How cross-chain swaps work

A cross-chain swap exchanges an asset on one blockchain for an asset on another — native BTC for native SOL, say — even though the two chains cannot read or verify each other. Every implementation on the market is one of three architectures: lock-and-mint bridges that issue wrapped IOUs, native settlement networks that hold real inventory on both chains, and RFQ or intent-based desks where a counterparty fronts the destination asset. Which one your route uses determines who holds your money mid-flight, what you pay, and how you lose funds if something breaks.

The problem: chains cannot verify each other

Bitcoin, Ethereum, and Solana are separate ledgers with separate validator sets. A Bitcoin node has no way to check the state of Ethereum, and vice versa. There is no shared settlement layer, so a “swap” across chains is really two independent settlements on two ledgers that know nothing about each other, plus some mechanism that coordinates them. The entire design space of cross-chain infrastructure is a set of answers to one question: who do you trust while your value is in flight between ledger A and ledger B?

The original answer was a centralized exchange — deposit BTC, sell it, buy SOL, withdraw. That works, but it means an account, full custody with a third party, and balance-sheet risk for the duration. Everything below is about doing the same job without handing over a deposit. If your actual question is whether to move an asset between chains versus change it into another asset, start with our cross-chain swap vs bridge comparison — the two get conflated constantly and the risk profiles differ.

Architecture 1: lock-and-mint bridges and wrapped assets

The oldest pattern. You lock the asset in a contract (or with a custodian) on chain A, and the bridge mints a wrapped representation on chain B — WBTC on Ethereum is the canonical custodial example. The wrapped token is not the asset; it is an IOU redeemable against the locked collateral, and it is exactly as good as the custody arrangement behind it. Verification of the lock event is done by whatever the bridge uses as its oracle: a multisig, an external validator set, or (rarely, because it is hard) an on-chain light client.

The defining property of this architecture is that the risk does not end when the swap ends. As long as you hold the wrapped asset, you hold exposure to the bridge’s validator keys and contract code. If the collateral is stolen, the IOU depegs, no matter how long ago you bridged. Lock-and-mint is how most liquidity historically moved into new ecosystems, and it is also where the largest losses in crypto infrastructure happened — more on that below. For a route-level walkthrough of a typical EVM-to-non-EVM leg, see ETH to SOL bridge options.

Architecture 2: native cross-chain settlement

The second pattern removes wrapped assets entirely. A settlement network runs vaults on every chain it supports, holding real inventory — actual BTC on Bitcoin, actual ETH on Ethereum. You send the native asset into the vault on chain A; the network observes the deposit, prices the swap against its liquidity pools, and pays you the native asset out of its vault on chain B. You start and end with real assets; nothing wrapped ever touches your wallet. The vaults are controlled by threshold signatures across bonded node operators, so the trust assumption shifts from a custodian to validator economics — ideally the value bonded by operators exceeds the value sitting in the vaults they control.

Three networks matter in this category as of mid-2026. THORChain, the largest, settles native swaps across roughly a dozen chains — Bitcoin, Ethereum, BNB Chain, Avalanche, Cosmos, Dogecoin, Litecoin, Bitcoin Cash, and Solana (re-added in early 2026) among them. It also just demonstrated the category’s risk profile: an exploit of roughly $11 million in May 2026 paused the network for 39 days before the v3.19 restart, which put native Monero support at the front of its integration queue. Maya Protocol, a friendly fork of THORChain, runs the same vault model over a different chain set — Bitcoin, Ethereum, Dash, and Arbitrum, with Zcash added as a privacy corridor and Cardano in development. Chainflip takes a different pricing approach (a just-in-time AMM) and settles native BTC, ETH, SOL, and major stablecoins across Bitcoin, Ethereum, Solana, Arbitrum, and Polkadot’s Asset Hub. Coverage per protocol is narrow by design — every added chain is a heavy engineering and security lift — so these networks carry the majors, not the long tail. For what this means for privacy-asset routing specifically, see the Monero swap guide and BTC to XMR routes.

Architecture 3: RFQ desks and intent-based systems

The third pattern is the simplest and the fastest growing: a counterparty just fills you from its own inventory. In the classic instant-exchange (RFQ — request for quote) model, you request a quote for A→B, send asset A to the desk’s address, and the desk sends asset B to yours. No account, no deposit held on platform, no pool mechanics — the desk prices its inventory and hedging cost into the quote. Because coverage is set by what the desk is willing to hold rather than by protocol integrations, these venues carry the broadest long tail, and aggregators of such desks commonly quote assets the settlement networks historically could not, including Monero. The trade-off is equally simple: mid-flight, the desk is your counterparty, and counterparty selection is the whole game — eXch, one of the larger no-account desks, was shut down by Germany’s BKA in April 2025 over money-laundering allegations, with about €34 million seized.

Intent-based systems generalize this into open competition. On NEAR Intents — the most prominent implementation, spanning 30-plus chains and past $20 billion in cumulative volume by June 2026 — you sign an intent (“swap 1 ETH on Ethereum for USDC on Solana”) and a network of solvers competes to fill it, with settlement enforced by a verifier contract rather than by trusting any single desk. Aggregators such as LI.FI now route through it alongside bridge and native-settlement backends. On privacy assets, our framing is deliberate: the legitimate case for these corridors is financial privacy — payroll confidentiality, commercial positioning, personal safety — and whatever route you use, you remain responsible for complying with the laws and tax obligations of your jurisdiction.

The three architectures, compared

How to read a row → Native settlement · vault holds for one swap · ~1–30 min: protocol vaults custody your funds only for the duration of the swap and settle in native assets, so your trust window closes the moment it lands — unlike a lock-and-mint IOU that stays live as long as you hold it.

Where funds actually get lost

The failure history of this category is dominated by bridges. Ronin (March 2022): attackers compromised five of nine validator keys and withdrew roughly $620 million in ETH and USDC from the bridge backing Axie Infinity’s sidechain — a pure key-custody failure. Wormhole (February 2022): a signature-verification flaw let an attacker mint 120,000 wETH — about $325 million — without collateral; Jump Crypto recapitalized the bridge to keep the IOUs whole. Note what failed in each case: not the swap UX, but the collateral layer that wrapped assets silently depend on.

Native settlement is not immune — THORChain’s ~$11 million exploit and 39-day halt in mid-2026 is the live example — but the blast radius differs: your exposure is one swap’s duration, not an open-ended IOU. RFQ desks fail differently again: the desk can misquote, disappear, or be seized (eXch). A working taxonomy of loss:

• Contract bugs — minting or verification flaws (Wormhole).
• Key compromise — validator or multisig keys stolen (Ronin).
• Economic failure — collateral shortfall depegs the wrapped asset even without a hack.
• Counterparty failure — an RFQ desk does not deliver, or delivers late into a moved market.
• Operational and legal shutdown — infrastructure seized or wound down mid-corridor.

The desk rule that falls out of this: size the position to the trust model, not the interface. A route that looks identical in a swap widget can sit on any of the three architectures underneath.

Fees, quotes, and how aggregators pick routes

Every cross-chain quote is built from the same four parts, whatever the marketing says. Network fees: you pay gas on the source chain in, and the route pays gas on the destination chain out — the outbound fee is deducted from what you receive, which is why small swaps price so poorly. Liquidity fee: on pool-based routes, your size moves the pool — THORChain-style slip-based fees charge you more the larger you are relative to pool depth. Spread: RFQ desks and solvers price inventory risk and hedging cost into the quote itself; the fee is invisible but always there. Affiliate fees: most interfaces and aggregators add basis points on top for order flow.

Aggregators exist because no single backend wins everywhere. They query bridges, native settlement, and RFQ desks for the same pair, normalize everything to a single number — destination amount after all fees — and rank routes, ideally also weighing speed and failure behavior. The number that matters when you compare venues is not the headline rate but the delta between quoted and realized: floating quotes can slide between quote and fill. That is what “guaranteed minimum received” fixes — the quote commits to a floor on the destination amount, and if the route cannot clear at or above the floor, the swap refunds rather than filling worse. It converts open-ended slippage into a binary fill-or-refund outcome, which is the property a desk actually wants when moving size. The strongest routers are built on this model — cross-chain swaps quoted with a guaranteed minimum received, the floor on what arrives before you commit.

FAQ

What is a cross-chain swap?

A cross-chain swap exchanges an asset on one blockchain for an asset on another — for example native BTC for native SOL — without a centralized exchange account. Under the hood it is always two settlements on two ledgers, coordinated by a bridge, a native settlement network, or a market-making desk.

What is the difference between a cross-chain swap and a bridge?

A bridge moves a representation of the same asset between chains, typically by locking it and minting a wrapped IOU. A cross-chain swap changes both the chain and the asset, and the better implementations settle in native assets on both sides, so you never hold a wrapped claim.

Are cross-chain swaps safe?

The mechanics are mature, but each architecture concentrates risk somewhere: bridges in the custody of locked collateral, native settlement in validator economics, RFQ desks in the counterparty. The two largest losses in the category — Ronin (about $620M) and Wormhole (about $325M) — were both bridge failures. Size positions to the trust model, not the interface.

How long does a cross-chain swap take?

Roughly the confirmation time of the slower chain plus routing overhead. Intent and RFQ routes that front inventory can settle in seconds to a few minutes; native settlement involving Bitcoin typically takes 10–30 minutes because the protocol waits for confirmations before releasing the outbound asset.

What does guaranteed minimum received mean?

It is a floor on the destination amount written into the quote. If the route cannot clear at or above that floor, the swap fails or refunds instead of filling at a worse rate — turning open-ended slippage risk into a binary fill-or-refund outcome.

The TT desk thoughts

The market is quietly repricing the three architectures, and the direction is one-way. Lock-and-mint had first-mover liquidity, but every year of exploit history compounds against a model where risk outlives the transaction — we treat wrapped assets as unsecured credit exposure to a validator set, and we hold them accordingly, which is to say briefly. Native settlement earns its slower speeds by ending the trade with real assets, but its chain coverage will always trail the market, and THORChain’s 2026 halt is a reminder that bonded-validator economics are a mitigation, not an absolution. Intents and RFQ, meanwhile, are winning flow for a plain reason: the trust window collapses to minutes, and competition between solvers prices the risk instead of a governance forum.

What we actually look for on a route, in order: does the quote commit to a floor (fill-or-refund beats any headline rate), how long is the custody window, and who eats the loss if the leg fails. Speed and fees come after. Most retail flow ranks those exactly backwards, which is why the bridge-hack graveyard is as large as it is.

Keep reading

Infrastructure hub · Monero swap guide · BTC to XMR routes · ETH to SOL bridge options · Cross-chain swap vs bridge · How perp DEXs work · Regulation · All notes