof system plus its operator multisig.
​

Optimistic verification with bonded relayers (Across, UMA-secured bridges): Relayers post bonds to front liquidity. If a relayer is fraudulent, the bond is slashed via UMA's optimistic oracle. Trust assumption is at least one honest watcher during the dispute window.
​

Multi-validator messaging (LayerZero, Wormhole, Hyperlane): A set of validators or oracles attest to message validity. Trust assumption is the validator set's honesty threshold (typically 2/3 or majority).
​

Bridge hacks have hit all three models historically. Ronin ($625M, 2022), Wormhole ($325M, 2022), Nomad ($190M, 2022), Multichain ($230M, 2023). The total bridge-hack loss exceeds $2.8B per rekt.news's leaderboard. Most losses came from configuration errors or compromised validator keys, not the underlying cryptographic protocols.
​

Eco is a stablecoin execution network that abstracts bridge selection. Eco Routes (CLI + API) accepts a stablecoin transfer intent and routes through whichever combination is cheapest and fastest — CCTP for native USDC, third-party bridges for cross-stable swaps, canonical bridges where appropriate. The network supports 15 chains including all major L2s plus Solana and other non-EVM chains. Production teams running stablecoin payments or treasury operations across multiple L2s use Routes to avoid maintaining separate integrations with Across, CCTP, Stargate, and a half-dozen other bridges.
​

Neither is categorically safer. Native bridges inherit rollup security, which depends on the rollup's stage and operator multisig. Third-party bridges have specific security models (optimistic, multi-validator, liquidity-bonded) and have suffered larger historical hacks but more often via operational errors than protocol breaks. Use native for L1↔L2 deposits where speed isn't critical; use third-party for fast L2↔L2 or fast withdrawals.
​

Optimistic rollups require a 7-day fraud-proof challenge window before withdrawals finalize on L1. ZK rollups don't have this delay but still need 30–90 minutes for the validity proof to verify on L1. Third-party bridges sidestep both by fronting liquidity on L1 and waiting for the canonical settlement themselves.
​

CCTP if both chains are CCTP-supported. As of April 2026 that includes Arbitrum, Base, OP Mainnet, Polygon, Avalanche, Solana, Sui, Aptos, and zkSync Era. CCTP fees are typically $0.30–$1 plus gas. Across Protocol is the next-cheapest option for any USDC route, including pairs CCTP doesn't yet cover.
​

Across completes in 15–30 seconds. Stargate completes in 2–4 minutes. Hop completes in 3–6 minutes. CCTP takes 15–25 minutes due to Circle's attestation step. Hyperlane and LayerZero vary by route configuration. Cross-VM routes (e.g., Ethereum to Solana) typically take longer than EVM-to-EVM.
​

For amounts under $10K, a third-party bridge is usually faster and cheaper. CEX deposit-and-withdraw routes work but require trust in the exchange and incur withdrawal fees that can exceed bridge fees. For amounts over $1M, OTC desks or stablecoin orchestration networks become cost-competitive with both bridges and CEXs.
​

Recovery depends on the bridge. Native bridge hacks usually result in protocol governance action (in some cases full reimbursement from the rollup treasury). Third-party bridge hacks have varied — Wormhole was reimbursed by Jump Crypto, Nomad recovered some funds via whitehat appeals, others have not been made whole. For large flows, treat bridge risk like counterparty risk: limit exposure per bridge.
​

Most third-party bridges charge a percentage fee (0.05–0.20% of notional) plus a flat gas component. CCTP charges only gas (no percentage). Native bridges charge only gas. For very small transfers (under $50), the flat gas component dominates and percentage fees barely register. For large transfers (over $100K), percentage fees become material — a 0.10% fee on $1M is $1,000.
​

The L1 gas cost is essentially the same across all L2 native bridges — roughly $4–18 depending on mainnet conditions. The marginal differences come from each rollup's deposit gas overhead, which is small (10–30%). Base, Optimism, and Arbitrum tend to have the lowest L1 deposit costs in practice; ZK rollups like zkSync sometimes cost slightly more due to additional state-init overhead.
​