​TL;DR. Crypto bridging fees are the total cost to move a token from one chain to another, and on the exact same route they can vary 10x or more depending on which protocol you pick, how deep liquidity is, and what the gas market looks like that minute. Only one of those numbers is ever published. Circle charges nothing on CCTP Standard Transfers and 0 to 13 basis points on Fast Transfers depending on the source chain (CCTP fees), and Stargate charges 6 bps on every non-STG transfer plus a variable rebalancing fee (Stargate protocol fees). The parts no operator publishes, source-chain gas and the solver or LP spread, are what actually move a quote. At institutional size the cheapest route often stops being a bridge at all. It is an OTC desk RFQ or a CEX withdrawal. This guide breaks the quote into its components and shows when each route wins.
​

If you have ever stared at a bridge quote and wondered why a small transfer carries such a large fee, you are not alone. The fee is never one number. It is at least four numbers stacked together, and each one moves independently. Understanding what sits underneath that quote is the difference between overpaying every time and routing intelligently across the full surface of available cost paths.
​

A bridging fee is the all-in cost to move value from Chain A to Chain B. It decomposes into four parts: source-chain gas to initiate, destination-chain gas to release, a protocol or service fee, and a spread paid to whoever fronts liquidity. Some rails collapse these into one quote. Others show them itemized. The total is what matters.
​

Here is the simplest way to think about it. A bridge is a coordinator. Your tokens either get locked and a representation is minted on the destination, they get burned and re-minted by an official issuer, or a third party fronts the asset on the destination and gets reimbursed from your deposit. Each mechanic carries a different cost structure, which is why two bridges quoting the same route can be 10x apart. The DefiLlama bridge dashboard tracks daily volumes across more than 40 protocols and is the cleanest place to see which rails are actually moving real money.
​

Every bridging fee decomposes into source-chain gas, destination-chain gas, a protocol fee, and a solver or LP spread. Source gas is usually the largest slice on mainnet origins. Protocol fees are almost always the smallest slice. The solver spread is the one users most often overlook, and on large transfers it dominates the bill.
​

The transaction that kicks off the bridge. The cost is gas consumed times base fee times the ETH price, and all three move independently. ETH traded near $2,402 on September 16, 2026 (DeFiLlama), so the identical deposit call costs materially more or less from one week to the next without any bridge changing a thing. Read a live gas tracker rather than any published estimate (Etherscan gas tracker). This is usually the single biggest variable in a mainnet-origin quote, which is why bridging from an L2 is almost always cheaper.
​

The transaction that releases or mints your tokens on the destination. On an intent-based rail, a solver pays this and rolls the cost into the spread. On a lock-and-mint bridge with a relayer network, the relayer pays it and the protocol charges you back. Either way you are funding it, and on L2 destinations it is small next to a mainnet-origin deposit.
​

The take rate charged by the bridge itself, and the only component operators publish. Across's fee page describes the quote as an LP fee plus a relayer fee, with the LP fee falling to zero when the relayer takes repayment on the origin chain (Across fees). Circle charges nothing on CCTP Standard Transfers and 0 to 13 bps on Fast Transfers by source chain: 1 bp from Ethereum and Solana, 1.3 bps from Base and OP Mainnet, 13 bps from Linea (CCTP fees). Stargate publishes 6 bps on every non-STG transfer, split 4 bps treasury, 1 bp veSTG holders, 1 bp liquidity providers, plus a separate rebalancing fee (Stargate protocol fees).
​

This is the fee hiding in plain sight. On intent-based bridges, a solver quotes you a price, fronts the asset on the destination, and collects your deposit on the source plus a spread. On pool-based bridges, you pay a swap fee and slippage into the liquidity pool. This spread absorbs the solver's capital cost, rebalancing cost, and risk premium. It varies most by route and size, and nobody publishes it, because it is set per quote by whoever fronts the capital. Across describes its relayer fee as covering gas, capital opportunity cost and risk, and exposes the components through a quote API rather than a rate card (Across fees). The only reliable way to know the spread on your route is to pull a live quote before you send. Solvers hedge inventory through OTC desks and CEX withdrawals themselves, so at size you can disintermediate by going to those sources directly, which is the crossover the next two sections cover.
​

Add those four numbers together and you have the quote. For deeper mechanics, our walkthrough on the difference between crypto bridging and swapping explains how these same components show up in swap routing too.
​

Fee tables that quote an exact dollar cost per rail are not verifiable after the fact, because source gas and solver spread are both live variables. What can be checked is the published protocol fee. Below is what each operator states on its own fee documentation. Everything else in a quote has to be read off a live API call.

A few patterns jump out of those published numbers. Intent rails put no protocol take in the quote at all; the cost is capital, gas and risk priced per fill. Pool-based rails carry an explicit take rate plus a balance-dependent rebalancing fee. Burn-and-mint sits between the two: no LP risk to price, but Circle charges for speed and the charge varies by source chain. The CCTP fee page and the Across fee page are where the actual formulas live. OTC desks and CEX routes publish no rate card at all, so comparing them means pulling live quotes. For a full ranked comparison of on-chain rails on speed, liquidity, and chain coverage, see our best crypto bridges guide.
​

The cheapest bridging path is a function of notional size, not just route popularity. At retail size the fixed component dominates, so rails that absorb destination gas into a single quote win. As notional rises the percentage component dominates instead, and OTC desk RFQs from Cumberland, FalconX or Wintermute settle bilaterally without ever touching a public bridge. Desks do not publish their spreads, so the only way to know whether a desk beats a rail on your size is to request a quote from both. For niche destinations where solver coverage is thin, a CEX withdrawal route is often the cleanest path.
​

OTC desks compete with on-chain bridges at institutional size. A desk like Cumberland or FalconX quotes a principal price, takes the asset on one chain, and delivers on another from inventory. The user pays a spread rather than a protocol fee and settles bilaterally without exposing size to a public mempool. Neither Cumberland nor FalconX publishes a spread schedule, so the numbers only exist inside a live RFQ.
​

The mechanism is simple. Desks run large standing inventory across major chains and CEX venues. When a client requests a USDC delivery on Base against Ethereum, the desk debits Ethereum inventory, credits Base inventory, and rebalances internally on its own schedule. Because they net flow across many clients, the per-trade cost drops below what a solver fronting a single fill can offer. The Cumberland desk overview and the FalconX product documentation describe the RFQ workflow.
​

The tradeoff is settlement risk and onboarding friction. A bridge is atomic in seconds. An OTC trade is bilateral. The desk needs your KYC, a master agreement, and an established credit relationship. Settlement typically lands within minutes to a few hours rather than seconds, and you carry counterparty exposure until both legs settle. For treasury operations moving stablecoins at scale on a recurring basis, any spread saving compounds across every transfer and can justify the operational setup. For one-off transfers, the friction usually outweighs it.
​

A centralized exchange can function as a free or near-free bridge for any asset it lists on multiple chains. The user deposits on one chain, the exchange credits the account internally with zero on-chain transfer, and the user withdraws on a different chain. Internal book transfers cost nothing. Only the withdrawal fee on the destination side applies. For USDC, many large venues currently charge $0 for native withdrawals to Base, Arbitrum, and Optimism.
​

Coinbase is the cleanest example. A deposit of USDC on Ethereum is credited instantly. A withdrawal of USDC on Base is free for native USDC at current schedules. The entire move costs Ethereum deposit gas plus zero withdrawal fee. Binance, OKX, and Bybit run similar internal-transfer mechanics with varying withdrawal-fee schedules per chain and per token. DeFiLlama tracked Coinbase Bridge at about $7.7 billion and the major CEX venues at about $161 billion (Binance), $30 billion (OKX) and $15 billion (Bybit) when pulled on September 15, 2026 (DeFiLlama protocols), which gives a sense of the inventory standing behind these routes.
​

The CEX route wins in two specific scenarios. First, when the destination is a chain with thin on-chain solver coverage but solid CEX support. Second, when the asset is a stablecoin and the destination CEX withdrawal fee is at or near zero. The tradeoff is custody and KYC. Your funds touch the exchange's books, you wear platform risk during the deposit-to-withdrawal window, and not every user is willing or able to use a regulated CEX. For users already holding funds on an exchange, the route is often unbeatable on fee for stablecoin transfers to supported L2s.
​

The same transfer can cost an order of magnitude more or less depending on five factors: source-chain gas, destination liquidity, transfer size, route popularity, and protocol mechanism. Only the last of those is published anywhere. The other four are read off live conditions at the moment you quote.
​

When Ethereum base fees spike from 10 gwei to 80 gwei (common during NFT mints or token launches), every mainnet-origin bridge quote rises together. L2-origin transfers barely budge. If you are bridging from Arbitrum, Base, or Optimism to another chain, your fee is almost entirely decoupled from Ethereum congestion.
​

Solvers and LPs charge more when their capital has to travel further to rebalance. Ethereum to Base has deep liquidity on both sides. Ethereum to a newly launched L2 has fewer solvers keeping float there, and the spread reflects it. Stargate makes the same effect explicit on the pool side: its rebalancing fee is a function of how far a transfer pushes a destination pool below its target balance (Stargate protocol fees). If you are bridging to a young L2, expect to pay for the thin book or consider the CEX route described above.
​

Fees do not scale linearly with size. On intent rails there is a fixed component, the destination gas the solver fronts, plus a percentage component, the spread. A small transfer pays the same destination gas as a large one, so the fixed part dominates at the bottom of the range and the percentage part dominates at the top. Across exposes both separately in its quote response, as relayerGasFee and relayerCapitalFee (Across fees). For small transfers, look for rails with no fixed fees. For large transfers, look for the tightest percentage spread and compare against an OTC RFQ.
​

Volume concentrates on a handful of routes, and solver competition follows it. The DeFiLlama bridge dashboard is the cleanest public view of which corridors actually move money. Off-the-beaten-path routes attract fewer bids, and the spread reflects that.
​

A solver fronting the asset on the destination only needs destination gas plus a spread. A lock-and-mint bridge running its own messaging layer pays relayers to observe the source event, validate it, and execute the mint. More on-chain steps means more gas means higher fees. This is structural, not tunable.
​

The fee profile of any bridge flows directly from how it is built. Five architectures dominate production in 2026, and each has a distinct cost shape. Intent rails are cheapest on small and mid-size transfers, burn-and-mint rails sit in the middle, pool and lock-and-mint designs run higher.
​

User signs a desired outcome ("pay X on chain A, receive Y on chain B"). A solver reads the intent, fronts the asset on the destination, and collects the deposit on the source. The user sees a single quote that bundles destination gas plus a small spread. These rails are the cheapest on most retail-size routes because solvers absorb destination gas and compete on spread. The ERC-7683 cross-chain intents standard formalized this design in 2025, and most intent bridges have adopted it.
​

Circle burns USDC on the source and mints native USDC on the destination via an attestation. No LP. No wrapped token. No slippage. You pay source gas, destination gas, and Circle's Fast Transfer fee, which is 0 to 13 bps depending on the source chain and is deducted from the transferred amount at mint unless you pay it upfront. Standard Transfers carry no Circle fee at all (CCTP fees). Circle tells integrators not to hardcode the rate and to re-read the fee API at least weekly, which is a good reason to distrust any fee table that quotes a fixed number.
​

LPs supply native assets into shared pools across chains. A user swaps into the pool on the source, receives native on the destination from the same pool. Fees cover the 6 bps protocol charge, the balance-dependent rebalancing fee, and the messaging layer (Stargate protocol fees). These rails handle non-stablecoin assets well but carry more cost layers than intent rails.
​

Tokens get locked on the source and a wrapped version is minted on the destination. Unwinding requires a reverse burn. Fees cover source gas, destination gas, relayer compensation, and a protocol fee. Historically most expensive and slowest, but sometimes the only option for obscure long-tail assets.
​

Each chain has its own AMM pool for a token and bridgers swap canonical USDC into the hToken, send messages across, then swap out of the hToken on the destination. Two swap fees plus slippage plus messaging. Fees are higher but the design handles EVM-to-EVM transfers with fast finality.
​

Bridging fees feel high mostly because of Ethereum gas, not bridge greed. Published protocol fees are the smallest line in the stack: zero on a CCTP Standard Transfer, single-digit basis points on a CCTP Fast Transfer from most chains, 6 bps on Stargate (CCTP fees, Stargate protocol fees). Source-chain gas is the line that swings, and it swings with the base fee and the ETH price rather than with anything the bridge controls (Etherscan gas tracker).
​

For very small transfers the math is uglier, because the fixed gas component is amortized over almost nothing. That is arithmetic, not a percentage take by the bridge. The fix is bridging larger amounts less often, originating from an L2 instead of mainnet, or using a rail whose quote carries no fixed component, which Across exposes separately as relayerGasFee (Across fees).
​

Seven levers move the bill: originate from an L2, use an intent rail on popular routes, run an aggregator quote, time around gas spikes, batch small transfers, use CCTP for canonical USDC delivery, and pick deeper-liquidity destinations. They stack, and the ones that touch source-chain gas matter most because that is the largest and most volatile line.

For teams building apps on top of bridging infrastructure, fee abstraction is increasingly table stakes. Eco's rail and app model bundles solver selection, liquidity routing, and settlement finality into a single integration. If you are comparing integration paths, the best ETH-to-stablecoin aggregator writeup covers where this fits in a typical stack.
​

Bridging native USDC is not the same as bridging a wrapped variant, and the fee profiles differ. Native USDC is the canonical Circle-issued token. Wrapped variants like USDC.e on Arbitrum, or bridge-specific wrappers like nUSD (Synapse) or hUSDC (Hop), are representations only redeemable through that bridge. The exit swap from wrapped to native often adds fees the original quote did not show.
​

Burn-and-mint rails (CCTP) give you native USDC on the destination. Pool-based rails often give you native too, after an internal swap. Lock-and-mint rails give you a wrapped version that you then have to swap to native if you want the canonical asset. The "cheap" bridge quote can turn expensive once you add the exit swap. Our deeper piece on bridging wrapped tokens walks through the redemption mechanics in detail.
​

Stablecoin routes carry the lowest fees because USDC and USDT have the deepest cross-chain liquidity. ETH bridges at similar cost on intent rails because solvers hold ETH inventory on most chains. Long-tail ERC-20s, LSTs, and governance tokens pay more for the same notional, because fewer solvers hold inventory and pool depth is thinner. NFTs use dedicated bridges with flat per-piece fee models.
​

If your transfer is a stablecoin, you have the best fee environment the market offers. If it is a long-tail asset, check an aggregator first. The route might not exist directly, and you might be better off swapping to USDC on the source, bridging the USDC, and swapping back on the destination. Two swaps plus a cheap bridge often beats one expensive bridge of the long-tail asset.
​

Cheaper is not always better. Bridges with minimal fees sometimes achieve that by running leaner security: fewer validators, fewer confirmation blocks, more trust in solver honesty. The Rekt leaderboard of exploits is a cautionary read, and bridges are heavily represented on it.
​

The working rule: for transfers you care about, pay a few bps more to use a rail with a strong security model (CCTP's Circle attestation, Across's optimistic oracle, Eco's solver slashing, LayerZero's DVN setup). For pocket-change transfers, optimize purely on fee. For institutional-size transfers, the OTC and CEX routes carry their own counterparty considerations that need their own underwriting. More on the on-chain side in our guide on crypto bridging safety.
​

In almost every case the answer is Ethereum gas, not the bridge itself. The published protocol fee is the smallest line: zero on a CCTP Standard Transfer and 0 to 13 bps on a Fast Transfer (CCTP fees), 6 bps on Stargate (Stargate protocol fees). Source-chain gas is the line that dominates a mainnet-origin quote, which is why originating from an L2 is the single biggest lever.
​

There is no stable answer, which is why this article does not print one. The only fixed part is the published protocol fee: nothing on a CCTP Standard Transfer, 0 to 13 bps on a Fast Transfer depending on source chain (CCTP fees), 6 bps plus rebalancing on Stargate (Stargate protocol fees), and no separate protocol take on Across (Across fees). Everything else depends on the gas market and the spread in your live quote.
​

At institutional size it is worth asking. Desks like Cumberland, FalconX and Wintermute quote a principal price from inventory rather than charging a protocol fee, but none of them publishes a spread schedule, so the comparison only exists once you have both quotes in hand. The tradeoffs are KYC onboarding, a master agreement, and settlement risk over minutes to hours instead of atomic seconds. For one-off retail transfers, stick with intent rails.
​

Originate from an L2 instead of mainnet, use an intent rail on popular routes, check an aggregator before committing, time transfers during low gas, batch small transfers, and at institutional notional compare on-chain quotes against an OTC RFQ or a CEX withdrawal route. The first lever matters most, because source-chain gas is the largest and most volatile line in the quote. Our best crypto bridges ranking highlights which rails win on fee for each common route.
​

Four components: source-chain gas to initiate, destination-chain gas to release, a protocol fee charged by the bridge, and a spread paid to solvers or liquidity providers. Source gas is usually the largest slice on mainnet-origin transfers. Protocol fees are almost always the smallest.
​

It depends on the route and the moment, and neither side publishes a number you can compare in advance. Circle's Fast Transfer fee runs 0 to 13 bps by source chain, and Standard Transfers are free (CCTP fees); Across charges no separate protocol take and prices the fill through its quote API (Across fees). CCTP's durable advantage is that you receive canonical Circle-issued USDC with no wrapped-token intermediary, which matters for treasury flows and regulated contexts.
​

No truly zero-fee bridge exists in production, because source-chain gas is always paid by someone. Across takes no separate protocol fee, but the quote still carries a relayer fee and source gas (Across fees), and Circle's free tier is the slower Standard Transfer rather than the Fast one (CCTP fees). The nearest-to-free experience is an L2-origin intent transfer on a deep route, or a CEX withdrawal route where the destination withdrawal fee is zero for native USDC.
​

Usually not meaningfully on the busiest routes, because both have deep liquidity and every major rail supports them. DeFiLlama put USDT supply near $183.3 billion and USDC near $73.4 billion when pulled on September 15, 2026 (DeFiLlama stablecoins). On newer L2s, USDC often has deeper solver coverage than USDT and quotes come in tighter. CCTP only supports USDC, so if you want the burn-and-mint path specifically you need USDC.
​