Flash loan arbitrage across Uniswap and Sushiswap

Flash loan arbitrage lets a trader borrow crypto without posting collateral, route it through several decentralised exchange pools, repay the loan in the same transaction, and retain any surplus. The transaction either completes in full or reverts, so the borrower does not carry conventional lending exposure when the trade fails.

A triangular route uses three assets and three swaps. For example, a bot might borrow USDC, exchange it for WETH on Uniswap, trade WETH for DAI on Sushiswap, and convert DAI back to USDC. The route is profitable only when the final USDC balance exceeds the original loan, the flash-loan fee, swap fees, gas, and any execution costs.

For Australian traders, this activity sits within a market shaped by ATO tax rules, AUSTRAC obligations for registered digital currency exchanges, and ASIC scrutiny of financial products. A profitable transaction can still create reporting and compliance work, so technical execution should be treated separately from legal and tax advice.

How the triangular route works

The first step is selecting a starting asset with deep liquidity, such as USDC, USDT, WETH, or DAI. The bot then compares prices across Uniswap and Sushiswap, considering pool reserves, fee tiers, token decimals, and the exact trade size. A quoted price that looks attractive on a dashboard may disappear after the transaction changes the pool balance.

Suppose the route is USDC to WETH on Uniswap, WETH to DAI on Sushiswap, and DAI back to USDC on Uniswap. The program calculates the output after each constant-product or concentrated-liquidity swap. It must also account for the flash-loan provider’s fee and leave enough USDC in the contract to repay the principal.

This is different from simply buying an asset cheaply in one market and selling it elsewhere. A triangular cycle can finish in the original token within one atomic transaction, avoiding the need to maintain inventory across venues. The trade remains vulnerable to slippage, sandwich attacks, failed calls, and a competing bot submitting a higher-priority transaction.

Finding a viable price discrepancy

A reliable scanner reads on-chain pool states directly or through a carefully maintained indexer. For Uniswap V3, this means assessing ticks, liquidity distribution, fee tiers, and the route’s impact across concentrated liquidity ranges. For Sushiswap pools, reserve-based pricing and the relevant pool fee must be included in the calculation.

The scanner should reject opportunities that depend on stale quotes. It should simulate the complete transaction with the intended amount, current block state, and realistic gas settings. Profit must be expressed in the borrowed asset, because a route showing a gain in WETH may still lose value when converted back to USDC.

Market conditions can shift quickly during Sydney and Melbourne trading hours when global desks are active at the same time as Australian users. The size of the opportunity is often smaller than the apparent spread once fees and price impact are included. Broader speculative cycles have produced similar illusions of easy profit, as seen in analyses of virtual land declines.

Building the smart contract

The execution contract typically receives the flash loan, approves the required routers, performs the three swaps, checks the final balance, repays the loan, and sends any remainder to the designated recipient. It should use strict token allowlists, fixed router addresses, reentrancy protections where appropriate, and a minimum-profit condition that causes the transaction to revert if the result falls below expectations.

A common implementation mistake is approving unlimited spending to an incorrectly configured router or accepting arbitrary token addresses from an external caller. The contract should validate every route parameter and avoid leaving borrowed funds or unexpected tokens trapped after execution. Safe transfer libraries are useful because ERC-20 tokens do not all behave identically.

Testing should begin on a fork of Ethereum mainnet rather than a public testnet, since testnet liquidity and contract behaviour rarely reflect production conditions. Foundry or Hardhat can replay recent blocks, estimate gas, and test failed swaps. An independent security review is sensible before real funds are exposed, especially when the contract combines flash loans, multiple routers, and callback functions.

Executing the flash loan atomically

A provider such as Aave or Balancer supplies the borrowed asset and invokes the contract’s callback. During that callback, the contract executes the planned swaps. Once the route ends, the provider pulls the principal plus its fee. If repayment is impossible, the entire transaction reverts and the state changes from the swaps are discarded.

The transaction should include a deadline, minimum output for every leg, and a final profitability check. These safeguards do not guarantee success, because another transaction can alter pool prices before inclusion. They do prevent the contract from knowingly accepting an arbitrary amount of slippage.

Public mempools expose profitable trades to searchers and validators. A competitor can copy the route, insert transactions around it, or cause the expected output to vanish. Private transaction submission and appropriate gas-price controls can reduce exposure, although they add infrastructure and trust considerations. Ethereum execution is global, so being based in Brisbane, Perth, or Adelaide does not automatically provide a local latency advantage.

Measuring fees, gas, and tax exposure

The break-even calculation should include the flash-loan fee, each DEX fee, Ethereum gas, approval transactions, failed-attempt costs, and any private relay charge. A route with a gross gain of 30 USDC may be unprofitable after a congested mainnet transaction costs more than expected. Gas should be estimated against the actual calldata and contract path rather than a generic swap.

Australian tax treatment requires careful records of wallet addresses, transaction hashes, asset quantities, Australian-dollar values, fees, and the purpose of each transfer. The ATO may treat disposals and exchanges as taxable events depending on the facts, and frequent trading can raise questions about whether activity is on revenue account rather than treated solely as capital gains.

Using an Australian exchange to move AUD into crypto also creates a practical recordkeeping trail. Bank transfers, exchange statements, and DeFi wallet activity should reconcile. The existence of a decentralised protocol does not remove obligations that may apply to a person’s broader financial activity, business structure, or service operation.

Managing operational and market risk

A production bot needs monitoring for changed pool liquidity, paused protocols, token-list updates, oracle failures, RPC outages, nonce conflicts, and abnormal gas markets. Alerts should report reverted transactions, balances, failed approvals, and contract events. A separate emergency process should be able to stop new trades without creating a rushed withdrawal path.

Stablecoins can depeg, wrapped assets can carry bridge or custodian risk, and governance decisions can alter router behaviour. Uniswap and Sushiswap also have different pool designs and fee structures, so a route that worked yesterday may become unprofitable after liquidity migrates. Small test amounts and strict exposure limits are more useful than assuming every apparent spread is exploitable.

Flash loan arbitrage is therefore an engineering system rather than a guaranteed yield strategy. The essential sequence is to model the route from live pool data, simulate the exact transaction, secure the contract, submit it with execution safeguards, and reconcile the result for Australian tax and compliance records. The reader should remember that a spread is only real after every swap, fee, gas charge, repayment obligation, and execution risk has been counted.