The rollup wars and what they mean for payment infrastructure builders.

Everyone keeps asking who wins the rollup wars. That's not actually the question. The question is which one wins for your specific problem. And the answer is different for every payment infrastructure company.

Let me be direct about the tradeoffs instead of pretending one rollup is objectively better.

The cost-versus-finality thing

Arbitrum is an optimistic rollup. Assumes transactions are valid, batches 10,000 of them, posts to Ethereum. Costs about 0.01 to 0.05 dollars per transaction. Your money settles back to Ethereum in seven days.

Base is the same architecture but with better data compression. 0.008 to 0.04 dollars per transaction. Still seven-day finality. Slightly cheaper overall.

Optimism does the optimistic approach too. Costs 0.02 to 0.06 per transaction right now. Seven days. They're building towards a decentralized sequencer and costs should drop when that ships.

StarkNet is a ZK rollup. That means cryptographic proof that transactions are valid. Finality is immediate on Ethereum. No seven-day wait. But verification is expensive. Costs per transaction are higher. Bridge infrastructure isn't mature yet.

Polygon zkEVM is another ZK rollup. Newer. Similar cost as StarkNet. Immediate finality. Still proving itself.

Matching rollups to actual problems

You're running a coffee shop accepting stablecoins? Base or Arbitrum. Seven-day finality doesn't matter because you're not moving money back to Ethereum daily anyway. Cost per transaction is what matters. Base wins by a hair.

You're processing 100 transactions per day averaging 50,000 dollars each? Arbitrum. Bridge liquidity is deep. Seven-day finality is acceptable because that's just 1 million dollars of settlement risk on a company worth more than that. Base also works but Arbitrum has more mature bridge infrastructure.

Exchange or settlement network? You need a ZK rollup. Immediate finality, no counterparty risk. Costs are higher, infrastructure is less baked, but finality guarantees are real. Unless you're okay accepting seven days of risk. Then Arbitrum works and costs way less.

Emerging markets? You probably don't want a rollup at all. Polygon or Solana. Sub-second finality. Costs in cents. Different tradeoff profile.

Gas fees hide the real cost

Here's where people get stupid about choosing a rollup. You see two numbers. L2 gas cost and base layer submission cost. Most teams only optimize for L2 gas because it looks bigger.

On Arbitrum, a basic transfer is 5,000 gas. Gas price is usually 0.1 to 0.2 GWEI. That's 0.0005 to 0.001 dollars per transaction. Sounds great. Except Arbitrum batches 10,000 transactions and posts them to Ethereum. That batch costs 500,000 gas on Ethereum. About 10 dollars at current prices. Divide by 10,000 and you get 0.001 dollar amortized cost just for posting.

Real cost is 0.002 per transaction, not 0.0005. The L2 gas is half the story.

Base is cheaper on the base layer because they compress data better. Arbitrum submits full transaction calldata. Base uses efficient encoding. That's why Base transactions end up 20 percent cheaper overall.

MEV is where this gets painful. MEV is maximal extractable value which is a fancy word for arbitrage. You submit a transaction. Gap exists between submission and inclusion. Someone sees your transaction, understands what you're doing, executes a similar trade ahead of you at a better price. That's MEV extraction. On Ethereum mainnet it's brutal. A 100,000 dollar trade loses maybe 1,000 dollars to MEV.

On rollups the sequencer controls transaction ordering so MEV should be eliminated. Except different rollups handle this differently. Arbitrum is building decentralized sequencing but right now Offchain Labs runs it. They could extract MEV. Optimism is also building decentralized sequencing but not there yet. Base uses Coinbase's sequencer. All three are opaque on MEV.

ZK rollups like StarkNet handle MEV differently. Transaction ordering comes from the proof system, not a sequencer, so MEV is limited. Tradeoff is longer latency and expensive proofs.

For payment companies this matters because MEV directly impacts settlement costs. Moving stablecoins or L2 tokens? MEV is probably trivial. Moving ETH or complex swaps? MEV could be real. Some infrastructure companies build MEV protection into their payment flows.

Bridge withdrawals are the actual constraint

You want to move money from a rollup back to Ethereum. Two paths. Standard withdrawal uses the fraud proof period (seven days on Arbitrum, Base, Optimism). Fast withdrawal goes through a liquidity bridge.

Fast withdrawal works like this. You want to move 10 ETH from Arbitrum to Ethereum. You submit a withdrawal. A bridge provider on Ethereum gives you 10 ETH minus 0.1 percent fee. They take the counterparty risk. After seven days your standard withdrawal confirms and pays them back.

So if you're a payment processor, actual withdrawal time isn't seven days. It's minutes with a fast bridge. But you pay fees and take counterparty risk. For most payment companies it's worth it. You need your stablecoins on mainnet. Can't wait seven days.

Fast bridges have limits though. They only work for standard assets with liquidity on both sides. ETH, stablecoins, major tokens. They charge fees. If the bridge provider runs out of liquidity on the destination, they slow down or stop.

Arbitrum has the best bridge liquidity because they're oldest and established. Base has been racing to improve since launch. Multiple bridge provider partnerships for liquidity diversity. Optimism has similar setup.

ZK rollups don't have this problem because they're pursuing instant finality differently. But they're new, bridge liquidity is thin. Try to move 10 million through a ZK rollup today and you hit limits.

Choosing a rollup for your use case

High-volume, low-value transactions (thousands of ten-dollar payments per minute)? Base is right. Per-transaction costs matter most. Fast bridges aren't essential because you're not moving millions. Seven-day finality is fine for small amounts.

Lower-volume, higher-value transactions (100 transactions per day, 50,000 dollars each)? Arbitrum. Your budget absorbs slightly higher costs. Bridge liquidity matters because you're moving large amounts. Arbitrum has the deepest liquidity. Seven-day finality means maybe 1 million dollars of settlement risk but that's manageable.

Exchange or settlement network needing instant finality? ZK rollup. Costs higher. Infrastructure newer. But finality guarantees are real. Some finance companies accept no other model. Others take Arbitrum or Base and accept seven days because cost savings versus Ethereum mainnet are huge.

Payments in emerging markets where speed beats cost? Probably not a rollup. Polygon or Solana. Sub-second finality. Costs in cents. Different tradeoff entirely.

Nobody wins, they just specialize

The "one rollup wins" story is wrong. What's happening is specialization by use case. Arbitrum owns developer mindshare and bridge liquidity. Base owns cost and Ethereum alignment. Optimism is building the most sophisticated tech. StarkNet is the only one with actual finality.

Real payment companies operate on multiple rollups. Route transactions based on real-time cost, finality requirements, bridge liquidity. Abstract the rollup layer. Single API to customers.

Arbitrum probably gets 40 to 50 percent of L2 payment volume. Oldest, best liquidity. Base gets 20 to 30 percent for cost. Optimism holds 10 to 20 percent while they build decentralized sequencing. ZK rollups get 5 to 10 percent for use cases demanding finality.

This is a stratified market, not a winner-take-all one.