Picking an L2 is not picking a winner. It's picking the chain your money is already sitting on at the moment you ask for it.
Base and Arbitrum get filed under the same mental bucket — "the cheap Ethereum L2s" — and then merchants spend a week reading throughput charts that have nothing to do with their problem. For settlement, the two chains are far more alike than the benchmarks suggest, and the one variable that actually decides the answer rarely shows up in a comparison table.
Both are EVM. Both clear a transfer for far less than mainnet gas. Both inherit Ethereum's security as their settlement layer, and both are reorg-aware the way every chain we settle on has to be. From the contract's point of view a transfer is a transfer either way. So the comparison that matters isn't "which L2 is faster." It's "which assets actually settle on each rail, and which one is where your customer's funds already live."
The thing that's the same
Start here, because it collapses half the decision.
Base and Arbitrum are both EVM-equivalent. The address format is identical, the wallet flow is identical, the confirmation model is the same shape, and on both the customer pays gas in ETH. From our side, the gate logic, the per-chain confirmation thresholds, and the reorg handling are cut from the same cloth. From the payer's side, MetaMask, Coinbase Wallet, Rabby, and the rest treat both as a network you switch to and send from.
So if a vendor tells you Arbitrum is "faster" or Base is "cheaper" as a tiebreaker, push back. At settlement amounts, the gas difference between the two L2s is rounding noise next to mainnet. Neither one is going to compress your margin. That's not where the decision is.
The thing that's different: which assets settle on each
Here's the variable that actually moves.
USDC settles on Base. On halfin, Base carries two settlement assets — USDC and native ETH — and the USDC contract on Base is its own canonical deployment, distinct from USDC on Ethereum and USDC on Solana. Base was built as Coinbase's L2, and a large share of retail and SMB dollars that enter crypto through Coinbase land on Base as USDC and stay there. When a US customer holds "some USDC," there is a good chance it is already on Base, in a wallet, a tap away from paying you.
Arbitrum is the other low-cost EVM rail we settle on, but the asset picture is different: on halfin, Arbitrum settles native ETH (and ETH is also the gas asset). We do not list USDC as a settlement asset on Arbitrum — if you want to take dollars as USDC on an L2, Base is the rail that carries it. So the Base-vs-Arbitrum choice is not "USDC here or USDC there." It's "do your customers want to pay in USDC (Base) or in ETH on a cheap L2 (Arbitrum)?"
That distinction is the whole post:
So which one
The honest answer is: expose the assets your customers actually hold and stop agonizing about chains. But if you're deciding which L2 rails to turn on, decide by the asset, not by a benchmark.
Turn on Base if you want to take dollars as USDC. It's the L2 where halfin settles USDC, and a large slice of retail and SMB dollars already sit there as USDC after coming in through Coinbase. For a customer paying in stablecoins, Base is the low-friction door — they pay from the funds exactly where they already are.
Turn on Arbitrum for customers paying in native ETH on a cheap L2. ETH-holding, crypto-native payers who want mainnet-shaped value without mainnet gas will settle on Arbitrum without a second thought. It's an ETH rail on halfin, not a second USDC rail — so reach for it when the asset your audience holds is ETH, not dollars.
Your treasury is a real input too. What lands on each rail — USDC on Base, ETH on Arbitrum — is what you reconcile and redeploy. We do balance conversion asset-to-asset, so you're not locked to the asset a payment arrived in — but a balance you don't have to convert is one less moving part.
The trap
The mistake we watch merchants make is picking one rail because a benchmark said it was "faster," advertising only that one on checkout, and then discovering their customers' money is somewhere else. A merchant whose audience holds USDC turns on Arbitrum-only and watches dollar-holders bounce, because USDC doesn't settle there. A merchant whose audience holds ETH leans on a single stablecoin rail and loses the ETH payers. Those customers now face a choice: move funds into the asset-and-chain you accept, or give up. A meaningful fraction give up. The ones who don't burn a bridge or swap fee to send you money they'd have sent cleanly if you'd shown the rail their balance was already on.
You optimized for a number on a chart and paid for it in abandoned checkouts. The fee delta you "won" between two L2s is rounding noise next to a single lost sale.
What we recommend
Bill in fiat, let the customer choose an asset, and accept the rails their wallets are already pointed at. The networks page shows the full set we settle on — Base and Arbitrum are two of seven — and the same "meet the money where it is" logic applies across all of them. For a checkout, the right set of supported rails is "the ones your customers actually hold balances on," not "one, picked by benchmark."
If you want the longer version of the closely related question — Base versus Ethereum mainnet for the same USDC — we wrote that up separately in USDC on Base vs Ethereum. The short version: mainnet is for institutional size where gas is rounding noise; the L2s are for everyone else.
Operating rule
Base and Arbitrum are not rivals you have to rank. They're two doors into the same building, and your customer already knows which one they walked in through. Read which asset their money is in — dollars as USDC on Base, or ETH on Arbitrum — accept that rail, and keep the other door open. The chain is plumbing. The asset and where it sits are the decision.