You have $5,000 of USDC sitting on Base and a better lending rate waiting on Arbitrum. You open a swap interface, pick the two chains, and it shows a quote: about 4,997 USDC out, arriving in seconds, no wrapped token, no seven-day wait. You sign one message. Thirty seconds later the USDC is in your Arbitrum wallet and you have paid a fee somewhere near $7.50.
Nothing was "bridged" in the sense most people picture. No vault opened, no IOU was minted, and no deposit is sitting on the far side waiting to be claimed. Instead, a professional market maker who already held USDC on Arbitrum handed you theirs, and a settlement layer quietly moved your Base USDC to reimburse them. You signed a request — an intent — and a private competition filled it. The entire exchange happened in a market you never saw.
That market is where your fees, your slippage and your MEV exposure actually live now. In 2026, intent-based routing has become a default for fast stablecoin and blue-chip moves between chains, and the numbers on your receipt are the output of an auction. This piece opens the hood: what an intent is, who the solvers are, how they get paid, how the two settlement paths differ, where your slippage and MEV go, and the failure modes that decide whether a good quote turns into a bad outcome.
TL;DR. An intent says what you want ("4,997 USDC on Arbitrum") rather than how to get it; solvers compete in an auction to fill it at the best net output. It feels like a bridge but is not one: an atomic filler fronts its own capital on the destination chain and is reimbursed on the source chain seconds later, so you hold native USDC, not a wrapped IOU. Because the order never touches a public mempool, the solver internalizes your slippage and MEV instead of a bot extracting it. You still carry new risks: solver inventory can run dry, an auction can win on a thin route, fills can be partial, refunds can lag, and you now trust the settlement layer. On $5,000 Base→Arbitrum, a classic canonical bridge cost about $4.60 but took days; an intent route cost about $6.00 net after a positive-slippage rebate and finished in seconds; a CEX round trip cost more and added KYC. Model the real net output in the DeFi yield calculator.
An intent is a request, not a transaction
A classic swap or bridge asks you to commit to a route. You approve a specific contract, on a specific pool, with specific calldata, and you broadcast that exact transaction to a chain. If the price moves before it lands, you eat the difference. If a bot sees it coming, it can sandwich you. You are specifying the how, and you wear every consequence of it.
An intent inverts that. You sign a message that states the outcome you will accept — deliver at least 4,997 USDC to this Arbitrum address, and I will give up 5,000 USDC on Base — and you stop there. The message is signed off-chain (the EIP-7683 cross-chain intent standard and EIP-712 signatures are the plumbing) and never becomes a public transaction until someone agrees to fill it. Whoever takes the other side chooses the route, the venues and the timing. You only judge the result.
That is the whole shift: from "execute this" to "achieve this." The consequence is that a competitive market now sits between your signature and your funds, and the quality of that market — not your gas settings — determines what you actually receive. Everything that follows is about how that market is priced.
Inside the solver auction: who bids, who wins, who pays
When you sign an intent, it goes to a solver network. The bidders are professional market makers with inventory on multiple chains, bridge liquidity providers, MEV searchers who turned negative-EV extraction into a service, and protocol-run solvers funded by the settlement layer itself. On a liquid stablecoin route like USDC Base→Arbitrum there might be a dozen or more; on an exotic long-tail pair, often one or two.
How they win: the auction is scored on net output to you, not on the lowest headline fee. A solver quoting 0.20% but sourcing the destination asset cheaply can beat one quoting 0.10% with a worse fill. Most systems run a Dutch auction — the price improves until a solver commits — or a first-price sealed bid. Typical solver fees in 2026 land around 0.05–0.15% on major stablecoin routes and widen past 0.3% on volatile or thin pairs.
What they are paid: the spread between the source asset they collect and the cost of the destination asset they deliver, plus any quoted fee, minus gas and inventory costs. Positive slippage — where the solver fills you better than your floor — is sometimes rebated to you as a sweetener. It is real money, but it is marketing as much as market outcome, and a rebate is never guaranteed on the quote.
Two settlement paths: atomic filler or lock-and-mint
Once a solver wins, how does the money actually move? Two architectures dominate, and they are not the same product.
Atomic filler (fast). The solver fronts its own capital on the destination chain immediately — you get native USDC from its Arbitrum wallet in seconds. Then, on the source chain, it collects your Base USDC to replenish that inventory, and the settlement layer enforces that both legs land or neither does. Your speed is bought with the solver's balance sheet, and the solver's exposure is inventory risk plus its trust in the settlement layer.
Lock-and-mint (slower). Your deposit is locked in a vault on the source chain and a matching token is minted on the destination — the classic bridge model. No solver capital is required, so it works for any pair, but you receive a claim, and a claim can depeg or stall if the vault pauses. This is the model behind most of the cross-chain bridge safety losses.
Where does your slippage go? In a classic bridge you broadcast a destination swap into a public mempool, and bots can front-run it — a plain MEV sandwich is exactly that. In an intent, the solver absorbs it: it routes privately, nets flows, and exposes you only to the fee it quoted. You traded a visible, unbounded MEV surface for a small fixed fee — usually a good trade, but a trade nonetheless.
Worked example: $5,000 USDC from Base to Arbitrum
Numbers make the difference concrete. Three ways to move $5,000 USDC from Base to Arbitrum in October 2026, all landing in a self-custodied Arbitrum wallet:
| Route | Fee | Gas (both sides) | Slippage | Net received | Time |
|---|---|---|---|---|---|
| Classic canonical bridge | 0.08% ($4.00) | $0.60 | None quoted | $4,995.40 | ~7 days (optimistic exit) |
| Intent route (solver) | 0.15% ($7.50) | $0.30 | +$1.80 rebate | $4,994.00 | ~30 seconds |
| CEX round trip | ~0.10% spread ($5.00) | $0.20 | $1.50 withdrawal | $4,993.30 | 10–30 min + KYC |
Read the table honestly. The classic canonical bridge is cheapest on paper at about $4.60, and if you actually want a canonical Ethereum exit it can take roughly seven days — fine for a treasury move, useless if you are chasing a rate today. The intent route is about $1.40 more expensive, but the positive-slippage rebate claws back part of the fee and it settles in seconds while keeping you in native USDC the whole time. The CEX round trip is the worst of both worlds: more cost, KYC, exchange counterparty risk, and an extra withdrawal step. The real choice is between the first two, and it is a price for time, not a safety contest.
Failure modes worth knowing before you size up
Fast routes hide their complexity until something breaks. These are the ways an intent can go wrong, roughly in order of how often they bite ordinary users.
- Solver runs out of inventory. The auction clears, then the winning solver's destination balance is short at fill time. Your intent waits or is re-auctioned, and the quote you saw is no longer guaranteed.
- An auction wins on a bad route. The best net output can come from a solver with the weakest settlement assumptions or a thin destination venue. Cheapest is not always the most robust route.
- Partial fill. Some systems let a solver fill only part of your order. You receive 60% at the quoted price and the remainder later — or never, if the rest expires.
- Refund delays. If a fill fails after you have committed funds, the return path can be slow and usually rides the same settlement layer that just failed you. Check the refund window before you size up; a "fast" route with a 24-hour refund is not fast when it matters.
- Settlement-layer trust. Atomicity is only as strong as the contract or committee enforcing it. You trust that layer for the seconds-to-hours your capital is in flight — a narrower window than a bridge vault, but not zero, and much of the same redemption-delay dynamic applies when a fill stalls.
Classic bridge vs intent routing at a glance
Different products for different jobs. Here is the side-by-side that matters for your money:
| Dimension | Classic bridge | Intent-based routing |
|---|---|---|
| Custody while in flight | Funds locked in a bridge vault or canonical contract; you hold a claim | Solver fronts its own inventory; native asset delivered from its balance |
| Failure mode | Vault exploit, pause, or wrapped-token depeg | Solver short on inventory, bad-route win, partial fill, refund lag |
| MEV exposure | Public mempool on both chains; destination swap can be sandwiched | Order flow is private and off-chain; solver internalizes the MEV |
| Refund path | Retry claim when the bridge resumes; may be delayed for days | Intent expiry refund, with a fallback to a canonical route |
| Typical cost | ~0.05–0.1% plus gas both sides; canonical is cheap but slow | ~0.05–0.30% solver fee plus gas; net rebate possible |
The pattern: intents win on speed, MEV exposure and the custody of the asset you actually hold; classic bridges win on simplicity, cost for large slow moves, and canonical-route safety. Neither is "safe" in the abstract — each concentrates risk in a different place, and the right pick depends on size and urgency. For how the slower route breaks down when it matters, the withdrawal queue mechanics are the piece worth reading before you commit a large amount.
Practical rules: size, read the refund terms, keep a fallback
Three rules cover most of the decision. Size to the route. Below roughly $10,000, an intent route's fixed solver fee and instant settle are usually worth the small premium; above that, the 0.1%-class savings of a canonical bridge start to matter and a few days of delay is often acceptable.
Read the refund terms before you sign. Find out what happens if the fill fails: is the refund automatic, how long does it take, and who holds your funds in between. A route with a clear fallback to a canonical bridge is safer than one that only promises speed.
Prefer routes with an obvious fallback. If the solver cannot fill, you want your intent to expire and your funds returned, not to sit in limbo. The best interfaces show the fallback up front.
You will not see the auction, but you can price it. Compare the net output against a canonical route, and treat the difference as exactly what you paid for time. On your next stablecoin move, which do you reach for — the instant intent or the slow canonical bridge?
Sources and further reading
- ethereum.org — Blockchain bridges — how bridges work, their trust trade-offs, and where intent-based designs fit.
- L2BEAT — Bridges — a risk-framework view of bridge designs: validation type, value secured and assumptions.
- Across Protocol documentation — how an intent settlement layer auctions fills and reimburses solvers on the source chain.
- UniswapX documentation — the off-chain intent model, filler competition and how settlement enforces the user's floor.
Frequently asked questions
What is a cross-chain intent?
A cross-chain intent is a signed, off-chain message that states the outcome you want — for example, deliver at least 4,997 USDC to this Arbitrum address in exchange for 5,000 USDC on Base — without specifying the route. Instead of broadcasting an exact transaction you control, you let competing solvers choose how to fill the request. If no solver fills it before the deadline, the intent expires and your funds are returned.
Who are solvers and how do they make money?
Solvers are professional market makers, bridge liquidity providers, MEV searchers and protocol-funded actors that hold inventory on multiple chains. They win an auction by offering you the best net output, then deliver the destination asset from their own balance and get reimbursed on the source chain through the settlement layer. They earn the spread between the source asset they collect and the cost of the destination asset they deliver, plus any quoted fee, minus gas and inventory costs. Typical fees run about 0.05–0.15% on major stablecoin routes.
Are cross-chain intents safer than bridges?
Not automatically — they concentrate risk differently. An intent with an atomic filler hands you the native asset rather than a wrapped IOU, and because your order never touches a public mempool, a solver absorbs your slippage and MEV instead of a bot extracting it. But you then trust the settlement layer to enforce both legs, and you depend on solver inventory. A classic canonical bridge is slower and can pause, but its worst case is usually a delay, not a loss. The right choice depends on size and urgency.
What happens if a solver can't fill my intent?
It depends on the design. In most systems the intent has a deadline; if no solver fills it, it expires and your committed funds are returned on the source chain. Some routes allow partial fills, so you might receive part of the order at the quoted price while the remainder expires unfilled. Refund speed varies and usually rides the same settlement layer that just failed, so check the refund terms — an automatic, same-day refund is worth more than a promised-fast route with a 24-hour fallback window.
Do intent routes remove slippage and MEV exposure?
For you, largely yes on the surface: the solver routes the swap privately and quotes you a fixed output floor, so a bot cannot sandwich an order it cannot see. The slippage and MEV do not disappear — the solver prices them into the fee it charges, and it keeps any positive slippage it does not rebate. So you trade an unbounded, visible MEV surface for a small, predictable fee. That is usually a good trade, but it is a transfer of risk, not an elimination of it.
Price your cross-chain move before you sign
DifiCalc's DeFi yield calculator models the fee, the slippage rebate and the time cost of an intent route versus a classic bridge — so you can see the real net output, not the sticker fee.
Open the DeFi Yield CalculatorRelated: run the cross-chain bridge safety checklist, read up on MEV sandwich protection for the swaps on either end, understand withdrawal queues and redemption delays, and model the numbers in the DeFi yield calculator.