Guide · Markets
Layer-2 Rollups Explained: Why Arbitrum, Optimism, Starknet and ZKsync Exist
Ethereum is secure but slow and expensive to use directly, so most activity has moved to 'layer-2' networks built on top of it. Here's how optimistic and zero-knowledge rollups work, and why their tokens behave the way they do.
Ethereum can only process a limited number of transactions per block, and when demand is high, users compete for that space by paying higher fees. For years, that made simple actions such as swapping tokens expensive. Instead of changing Ethereum itself dramatically, developers built networks that sit on top of it. These are called layer-2 networks, or L2s, and the most common design is the rollup.
A rollup does most of the work off Ethereum. It collects many transactions, executes them on its own system, and then posts a compressed record of the results back to Ethereum — 'rolling up' hundreds of transactions into one. Because Ethereum stores that record, users can, in principle, rely on Ethereum's security while paying a fraction of the cost. The key question is how Ethereum knows the rollup's results are correct, and that's where the two main families differ.
Optimistic rollups, such as Arbitrum and Optimism (and Base, which is built with Optimism's software), assume the posted results are valid unless someone proves otherwise. There's a challenge window, typically around a week, during which anyone can submit a fraud proof. It's a simpler design, but it's the reason withdrawals back to Ethereum through the official route can take days. Zero-knowledge rollups, such as Starknet and ZKsync, take the opposite approach: they attach a cryptographic proof to each batch showing that the computation was done correctly. The proofs are complex to produce, but once verified on Ethereum, the results are final much faster.
Some networks go further and store their data somewhere other than Ethereum to cut costs. Celestia was built for exactly that: a blockchain that specialises in making data available cheaply for rollups. Services such as Caldera help projects launch their own rollups with these components already assembled.
For traders, the important point is that the success of a layer-2 network and the price of its token are only loosely connected. Most L2 tokens are used mainly for governance, while users pay transaction fees in ether. Many were launched with a large share of their supply locked for early investors and team members, and the steady release of those tokens has weighed on prices for long periods. That's a big part of why many L2 tokens fell 80-90% from their highs even as activity on the networks grew.
It also explains why their rallies can be so sharp when they do come. After long declines, these tokens have low prices, plenty of sellers already gone, and a clear narrative to rally around. In October 2026, Starknet more than doubled in three days, from $0.048 to above $0.12, and Celestia gained 27% in a single session, while ether barely moved. Moves like that are driven by speculative rotation rather than a sudden change in how the networks are used, so it's worth checking upcoming token unlocks and comparing the token's performance with ether's before reading too much into them.