The second floor

A blockchain that every node must fully verify cannot be fast, because every node must fully verify it. That is not an engineering failure; it is the design. Scaling therefore means moving execution somewhere else while keeping the security guarantee of the base chain.

A rollup executes transactions on its own network, compresses the results and posts them to the main chain along with enough information for anyone to check or reconstruct them. Users get low fees, and the guarantee that their funds can be recovered from the base chain even if the rollup operator disappears.

The two families, and the trade between them

Optimistic rollups assume the posted result is correct and allow anyone to challenge it during a window, typically a week. Cheap, compatible with existing Ethereum tooling, and the withdrawal delay is the price. Most users never notice it, because liquidity providers front the funds for a fee.

Zero-knowledge rollups post a cryptographic proof that the computation was performed correctly. No challenge period, so withdrawals are fast, and the proof itself costs computation to generate. The technology was harder and has matured quickly, and the long-run consensus is that this is where scaling ends up.

What made fees collapse

Rollups pay the base chain for data. In 2024 Ethereum added a dedicated, cheaper form of data specifically for this purpose, which cut rollup costs dramatically. A transaction that cost several euros now costs cents, and for the first time payments and small transfers on Ethereum-based networks became sensible.

This was a deliberate architectural decision: the base layer sells security and data availability, the rollups sell execution. Whether it is a good business for the base layer is exactly the open question discussed in our read on Ethereum after the Merge.

The fragmentation problem

There are now dozens of rollups. Each has its own liquidity, its own applications, its own bridge and its own token. A user with funds on one cannot use an application on another without bridging, which costs time, money and, historically, security: bridges have been the single largest source of theft in crypto.

The ecosystem's answer is a set of standards for shared settlement, common bridging and chain abstraction, where a wallet handles the routing invisibly. These are real and partially deployed, and they add complexity to fix complexity. A user comparing this with a single fast chain that simply works is not being unreasonable.

The centralisation nobody advertises

Most rollups run a single sequencer, one operator that orders transactions. That operator can censor, reorder and extract value from ordering, and if it stops, the network stops until users force an exit through the base chain. Decentralised sequencing is on every roadmap and shipped by few.

This is worth saying plainly because rollups are marketed with the security of the base chain, and the security inherited is about funds, not about liveness or fairness. Your money is safe; your transaction is at the sequencer's discretion.

Bitcoin's version

The Lightning Network takes a different approach: payment channels between parties, settled on the base chain only when opened or closed. It works well for what it was designed for, fast small payments, and it is not a general execution environment. Comparing it to a rollup is comparing a payment rail to a computer.

What to watch

Watch sequencer decentralisation, which is the difference between a rollup and a company with a blockchain-shaped API. Watch whether chain abstraction actually removes bridging from the user experience. And watch proving costs for zero-knowledge systems, which decide how fast the better architecture becomes the default.