Ethereum Layer 2 networks process transactions separately from Ethereum's main chain, then settle the results back to it, allowing far cheaper and faster transactions while still relying on Ethereum's underlying security. They exist to solve the capacity limits of Ethereum operating alone. The Ethereum Foundation maintains an overview of the networks and how they differ.
Why L2s exist
Ethereum's base layer can only process a limited number of transactions per block, which pushes fees higher during busy periods. That pressure shows up directly in the gas fee you pay. Layer 2 networks were built specifically to relieve that bottleneck.
The limit is deliberate rather than a failure of engineering, and this is the part that makes the whole design make sense. Ethereum could raise its capacity tomorrow by allowing bigger blocks. The cost would be that keeping up with the chain requires more powerful hardware, which prices ordinary people out of running nodes, which concentrates verification into a handful of professional operators. A chain only a few companies can check is a chain those companies effectively control.
So the base layer stays small on purpose, and the scaling happens above it. Rather than making the foundation carry more, the strategy is to do the work elsewhere and use the foundation to settle the result. That is the entire logic of Layer 2, and it is also why "why doesn't Ethereum just get faster?" has a real answer rather than an excuse.
Rollups
Most major Layer 2 networks use a technique called rollups, which bundle many transactions together off the main chain and post a compressed summary back to Ethereum, inheriting much of its security while dramatically cutting costs.
The economics are simply amortization. Posting a transaction to Ethereum costs the same whether it represents one transfer or a thousand batched together, so splitting one base-layer fee across thousands of users makes each one's share negligible. The security argument is the more interesting half: the transaction data goes onto Ethereum, so anyone can reconstruct the rollup's entire state from the base chain alone. Even if every server the rollup runs on vanished tomorrow, the record survives, and users can recover their funds.
Two flavors exist and the difference is how they prove they did not cheat. Optimistic rollups assume the operator is honest and allow anyone to challenge a bad result during a dispute window, which is typically about a week; that window is why withdrawing from Arbitrum or Optimism to Ethereum takes seven days unless you pay a service to front you the funds. Zero-knowledge rollups instead post a cryptographic proof that the computation was correct, which is mathematically stronger and allows fast withdrawals, at the cost of far more complex engineering.
Major L2s
Arbitrum and Optimism are among the most widely used Ethereum rollups, each hosting a large share of DeFi and app activity that would otherwise run more expensively on the base layer.
Both are optimistic rollups, both are broadly compatible with Ethereum's tooling, and from a user's seat they feel like Ethereum with the fees removed: the same wallet, the same addresses, the same applications, mostly the same interfaces. The zero-knowledge camp, including networks built on newer proving systems, is smaller today but improving quickly, and the general expectation is that proofs eventually win on merit.
The strategic thing to understand is that Ethereum has explicitly organized its roadmap around this outcome. The base layer's job is increasingly to be a settlement and data-availability layer for rollups rather than a place where ordinary activity happens. That is not a stopgap while something better is built; it is the plan.
Trade-offs
Layer 2s are generally faster and cheaper, but they add some complexity: moving funds between layers takes an extra step, and each L2 has its own trust and security nuances worth understanding before relying on it heavily.
The nuance that matters most is centralization. Nearly every major rollup today runs a single sequencer, one server that decides transaction order, usually operated by the team that built it. If it goes down, the network stops, and this has happened. If it censors you, your recourse is a slow escape hatch through the base layer. Most rollups also retain upgrade keys, held by a multisig, that can alter the contracts governing your funds. The security is real but it is not yet Ethereum's, and honest projects publish exactly where they sit on that spectrum.
The practical hazards are more mundane and cost people money more often. Assets do not automatically exist across layers: sending funds to the right address on the wrong network is a standard way to lose them permanently. Each L2 needs its own gas, and each maintains its own version of tokens like USDC, which are not interchangeable with the base-layer version despite the identical name. Bridges between layers have been the single most-attacked component in crypto, and a bridge is exactly what you use to get in and out.
None of which is an argument against using them. For anything short of large, infrequent transfers, an L2 is the sensible default and the fee difference is not close. It is an argument for knowing which network you are on before you press send.
Beyond rollups
Not everything marketed as a Layer 2 is one, and the distinction has real consequences for your money. A genuine rollup posts its transaction data to Ethereum, which is what lets anyone reconstruct the state and exit without the operator's cooperation. Some networks skip that step, keeping data on their own servers to cut costs further, and while the savings are real, so is the change in what you are trusting: if that data becomes unavailable, proving what you own becomes someone else's problem to solve for you.
Sidechains sit further out still. They are separate blockchains with their own validators and their own security, connected to Ethereum by a bridge, and they inherit nothing from Ethereum at all. Calling them Layer 2s is common and misleading. Your funds there are protected by that chain's validator set, not by Ethereum, which is a completely different proposition regardless of how the marketing reads.
The reasonable posture is to treat the label as a starting question rather than an answer. Ask where the data goes, who can stop the network, and whether you could withdraw if the team disappeared tomorrow. Independent trackers now publish exactly this for each network, graded rather than asserted, and the honest ones make for sobering reading.
The direction of travel is encouraging even so. Sequencers are being decentralized, proof systems are maturing, and Ethereum's own upgrades have made posting rollup data dramatically cheaper, which is why L2 fees fell by an order of magnitude rather than gradually. The gap between "secured by Ethereum" as a slogan and as a fact is closing, and it is closing faster than most of the surrounding argument suggests.