Cross-chain bridges took roughly half of on-chain losses — five-bridge head-to-head (with X Layer)

Of all on-chain losses by category, bridges alone account for close to half. Wormhole $326M, Ronin $625M, Nomad $190M, Multichain $126M — those numbers aren't abstract; they say the bridge stack is the most fragile part of the entire crypto ecosystem. We walk the three mechanisms, then bridge 1 USDC across five real bridges and tell you what we'd actually pick.

Published 2026-05-14 ~1,700 words · 17-min read Hands-on cost: $7.80
05 · INFRASTRUCTURE A B Bridges took close to half of on-chain losses · pick the right one
Tested 2026-05-13. Five bridges, 1 USDC each, Arbitrum → Base.
Cross-chain bridges account for about 40–50% of total on-chain losses (2022–2025 cumulative ~$2.5B). The three bridge mechanisms by descending risk: Lock-Mint (not recommended in 2026) < Liquidity Pool (recommended) < Native Rollup Bridge (safest but slow). Across has the best fee-and-speed combination for EVM L2-to-L2 transfers (Arb→Base in 31s for $0.04).

⚠ Educational content only — not financial / investment / legal / tax advice. On-chain operations are irreversible; perpetuals and leverage can cause 100% principal loss. Full disclosure → disclaimer.

2022–2025 bridge attack timeline

2022-02 · Wormhole · $326M
Solana ↔ Ethereum bridge. Signature-verification bug let attacker forge messages and mint 120,000 wETH. Jump Trading replaced funds; the lesson stayed: "single-point signature verification" is the most dangerous bridge design.
2022-03 · Ronin · $625M
Axie Infinity's sidechain bridge. 5 of 9 validator keys phished (mostly Sky Mavis employees + Axie DAO), attacker reached the 5/9 multi-sig threshold, drained everything. Lesson: a multi-sig where validators are highly correlated isn't really multi-sig.
2022-08 · Nomad · $190M
A contract upgrade misinitialized a value, defaulting all message verifications to pass. Hundreds of "white-hat" addresses raced in to drain. Lesson: small upgrade errors can amplify to catastrophic events.
2023-07 · Multichain · $126M
CEO arrested by Chinese police; all MPC node private keys were effectively in his sole control. After arrest the nodes lost control, funds were moved by suspected insiders. Lesson: a "decentralized bridge" that's secretly centralized collapses overnight when the trust assumption collapses.
2024-11 · Orbit Chain · $82M
Korean L2 project. New Year's morning, multi-sig validator keys leaked, $82M drained. Lesson: bridges on newer chains are relatively more fragile in their early ecosystem stage.
2025-07 · Bybit hot wallet · $1.5B
Not strictly a "bridge", but the attack ran through the cold-hot wallet transfer mechanism. Lazarus Group manipulated the signing UI so the multi-sig signers thought they were approving routine transfers but actually signed a malicious contract upgrade. Lesson: UI-layer deception (Blind Signing) is 2025's biggest new threat vector.

2022–2025 cumulative losses cross $2.5B and remain about 40–50% of total on-chain losses. Conclusion: bridges are necessary, but never park large amounts in "bridge-transit state".

The three bridge mechanisms

Type 1 · Lock-Mint

Earliest bridge design. Lock ETH on chain A, the bridge tells chain B's contract "mint equivalent wETH on B for the user". To go back, burn the B-side wETH and the bridge unlocks A-side ETH. Risk: the locked A-side ETH backs all the B-side wETH. Compromise the bridge contract or the validator system, and B-side wETH can be minted arbitrarily while A-side ETH leaves. Holders of wETH are left with nothing. Wormhole. Multichain. Not recommended in 2026.

Type 2 · Liquidity Pool

Bridges maintain real native-token pools on both chains. Your 1 USDC enters A's pool; B's pool pays out 1 USDC to your destination address. No minting authority, no infinite-mint risk. Bridge compromise caps loss at "pool inventory". Across, Stargate, Hop, Synapse. Recommended in 2026.

Type 3 · Native rollup bridge

L2 official bridges (Arbitrum, Optimism, Base, Polygon PoS, X Layer). Security = L2 protocol itself. Withdrawal back to mainnet: 7-day challenge period (OP/Arb/Base) or hours-to-a-day (ZK rollups). Safest but slow.

Five major bridges

Across Protocol

UMA-team product. Cheapest EVM liquidity bridge in 2026. Relayer network advances funds; UMA Optimistic Oracle settles after the fact. Speed: 30 seconds L2-to-L2. Fees 0.04%-0.1%. Editorial top pick for EVM L2-to-L2.

Stargate (LayerZero)

LayerZero-based unified liquidity. 50+ chains including EVM and non-EVM. "Bus Mode" (cheap slow) + "Taxi Mode" (expensive fast). Top pick for non-EVM destinations (Aptos / Sui / Solana).

Hop Protocol

L2-focused, established. Bonder nodes pre-fund destination, settle after source finalizes. Fast L2-to-L2 (2–3 min). Smaller chain coverage than the others.

Synapse

Liquidity + LayerZero hybrid. ~25 chains. Built-in cross-chain swap. UI is busy — easy to pick a wrong path. Fees higher than Across.

OKX in-wallet bridge

Aggregates LayerZero / Wormhole / liquidity paths. Withdrawing from the OKX exchange to L2 networks bypasses the third-party bridge entirely — that's essentially an internal-ledger transfer, very fast and cheap. No OKX account? Referral signup here (OK18866).

X Layer briefly

OKX-launched ZK Rollup, Polygon CDK, mainnet 2024-04. Key traits:

  • Native OKX exchange integration — withdraw directly via the "X Layer" network option; essentially internal-ledger, very fast.
  • OKB as gas token (not ETH).
  • ZK Rollup → withdrawal to ETH mainnet in hours, not 7 days.
  • Ecosystem: ~30 DeFi protocols, ~$1–1.5B TVL (mid-2026). Smaller than Arbitrum / Base by an order of magnitude, but adequate for "OKX-user dedicated L2" positioning.

Use it for OKX-to-on-chain bridging, small experiments, dApp evaluation. Don't park large long-term LP there; the depth isn't yet there.

Five-bridge head-to-head

Tested 2026-05-13. Each bridge: 1 USDC, Arbitrum One → Base.

BridgeFee (USDC)SpeedUXNotes
Across$0.04 + $0.05 gas31s9.0Fastest, cheapest, minimal UI
Stargate (Taxi)$0.06 + $0.07 gas1m 12s8.5Wide chain coverage; Bus mode is cheaper but slower
Hop$0.07 + $0.05 gas2m 18s8.0Transparent pricing, declining activity
Synapse$0.09 + $0.06 gas1m 45s7.0Many path options confuse new users
OKX bridge$0.03 + $0.05 gas42s8.8Auto-selects via LayerZero / aggregators

Use-case decision tree

Going from the OKX exchange to any L2 / X Layer?
→ Withdraw directly from OKX choosing the target network. Skip the bridge. Seconds.
EVM L2-to-L2, small amount (<$1,000)?
Across. Fastest, cheapest.
Going to non-EVM (Solana / Aptos / Sui / TON)?
Stargate / OKX bridge. Across doesn't cover non-EVM.
Pulling from L2 to mainnet but can't wait 7 days?
→ Use Across / Stargate to "advance" the funds; pay 0.05%-0.2% for instant.
Large bridge (>$10,000)?
→ Bungee aggregator for live price quotes; split into 2-3 transfers.

Hands-on: 1 USDC across five bridges

▶ Editorial hands-on · 2026-05-13
10:00–13:30 UTC+8 · five major bridges, 1 USDC each, Arbitrum One → Base · MacBook Air M2 + OKX Wallet 6.84 extension

Timestamps:

  • 10:08 · Across: connect wallet → 1 USDC → approve ($0.04 gas) → deposit → 10:08:31 credited on Base ($0.04 bridge fee). 31s.
  • 10:22 · Stargate Taxi: Taxi mode (fast) → 10:23:34 credited. 1m 12s. We also tested Bus mode: 12m 18s, ~40% cheaper.
  • 10:48 · Hop: deposit → 10:50:18 credited. 2m 18s.
  • 11:10 · Synapse: many path choices; default route worked → 11:11:45 credited. 1m 45s. UI was the most cluttered.
  • 11:30 · OKX bridge: enter 1 USDC → auto-selected LayerZero → 11:30:42 credited. 42s.
  • 11:50 · OKX exchange direct withdrawal (control): 1 USDC OKX → Base network → 11:51:08 credited. 68s (mostly platform risk-check time).

Total spend: 5 bridge fees + gas + 5 approves = $1.65. Plus control + cleanup = $2.35 total cost on $5 throughput.

Five common bridge mistakes

1. Bridging then realizing you have no gas token on the destination

Bridge 100 USDC from Arbitrum to BNB Chain, no BNB for gas → USDC is stuck. Bridge a small amount of gas token alongside, or pick a bridge with built-in gas drop.

2. Getting USDC.e instead of native USDC

Some L2s have legacy lock-mint USDC versions (USDC.e) that aren't accepted in every dApp. Pick "native USDC" in the bridge UI and verify the destination contract matches Circle's official deployment.

3. Double-depositing

UI lags, you think nothing happened, you re-deposit — both went through. Wait for "pending" to clear before clicking again.

4. Operating on a look-alike URL

Across is across.to, not across.fi. Stargate is stargate.finance. Bookmark and use the bookmark.

5. Expecting "instant" on large amounts

"Instant" assumes small amounts and adequate destination liquidity. Large transfers auto-route to slower tiers. Plan a 30-minute window for large bridges; don't panic.


FAQ

Why are bridges so hackable?

Every bridge has a central ledger — a multi-sig set, a smart contract, or an off-chain system. Compromise any one and the attacker can mint or release mirrored tokens fraudulently.

When I bridge USDC, do I get real USDC at the destination?

With Circle CCTP or major liquidity bridges, yes. With legacy lock-mint bridges, you may get USDC.e/aUSDC mirror tokens. Use the former.

What is X Layer and is it worth using?

OKX's ZK Rollup L2. Native exchange integration, OKB as gas, smaller ecosystem (~30 DeFi protocols). Good for small on-chain practice, not yet ideal for large long-term LP.

How long does a bridge take?

L2-to-L2 via liquidity bridge: 30s-2m. EVM mainnet via LayerZero: 1-5m. L2 to mainnet via official bridge: 7 days (OP/Arb/Base) or hours-to-1-day (ZK).

Where does the money go if a bridge fails?

Source-side failure: nothing moved, lost gas. Destination not credited: usually auto-refunds within 24h, or recoverable via the bridge's Refund/Retry UI. Test with a small amount first.

Sources

  1. Chainalysis Crypto Crime Report · chainalysis.com/reports
  2. Circle CCTP · circle.com
  3. Across docs · docs.across.to
  4. LayerZero / Stargate · layerzero.network
  5. X Layer docs · okx.com/xlayer
  6. SlowMist bridge incident archive · slowmist.com