Solver netting is the practice of matching opposing cross-chain intents inside a solver's own balance sheet, so the underlying funds rarely move across a bridge at all. A user who wants USDC on Base sourced from Ethereum and another user moving USDC from Base to Ethereum can be settled against each other locally. The solver pays each one out from inventory it already holds on the destination chain, and the imbalance, if any, gets bridged later. The user experience looks like a single signed intent and a fast fill. The plumbing underneath looks more like a market maker reconciling its book than a bridge moving tokens.
​

That distinction matters because intent-based protocols now route a meaningful share of cross-chain stablecoin volume. Across alone has settled over $19B in cumulative bridge volume since launch, with median fill times under 30 seconds for stablecoin transfers across Ethereum, Base, Arbitrum, Optimism, and Polygon. Most of those fills never trigger a canonical bridge transaction at the moment of settlement. The solver fronts the inventory and rebalances on its own clock. Solver netting is what makes that economically viable.
​

A solver, sometimes called a filler or a market maker, is a permissionless agent that competes to fulfill user intents in exchange for a small spread. When two intents flow in opposite directions across the same chain pair, a solver can settle them by swapping internal balances rather than moving funds onchain. The user on Ethereum hands their USDC to the solver, and the user on Base receives the solver's existing USDC there. The solver's net position is unchanged. No bridge fired. No CCTP attestation requested. No 13-minute Optimism withdrawal window.
​

The term borrows from traditional finance. Continuous Linked Settlement, the FX settlement system run by CLS Bank, nets roughly $6.5T in average daily FX activity down to settlement instructions worth a fraction of that. SWIFT messaging volumes hit 11.7B annual messages in 2024 according to SWIFT's traffic figures, but the actual cash movement is a small fraction of that gross flow because correspondent banks net positions internally. Solver netting brings the same logic onchain: gross intent volume can be many multiples of the actual cross-chain settlement that happens.
​

Eco Routes, Across, CoW Protocol, UniswapX, 1inch Fusion, deBridge DLN, and Hashflow all operate on this principle to varying degrees. They differ in auction format, in how solvers source liquidity, and in what guarantees the user receives, but the underlying inventory math is the same.
​

An intent is a signed message describing the outcome the user wants, not the path. A typical cross-chain stablecoin intent encodes the source chain and asset, the destination chain and asset, the recipient, the maximum acceptable slippage or fee, and an expiry timestamp. The user signs once with a permit-style signature, usually EIP-2612 or EIP-3009 for stablecoins, so no separate approval transaction is required.
​

The intent enters a solver auction. Solvers compete on price, speed, or both, depending on the protocol. The winning solver pays the recipient on the destination chain immediately from its own inventory. It then reclaims the user's source funds, either by pulling them via permit at the moment of fill or by having a settlement contract release them after a verifiable proof. ERC-7683 calls this an Origin Settler and Destination Settler model. The user signs an intent into a standardized order format; an Origin Settler on the source chain locks the user's funds; a solver fills on the destination; the Destination Settler releases the source funds to the solver after settlement is verified.
​

The netting happens at the solver's inventory layer. A well-capitalized solver may hold tens of millions of USDC across all major chains. When user flows are roughly balanced in both directions, the solver's per-chain inventory drifts only slightly with each fill. The solver only triggers an actual cross-chain rebalance, via CCTP, LayerZero OFT, Hyperlane, or a canonical bridge, when its inventory on a given chain falls below a threshold. That rebalance can be batched, scheduled for off-peak gas, and routed via the cheapest rail available at the moment.
​

The result for the user is the dominant economic property of solver netting: the cost and latency of the average intent are decoupled from the cost and latency of the underlying bridge. The solver pays bridge fees and waits out finality on its own balance sheet, amortized across thousands of fills. The user pays a spread that reflects the marginal cost, not the gross cost.
​

A solver's profit-and-loss reduces to four lines: spread captured on each fill, cost of cross-chain rebalancing, cost of capital tied up in inventory, and any MEV or routing edge captured during execution. The first is positive. The next two are negative. The fourth is sometimes positive, sometimes neutral, and occasionally a leak.
​

The spread on a typical USDC-to-USDC cross-chain transfer through an intent protocol sits in the 1 to 8 basis-point range, well below the 10 to 30 bps that canonical bridges plus DEX hops typically extract. Across publishes live referral and fee data showing relayer fees of roughly 4 bps on stablecoin routes between Ethereum, Base, Arbitrum, and Optimism. UniswapX's Dutch auction mechanism lets solvers bid down the spread further on small-to-medium orders, with the auction starting at a worse price for the user and improving over a 30 to 60 second window until a solver fills.
​

Capital cost is the binding constraint on solver scale. A solver running $20M of inventory across 10 chains at a 5% blended cost of capital absorbs roughly $1M of annualized funding cost. To break even on capital alone, that solver must generate $1M of net spread across all fills, which at 4 bps implies clearing $25B of annual gross intent volume. That math is why solver-network operators including Wintermute, Flow Traders, and crypto-native firms operate solvers as their primary order flow surface in DeFi: the volume threshold is non-trivial and rewards specialization.
​

The net inventory model is the operational expression of these constraints. Each solver maintains a target balance per chain per asset, derived from observed flow distributions. When a fill leaves the solver overweight on Base USDC and underweight on Ethereum USDC, the solver records the imbalance but does not act on it immediately. The next opposing intent, paying out Base USDC and receiving Ethereum USDC, restores the position without any onchain rebalance. Only when the cumulative imbalance crosses a configured threshold, often $250K to $1M depending on the route and capital allocation, does the solver fire a rebalance transaction. In well-behaved bidirectional corridors the netting ratio routinely exceeds 8 to 1, meaning gross intent volume settles at one-eighth the actual cross-chain transfer volume.
​

Rebalancing cost is the least visible line and the one solver design most aggressively optimizes. A solver that can defer rebalancing until imbalances are large enough to use the cheapest bridge for that route, and that can route via CCTP on canonical USDC corridors instead of paying liquidity fees, pushes its rebalancing cost from 8 to 15 bps down to 1 to 3 bps. The difference flows directly to user spreads.
​

The fourth P&L line, MEV and routing edge, is the most disputed. Solvers can route a user through a sequence of pools that maximizes user receive amount, splitting the order across Uniswap v3 and v4 pools, Curve stableswap pools, and protocol-specific liquidity, and capture the routing alpha as part of their spread. They can also extract value through worse routing or sandwich-style ordering on adjacent transactions in the same block. Honest solvers compete on price; less honest ones use private order flow access to extract on the margin. Auction structure determines which behavior dominates.
​

Across uses a single relayer fills the intent on the destination chain immediately after the user deposits on the source chain. The relayer is reimbursed from a unified liquidity pool after a 30 to 60 minute optimistic verification window managed by the Across protocol's canonical bridge and UMA's optimistic oracle. Solvers do not net peer-to-peer; they net against the unified pool, which itself rebalances across chains on its own schedule.
​

Eco Routes is the developer-facing intent execution surface for the broader Eco stablecoin orchestration network. Routes accepts permit-signed stablecoin intents and clears them through a permissionless solver set that competes on price and finality. Solvers maintain inventory across the 15 chains Eco supports today, including Ethereum, Base, Arbitrum, Optimism, Polygon, BNB Chain, Avalanche, and Solana, and net opposing flows internally before triggering rebalances via CCTP, Hyperlane, or LayerZero. The orchestration platform selects the rail per route based on cost and finality. Teams integrate the Routes API once and get netted execution across the supported chain set.
​

CoW, short for Coincidence of Wants, runs batch auctions every block. Solvers submit competing solutions to a batch of orders, and the protocol picks the solution that maximizes total surplus. When two orders within a batch can be matched directly without touching external liquidity, that pair settles peer-to-peer and the solver only sources external liquidity for the residual imbalance. CoW's batch auction docs describe the matching algorithm in detail. CoW originally ran on Ethereum mainnet and now extends across Gnosis Chain, Arbitrum, Base, and Polygon.
​

UniswapX runs a Dutch auction per order. The order starts at a price unfavorable to the user and improves linearly over a 30 to 60 second window. Solvers, called fillers in UniswapX terminology, monitor the auction and fill the moment the price crosses their break-even. Cross-chain UniswapX, launched in 2024 and described in the UniswapX V2 announcement, extends this to multi-chain orders. Fillers fill the destination side from inventory and reclaim source funds via the auction settlement contract.
​

1inch Fusion uses a similar Dutch auction structure for single-chain swaps. Fusion+, the cross-chain extension, uses an HTLC-style atomic swap pattern combined with the auction. The 1inch resolver network competes per order. The Fusion+ technical overview describes how resolvers maintain liquidity across chains and net flows internally.
​

deBridge DLN, the DLN standing for deBridge Liquidity Network, runs a 0-TVL model where solvers source destination-chain liquidity on demand rather than pre-funding pools. deBridge stats show DLN cleared roughly $11B in cumulative volume by late 2025 across over 20 chains, with median fill times under 8 seconds for major chain pairs.
​

Hashflow uses a request-for-quote pattern instead of an auction. Market makers stream firm two-sided quotes, the user accepts a quote, and the trade settles against the maker's inventory. Hashflow's RFQ design is closer to traditional FX market making than to the auction-based intent protocols, but the netting math at the maker's book is identical.
​

ERC-7683 is the cross-chain intent settlement standard authored in 2024 by Across and Uniswap Labs. It specifies a standardized order format, a generic Origin Settler interface for locking source funds, and a generic Destination Settler interface for releasing those funds after a fill is verified. The goal is solver liquidity portability: a solver that supports the standard can fill orders from any compliant origin protocol without writing protocol-specific integrations.
​

Adoption is uneven but growing. Across, Eco, Uniswap Labs, Relay, and several other intent protocols have shipped or committed to ERC-7683 support. The reference implementation on GitHub shows the canonical interface and a sample Origin Settler. The standard does not specify how proofs of fill are generated; that is left to each protocol, which means a solver supporting two ERC-7683 origins still needs separate proof verification logic for each. The standardization is at the order format and settlement-contract layer, not at the cross-chain messaging layer.
​

The practical impact for solver netting is consolidation. As more origins adopt the standard, solvers can spread fixed inventory across more order flow, which raises the netting ratio and lowers spreads. The intent settlement layer landscape already shows convergence around a small set of standards.
​

Three cross-chain mechanisms account for nearly all USDC flow today: solver-netted intent fills, CCTP burn-and-mint, and canonical lock-and-mint bridges. Each has a different cost, latency, and risk profile.
​

Canonical bridges lock funds on the source chain and mint a wrapped or pegged representation on the destination. Speed depends on the bridge's finality assumption: optimistic bridges like the original Optimism canonical bridge wait 7 days, while light-client and validator-set bridges complete in 5 to 30 minutes. Fees range from 5 to 50 bps depending on route and bridge model. Risk is dominated by the bridge's validator set or proof system; Rekt's incident archive documents over $2.8B in cumulative bridge exploits across Wormhole, Ronin, Nomad, and others.
​

CCTP, Circle's Cross-Chain Transfer Protocol, burns USDC on the source chain and mints native USDC on the destination after Circle's attestation service signs the burn. CCTP V2 is live across 11 chains as of Q1 2026. Latency is roughly 60 to 90 seconds on V2 fast routes. Fees are zero protocol-side; users pay only gas. The trust assumption is Circle's attestation service signing honestly, which is a centralized trust point but a well-capitalized one.
​

Solver-netted intents settle the user-visible portion in 5 to 30 seconds. The user pays a spread of 1 to 8 bps. The solver absorbs bridge or CCTP fees and waits out finality on its own balance sheet. The trust assumption splits across the solver, the intent protocol's proof system, and any underlying bridge the solver uses to rebalance. For most retail and SMB-sized stablecoin transfers, the latency and cost win is decisive: a $50K USDC intent from Ethereum to Base settles in seconds for under $4 of total cost on most intent protocols, versus 2 to 15 minutes for CCTP V2 and roughly $25 to $150 for canonical-bridge plus DEX-hop combinations.
​

Solver netting introduces failure modes that pure bridges do not. The first is solver insolvency. A solver fills the destination side from inventory and waits for source-side reimbursement. If the solver miscalculates capital allocation or gets caught in a directional move, it may hold an unbalanced book for hours. Most production intent protocols mitigate this with bonding requirements, reputation scoring, and fast slashing for missed fills, but the user is still implicitly relying on solver solvency for the duration between fill and settlement.
​

The second is MEV extraction. Solvers see intents before they settle. A poorly designed auction or a private-mempool arrangement can let a solver extract value through sandwich-style ordering or by routing the user through a longer path than necessary. CoW's batch auction MEV protections and UniswapX's Dutch auction structure are explicit attempts to contain this. The 2023 paper Quantifying MEV on Decentralized Exchanges by Heimbach and Wattenhofer documents the scale of the problem: roughly $675M in MEV extracted on Ethereum DEXs in the 18 months covered.
​

The third is liveness. Solver auctions need solvers. If solver participation drops, fills slow, prices widen, and user experience degrades. The 2024 paper Decentralized Finance: Risks, Regulation, and the Role of Financial Innovation from the Federal Reserve's research staff discusses the dependence of intent-based DEXs on solver liveness as a structural fragility, separate from the credit risk of any individual solver.
​

The fourth is settlement-layer assumption. ERC-7683 standardizes the order format but each protocol still chooses its own proof-of-fill mechanism: optimistic verification with a fraud-proof window, light-client proofs, validator-set attestations, or oracle-based attestations like UMA. Each has a different trust profile and a different failure surface. A user signing an intent should know which proof mechanism the protocol uses, because that mechanism, not the bridge name, is the binding trust assumption.
​

Stablecoin payments are the workload solver netting is best suited for. Stablecoin flows are large, repeated, and bidirectional. A treasury that pays out USDC on Base and receives USDC on Ethereum runs net flows that solvers can absorb cheaply. Payroll, vendor settlement, OTC clearing, and exchange-to-exchange transfers all share this property. The 1:1 nature of stablecoin-to-stablecoin transfers also removes the price-discovery overhead of volatile-asset intents, which simplifies the auction and tightens spreads.
​

Eco is built around this thesis. The orchestration platform routes stablecoin intents across 15 chains and selects between CCTP, Hyperlane, and LayerZero per route based on cost and finality. Eco's solver network nets opposing flows before triggering any underlying rebalance, and the developer surface, Routes, exposes that as a single API call. Teams shipping stablecoin payment products, treasury automation, or programmable-address infrastructure integrate once and inherit the netting math. For deeper context on how this maps to specific use cases, the writeups on 1:1 stablecoin swaps and stablecoin payroll automation show the production patterns.
​

A bridge moves funds across chains every transaction. Solver netting matches opposing intents inside a solver's inventory so the underlying funds rarely cross a bridge. The user sees a fast fill; the solver rebalances on its own clock when imbalances accumulate, often via canonical bridges or CCTP.
​

Solvers capture a spread of roughly 1 to 8 basis points per fill on stablecoin routes. They pay bridge or CCTP fees out of that spread when they rebalance, plus their cost of capital on inventory. Profit comes from running enough volume that the spread covers funding and rebalancing costs, with edge from MEV-aware routing.
​

ERC-7683 is a cross-chain intent settlement standard authored by Across and Uniswap Labs. It defines a generic order format and Origin and Destination Settler interfaces. Adoption lets solvers fill orders from any compliant protocol without protocol-specific integration, which raises the netting ratio across the ecosystem.
​

No. Solvers still rebalance via canonical bridges, CCTP, or messaging protocols when their inventory drifts. Solver netting is a layer above the bridge stack, not a replacement. The economic win is that user-facing transactions are decoupled from bridge cost and latency, not that bridges go away.
​

Across, Eco Routes, CoW Protocol, UniswapX, 1inch Fusion and Fusion+, deBridge DLN, and Hashflow all use solver-based execution with internal netting. They differ in auction format, in whether liquidity is pooled or peer-to-peer, and in proof-of-fill mechanism, but the inventory math is consistent across the set.
​