Cross-Chain Crypto Bridges: A Research Guide

What is a crypto bridge?

A crypto bridge moves tokens between blockchains that can't talk to each other natively. In 2026 the dominant approaches are intent-based settlement (Across V4, deBridge DLN), native burn-and-mint (Circle CCTP V2), and guardian/committee messaging (Wormhole, LayerZero/Stargate) — but the April 2026 Kelp DAO exploit ($292M via a LayerZero 1-of-1 DVN) proved that committee-based verification is only as strong as its configuration. Last verified: 2026-05-27.

ETH on Ethereum mainnet, ETH on Arbitrum, and ETH on Base are technically different tokens — to move between them you need a bridge. Bridges have lost users over $2.8B historically. Ronin (Ronin is the Axie Infinity sidechain bridge that was drained for $625M in March 2022 by Lazarus Group) bled $625M in March 2022 when Lazarus Group compromised five of nine multisig validators. Poly Network lost $611M in August 2021 to a contract-permission bug (most funds returned by the attacker). Wormhole (Wormhole is a generic cross-chain messaging protocol covering 35+ chains including Solana, hacked for $325M in February 2022) was drained for $325M in February 2022 via a signature-verification flaw and only made whole because Jump Crypto refunded users. BNB Bridge bled $586M in October 2022. Nomad lost $190M in August 2022 to a one-line initialization bug. Multichain (Multichain, formerly Anyswap, was a major lock-and-mint bridge that collapsed in mid-2023 after the CEO disappeared, freezing $1.2B in user funds) didn't even get hacked — its CEO disappeared in mid-2023 and $1.2B in user funds froze. Kelp DAO lost $292M in April 2026 when Lazarus Group (TraderTraitor sub-group) compromised two LayerZero RPC nodes, DDoS'd external nodes, and exploited Kelp's single-verifier (1-of-1 DVN) bridge setup to approve phantom withdrawal messages — the largest DeFi hack of 2026. Bridge picks matter more than any other DeFi category.

The three bridge architectures

1. Lock-and-mint (legacy): tokens locked on chain A, wrapped versions minted on chain B. Every major nine-figure bridge exploit through 2022 lived in this category. The wrapped token is a synthetic IOU — if the source-side custody fails, the IOU is worthless. Avoid in 2026.
2. Liquidity pool / intent-based (modern): relayers front your funds on the destination chain instantly and reclaim from the source side after a settlement window. Across uses UMA's optimistic oracle; deBridge uses its DLN intent layer. Sub-minute finality, single-digit basis-point fees, far smaller attack surface. ERC-7683 (ratified early 2025) is now the cross-chain intents standard — Across, UniswapX, and CoW Protocol have production endpoints; 88% of Across volume runs ERC-7683 as of Q3 2025.
3. Native message passing: cross-chain messaging via LayerZero (LayerZero is a cross-chain messaging protocol with a configurable DVN security model — the Kelp DAO $292M April 2026 exploit showed that a 1-of-1 DVN setup is catastrophically insecure), Chainlink CCIP, Hyperlane, or Wormhole's guardian set. These are the rails — most user-facing bridges sit on top. Security depends entirely on the DVN/guardian configuration: after the April 2026 Kelp exploit, LayerZero Labs committed to migrating all defaults to 5/5 DVN where possible, no less than 3/3.
4. Native burn-and-mint (CCTP): Circle burns USDC on the source chain and mints fresh USDC on the destination via signed attestation. No wrapped asset, no pool, no relayer counterparty. Only works for USDC — but for USDC it is the lowest-trust option available. CCTP V2 launched March 11, 2025; V1 enters manual deprecation July 31, 2026.

Top 6 crypto bridges in 2026

Across V4 (intent + ZK, 18+ chains including Solana), Stargate (LayerZero-owned post-August 2025 acquisition, 80+ chains), CCTP V2 (17+ chains, free for USDC), deBridge DLN (intent, 10+ chains), Hop (AMM, EVM L2s, lower volume), Wormhole (guardian set, 35+ chains). Last verified: 2026-05-27.

Across — deep dive

Full guide: Across.

Best for

Default EVM bridge. Across V4 (announced July 2025) added ZK-proof verification for non-EVM chains, eliminating the custom adapter dependency that limited V3. Solana mainnet integration launched August 20, 2025, making Across viable for EVM ↔ Solana. Relayers compete to front your funds on the destination chain in 30–90 seconds; UMA's optimistic oracle (UMA is an optimistic oracle — assertions are accepted unless challenged within a dispute window, with stake-backed economic guarantees) settles the source side over a longer window the user never sees. ERC-7683 orders now represent 88% of total Across volume.

How it actually works

You deposit on the source chain; a permissionless relayer sees the deposit and pays you out on the destination immediately. Hours later, the relayer claims back from the Across pool on the source chain, but only after UMA's optimistic oracle confirms the deposit was real. If a relayer lies, anyone can dispute and the relayer loses their bond. You get speed; the relayer takes the timing risk.

Trade-offs

The fee is competitive but a few basis points above CCTP for plain USDC routes. Relayer competition matters: on rarely-used chains you'll occasionally see thinner quotes. Hart Lambur's Bridge Across proposal (late 2025) floated converting ACX tokens to equity in a new C Corp at a 1:1 ratio — the DAO structure has become a bottleneck for enterprise contracts. Governance uncertainty is a risk to monitor.

Audit history & trust

Built on UMA's optimistic oracle, audited by OpenZeppelin and ChainSecurity. Has not been hacked through $34B+ in cumulative volume. Team is doxxed and US-based.

Cost

0.05–0.10% bridge fee plus gas on both sides. 30–90 seconds typical.

Circle CCTP V2 — deep dive

CCTP isn't a bridge; it's the USDC issuer doing the bridging itself. That single architectural choice eliminates the entire wrapped-asset attack surface.

Best for

USDC transfers between supported chains. CCTP V2 (CCTP V2 is Circle's Cross-Chain Transfer Protocol V2, launched March 11, 2025 — burns native USDC on source, mints fresh USDC on destination via signed attestation; adds Fast Transfer with 8–20s finality and programmable hooks) burns native USDC on the source chain and mints fresh native USDC on the destination via a signed attestation from Circle. No wrapped USDC, no liquidity pool that can be drained, no bridge counterparty that can vanish.

V2 vs V1

CCTP V2 launched March 11, 2025. It added three things over V1: Fast Transfer (8–20 second finality via Circle attestation guarantee, vs. 13+ minutes for V1 Ethereum hard finality), programmable hooks (atomic actions on USDC arrival — swap, deposit to Aave, etc.), and a versioned message format. CCTP V1 is now "Legacy" — deprecation announced November 14, 2025, manual phase-out begins July 31, 2026. If you are integrating CCTP, use V2.

Supported chains (as of May 2026)

Ethereum, Arbitrum, Base, Optimism, Polygon PoS, Avalanche, Solana, Linea, Unichain, World Chain, Sonic, HyperEVM, and others — 17 chains live. Aptos and Sui were expected by mid-2026 per Circle's roadmap.

Trade-offs

USDC only (no USDT, no ETH, no anything else). Trust is concentrated in Circle — but if you hold USDC, you already trust Circle. The only real downside is that CCTP is progressively slower to add chains than permissionless bridging protocols.

Cost

Zero protocol fee — you pay only gas. Cheapest USDC bridge in crypto.

Stargate (LayerZero-owned) — deep dive

Stargate is the breadth play — and is now fully owned by LayerZero. The August 24, 2025 acquisition (95% DAO vote in favor, $110M deal) dissolved Stargate's independent governance. STG holders receive 0.08634 ZRO per STG. Stargate stakers receive 50% of protocol revenue for six months post-acquisition; the remaining 50% funds ZRO buybacks.

Best for

Broadest chain coverage (80+ chains). Unified liquidity pool means the destination receives the canonical asset. Good fallback when Across or deBridge don't support your route. In December 2025, Stargate partnered with Ondo Finance for cross-chain transfers of 100+ tokenized assets including stocks and ETFs.

Trade-offs

Post-acquisition, Stargate's bridge trust model is layered directly on LayerZero's DVN infrastructure. The April 2026 Kelp DAO exploit ($292M) exposed that LayerZero DVN configurations with a 1-of-1 verifier are catastrophically insecure — 47% of active LayerZero OApp contracts were running this setup at the time. LayerZero Labs has since committed to migrating all defaults to 5/5 DVN where possible and no less than 3/3. Before bridging large amounts via Stargate, verify the route's DVN configuration. For high-value transfers, prefer CCTP for USDC and Across for ETH.

Audit history & trust

Multiple audits via the LayerZero ecosystem. LayerZero has been live since 2022 with no direct infrastructure exploit — but the Kelp DAO case is a cautionary tale about configuration defaults. The fee switch activated December 2025 directs all ZRO ecosystem fees to token buybacks and burns; by February 2026, all message fees convert to ZRO and are burned.

Cost

0.06–0.12% fee plus gas. ~30s with V2.

deBridge — deep dive

deBridge's DLN intent layer routinely undercuts Across on fees by a couple of basis points, and it covers Solana and Tron — which increases its relevance for stablecoin corridors. Zero security incidents since 2022 launch. 30+ audits on record; $200k bug bounty unclaimed.

→ Bridge on deBridge — referral link, supports this site at no extra cost.

Best for

EVM-to-EVM transfers when you want the cheapest intent quote, and EVM-to-Solana or EVM-to-Tron. November 2025 volume exceeded $1.53B monthly, with 40% of that routing through Tron's USDT reserves. Aggregators (LI.FI, Bungee) route through deBridge when it wins on price. TRON DAO integrated deBridge's MCP server in April 2026 for AI-agent cross-chain execution.

Trade-offs

Smaller relayer set than Across, so quote quality can vary more by route. Brand recognition is lower; some risk teams treat it as a tier-two venue despite a clean security record.

Cost

0.04–0.08% fee plus gas. 30–60 seconds typical.

Wormhole + Mayan — deep dive

Wormhole is the messaging layer most Solana bridges sit on. Mayan Finance (Mayan Finance is a cross-chain swap and bridge layer built on Wormhole specifically optimized for Solana ↔ EVM routes) is the user-facing front end most users should actually use.

Best for

Solana ↔ EVM routes. Wormhole's 19-node guardian set verifies cross-chain messages. Wormhole added XRP Ledger support on June 26, 2025, bringing supported networks to 35+. The Native Token Transfers (NTT) framework, shipped in late 2025, enables tokens to transfer natively without wrapping — over 100 tokens on 40+ chains via NTT. Mayan Finance has processed $18B+ in total volume with 8M swaps and 3M unique wallets; average transfer time 18 seconds.

Trade-offs

Wormhole was hacked for $325M in February 2022 via a signature-verification flaw — Jump Crypto refunded users, but there is no contractual guarantee they'd do so again. The 19-node guardian set is committee-based, not cryptographically trust-minimized. For sizes above $50k on Solana routes, split into multiple transfers.

Cost

Varies by route. Mayan typically clears 0.05–0.15% all-in on Solana-EVM transfers.

Hop Protocol — deep dive

Hop is the oldest AMM-based EVM L2 bridge still operating, but it is losing ground to intent-based alternatives. TVL is approximately $3.8M as of early 2026 — a fraction of Across or Stargate. Still operational and maintained (v2 monorepo updated January 23, 2026), but not competitive on fees or speed for most routes.

Best for

Legacy use cases, or routes where Across and Stargate don't have relayer coverage.

Cost

0.10–0.40% fee plus gas. 1–10 minutes typical.

Best bridge by use case

CCTP V2 for USDC, Across for ETH between L2s, Wormhole or Mayan for Solana, IBC or Squid for Cosmos, third-party bridges to skip the 7-day L2 withdrawal, official bridges for first L1→L2 deposits. Last verified: 2026-05-27.

• Best bridge for USDC — Circle CCTP V2 (zero protocol fee, native burn-and-mint, 17+ chains). Use V2 only — V1 deprecated July 31, 2026.
• Best bridge for ETH between L2s — Across V4 (intent-based + ZK, ~30–90s, $34B+ cumulative volume with zero exploits).
• Best bridge for Solana ↔ EVM — Mayan Finance (18s average, $18B+ volume) or deBridge (DLN intent, covers Tron too).
• Best bridge for Cosmos ecosystem — IBC native or Squid Router.
• Best bridge to skip the 7-day L2 withdrawal — Across or Stargate (third-party bypass).
• Best bridge for first deposit L1 → L2 — The official L2 bridge (Arbitrum Bridge, Base Bridge) — gas-only, no fee.
• Best bridge aggregator — LI.FI or Bungee (compares quotes across bridges in one interface).
• Best bridge for large amounts ($25k+) — Split into multiple smaller transfers across CCTP + Across. For any LayerZero-based route, verify DVN config is at minimum 3/3.
• Best bridge for new chains — Whatever the chain's official integration lists. Avoid bridges with under 6 months of audits.
• Best bridge to avoid entirely — Multichain (defunct), any lock-and-mint bridge with no 2024+ audit, any LayerZero OApp with a 1-of-1 DVN configuration.

How to choose a bridge

Decide on three axes: asset type (CCTP V2 for USDC, Across for ETH, Mayan/deBridge for Solana), trust assumption (CCTP lowest, lock-and-mint highest, committee-based middle), and required speed (sub-minute for intent bridges, 1–10 min for AMM/Hop). Last verified: 2026-05-27.

By asset type:

• USDC: Always use CCTP V2 first — zero bridge fee, native USDC (not wrapped), 17+ chains.
• ETH or majors on EVM: Across V4 or Stargate (verify Stargate DVN config for high-value transfers).
• Solana ↔ EVM: Mayan Finance or deBridge DLN.
• Cosmos ecosystem: IBC native, or Squid Router for cross-chain into Cosmos.

By trust assumption:

• Lowest trust: CCTP V2 (Circle is the USDC issuer — no extra trust needed beyond holding USDC).
• Low trust: Across V4 (UMA optimistic oracle, ZK-proof verification, well-audited, $34B+ volume no exploit).
• Medium trust: Wormhole (19-node guardian committee), Stargate (LayerZero DVN — risk depends on configuration, minimum 3/3 post-Kelp).
• Higher trust / avoid: Lock-and-mint bridges, any bridge with 1-of-1 or single-verifier configuration, any new/unaudited bridge.

By speed:

• <30 seconds: CCTP V2 Fast Transfer
• 30–90 seconds: Across, deBridge, Stargate
• 1–10 minutes: Hop, Wormhole standard
• >10 minutes: Anything using optimistic finality without speed-up, or legacy bridges on Ethereum finality

Step-by-step: bridging safely

Aggregator quote → URL check → wallet connect → quote breakdown review → destination address verify → approve + sign → status track → revoke approval after use. Last verified: 2026-05-27.

1. Pick the right bridge for your asset and route. Use a bridge aggregator like LI.FI or Bungee's multi-bridge quote tool to compare quotes from multiple bridges in one place.
2. Verify the bridge's official URL. Phishing fakes of bridge frontends appear above real results in Google Ads routinely. Bookmark official domains.
3. Connect your wallet, select source chain → destination chain → token → amount.
4. Check the quote breakdown: bridge fee, gas, expected receive amount, ETA.
5. Confirm the destination address — your own wallet on the destination chain. Never bridge to an exchange deposit address unless that exchange explicitly supports the bridge.
6. Approve the token (one-time per token+bridge). Then sign the bridge transaction.
7. Track the transaction on the bridge's status page or via a cross-chain message explorer like LayerZeroScan.
8. After arrival, revoke approval if it's a bridge you won't use again.

Bridges to avoid in 2026

Multichain (defunct, $1.2B frozen), any lock-and-mint bridge without a 2024+ audit, any LayerZero OApp running a 1-of-1 DVN setup, anything requiring a "deposit address" instead of a smart contract, anonymous teams. Last verified: 2026-05-27.

• Multichain (formerly Anyswap) — collapsed mid-2023 after the CEO disappeared and Chinese authorities reportedly detained him. $1.2B in user funds frozen indefinitely. Still listed by some aggregators and chain documentation; do not use under any circumstances.
• Old lock-and-mint bridges with no recent audits. If a bridge hasn't been audited in 18+ months, assume it's untrusted. The Nomad team thought they were safe right up until the August 2022 exploit drained them in a free-for-all.
• Any LayerZero OApp with a 1-of-1 DVN configuration. The April 2026 Kelp DAO exploit drained $292M precisely because a single compromised verification node could approve phantom withdrawals with no check. LayerZero Labs admitted fault and is migrating defaults to 5/5 — but the responsibility to verify sits with you. If a bridge won't disclose its DVN setup, treat it as 1-of-1 and avoid for any significant amount.
• Any bridge that requires you to send tokens to a "deposit address" instead of using a smart contract. That's a 2018-era pattern and almost always a scam — a real bridge interacts with your wallet via signed transactions.
• Anonymous teams — a bridge holds your funds in custody for the settlement window. Anonymous custody is unacceptable risk. Doxxed teams can be sued; pseudonymous teams can vanish.
• "Bridges" that are actually centralized swaps — some products labeled "bridge" are really a custodial exchange in a trench coat. If the UI asks you to "deposit and wait," not "approve and bridge," you're on a custodial venue.

How to actually use a bridge (the workflow)

1. Open an aggregator first. LI.FI and Bungee compare quotes from Across, Stargate, deBridge, CCTP, Hop, and others side-by-side. You'll occasionally find a 20-basis-point spread you wouldn't have caught using one venue.
2. Sanity-check the winning quote by opening the underlying bridge directly. Aggregators add a thin fee; for transfers over $10k it's often cheaper to go direct once you know who's winning.
3. Verify the URL. Bookmark the official domains; phishing copies of bridge frontends are common and often appear above the real result in Google Ads. Across is across.to, Stargate is stargate.finance, CCTP UI is bridge.circle.com or any wallet/aggregator that integrates CCTP V2 natively.
4. Pre-fund gas on the destination chain. If you're bridging USDC to a chain where you have no ETH for gas, you'll arrive with no way to move. Either bridge a small ETH amount first, or use a bridge with built-in gas-drop (Across, Stargate, deBridge all support this).
5. Confirm the destination address. Send to your own self-custody wallet. Never bridge directly to an exchange deposit address unless the exchange explicitly supports the bridge — many exchanges credit only on the canonical chain and your bridged tokens vanish into support-ticket purgatory.
6. Approve once, bridge twice if needed. First-time bridges require a one-time ERC-20 approval; sign that, then the actual bridge transaction. For repeated use, set a finite approval rather than infinite.
7. Track via the bridge's status page or LayerZeroScan for LayerZero-based bridges. Don't refresh the wallet view in panic; cross-chain confirmations take longer than same-chain.
8. Revoke approvals after one-time use via Revoke.cash. Every dormant approval is a potential exit if the bridge contract is ever compromised.

Bridge fee math (real example)

On a $5,000 USDC Ethereum→Base transfer, CCTP V2 delivers ~$4,995.90 (zero protocol fee, gas only), Across ~$4,990.90, Stargate V2 ~$4,989.40. CCTP V2 wins for USDC; Across wins for ETH and other tokens. Last verified: 2026-05-27.

Bridging $5,000 USDC from Ethereum mainnet → Base (estimates; use an aggregator for live quotes):

Winner for USDC: CCTP V2. For ETH or other tokens: Across.

Risk summary

Bridges remain the highest-risk DeFi category by historical loss. $328.6M in bridge exploits were recorded through mid-May 2026 alone, led by the $292M Kelp DAO hack. Even with intent-based architectures, custody risk during the settlement window, DVN configuration risk, smart-contract bugs, and phishing remain real. Last verified: 2026-05-27.

• DVN configuration risk (2026's new threat vector). The Kelp DAO exploit — $292M, April 18, 2026 — was not a smart-contract bug. Lazarus Group compromised LayerZero's RPC infrastructure feeding a 1-of-1 verifier, DDoS'd external nodes, and fed phantom burn events to release real funds. LayerZero Labs acknowledged fault and is migrating all defaults to 5/5 DVN. Until that migration is verified complete and covers your route, treat any LayerZero OApp with unknown DVN configuration as high-risk for large amounts.
• Smart-contract risk. Even audited bridges have been hacked — Nomad's $190M exploit in August 2022 was a one-line bug in a function that had been audited multiple times. Treat every bridge contract as a potential single point of failure; never leave large balances approved to the bridge.
• Liquidity-pool / relayer risk. Intent bridges depend on relayers being capitalized and competitive. In a tail event (chain outage, gas spike), relayer participation thins and quotes widen. Stargate's V2 pools have specific liquidity gating in low-utilization conditions — read the docs before bridging niche routes.
• Issuer / wrapped-asset risk. Wrapped tokens minted by older lock-and-mint bridges are IOUs against source-side custody. Most Wormhole-wrapped assets from pre-2022 still trade at par because Jump refunded the hack — there's no contractual guarantee they'd do it again.
• Phishing risk. Fake Across, fake Stargate, and fake LI.FI sites regularly appear above official results in Google Ads on bridge-related keywords. Bookmark official URLs and connect from your bookmark, not from search.
• Regulatory risk. Bridges that don't enforce sanctions screening have drawn OFAC attention since the Tornado Cash precedent. CCTP screens addresses at the Circle layer; most intent bridges rely on per-chain enforcement.
• Operator / multisig risk. Most bridges still have an upgrade key with timelock. A compromised multisig could swap the contract for a malicious one after the timelock elapses. Check the bridge's upgrade controls before sizing.

Looking ahead to 2027

• CCTP coverage expansion — Circle has been adding chains consistently. CCTP V1 phase-out begins July 31, 2026, which forces integrators to V2's faster-finality and hooks design. If CCTP reaches every major L2, USDC stops needing third-party bridges entirely for that asset.
• ERC-7683 as the new default intent API — Ratified in early 2025 and backed by the Ethereum Foundation's Open Intents Framework (30+ teams, including Arbitrum, Optimism, Polygon, zkSync). 88% of Across volume already runs ERC-7683. Expect wallets and aggregators to surface "intents" as the standard user path within 12 months.
• ZK light-client bridges going mainstream — Across V4 added ZK proofs for non-EVM chain verification in July 2025. Succinct's SP1 and Hyperlane are shipping ZK-verified cross-chain messaging. If gas costs decline further, cryptographically-verified bridges make committee-based models (LayerZero DVN, Wormhole guardians) look dated.
• Solana ↔ EVM convergence — Mayan ($18B+ volume, 18s average), deBridge (Tron + Solana + EVM), and Wormhole's NTT framework all target one-click Solana ↔ EVM in under 30 seconds with stablecoin-level fees. Across V4 Solana launched August 2025 adds a fourth serious competitor.
• LayerZero centralization questions — The Kelp DAO exploit renewed debate about LayerZero's DVN security model. LayerZero's planned Zero Chain L1 mainnet (Fall 2026) and strategic investments from Citadel Securities, ARK, Tether, DTCC, and ICE signal institutional ambition — but the 1-of-1 default configuration problem isn't fully resolved in May 2026.

Verdict

Default to Across V4 for EVM. Use CCTP V2 for USDC (V1 deprecated July 2026). Use Mayan Finance or deBridge for Solana. For any LayerZero-based route, verify DVN config is 3/3 or higher before bridging significant amounts. Use LI.FI or Bungee to comparison-shop before committing. Last verified: 2026-05-27.

• Default: Across V4 (intent + ZK, 18+ chains, $34B+ volume, zero exploits, now covers Solana).
• For USDC: CCTP V2 (zero protocol fee, native USDC, 17+ chains, 8–20s Fast Transfer). Do not use V1 — deprecated July 31, 2026.
• For Solana routes: Mayan Finance (18s average, $18B+ volume) or deBridge DLN.
• For breadth: Stargate (80+ chains) — but verify DVN configuration for high-value transfers post-Kelp.
• For comparison shopping: LI.FI or Bungee aggregators.
• Never use any bridge you can't find an audit for from 2024+, with a doxxed team, and a recent commit history. Never use a bridge whose LayerZero DVN setup is 1-of-1.

---

Related: Best Ethereum L2s 2026 · Best Bitcoin L2s 2026 · Best Stablecoins 2026 · Best DEXs 2026 · DeFi Yield Farming Guide 2026

This guide is updated as bridge protocols change fees, security models, and chain coverage. Last review: May 2026.