Layer 1 vs Layer 2 (and why everyone talks about 'scaling')

Two layers, one ecosystem

If you only remember one architectural fact about the modern Ethereum ecosystem, remember this. By 2026, most actual user activity does not happen on Ethereum itself. It happens on layer 2 networks built on top of Ethereum.

This is by design. Ethereum mainnet can process roughly fifteen transactions per second at its base layer. That is the price of its security and decentralization. Layer 2 networks process thousands of transactions per second by handling execution off the base chain while inheriting Ethereum's security guarantees.

The transition is not a competition between layers. It is a deliberate division of labor. Mainnet is for high-value settlement. Layer 2 is for everything else.

What "layer 1" actually means

A layer 1 blockchain is an independent network with its own consensus mechanism and its own validators. Bitcoin is a layer 1. Ethereum is a layer 1. Solana, Avalanche, and Cardano are layer 1s. They each have their own security model, their own native token, and their own community.

Each layer 1 makes its own tradeoff on the blockchain trilemma (covered in detail in the readings). Bitcoin trades throughput for maximum security and decentralization. Ethereum trades raw speed for programmability and a mature security model. Solana trades some decentralization for very high throughput. None of these tradeoffs is wrong; they reflect different priorities.

What unites layer 1s is that each one is self-contained. The integrity of the chain depends on the chain's own validators and consensus. There is no fallback to anyone else.

What "layer 2" actually means

A layer 2 network is structurally different. It is a separate blockchain that processes transactions independently but settles its final state back to a layer 1, almost always Ethereum.

The key word is inherits. An L2 borrows the layer 1's security. The integrity of the L2 ultimately depends on the integrity of the L1 underneath. This is what distinguishes an L2 from an alternative L1.

Practically, this means the L2 can be much faster and much cheaper than the L1 while still being economically secured by the L1's validators. Users get the throughput benefits of the L2 plus the security guarantees of the L1.

The architectural insight is that not every transaction needs every layer 1 validator to process it independently. Most transactions just need a final settlement record. The L2 does the work; the L1 verifies the record.

Optimistic rollups vs zk-rollups

The dominant L2 architecture in 2026 is the rollup. Rollups bundle many transactions into a single batch and commit the result to the base layer. There are two main flavors that differ in how they verify the batch.

Optimistic rollups assume the batch is valid by default. The L2 publishes its transactions to the L1 and provides a window — typically seven days — during which anyone can submit cryptographic proof if they believe a transaction was processed incorrectly. If no challenge appears, the batch is considered final.

The advantage is simplicity and low cost. The drawback is the seven-day withdrawal window. To move funds back to mainnet, you wait for the fraud-proof window to elapse. Arbitrum and Optimism are the two largest optimistic rollups.

Zero-knowledge rollups (zk-rollups) take a different approach. Each batch comes with a cryptographic proof that all transactions were processed correctly. The proof is small and cheap for the L1 to verify, but computationally expensive to generate. Once the L1 accepts the proof, the batch is final — no waiting period.

zk-rollups offer faster withdrawals (minutes, not seven days) but have historically been more complex to implement, especially for general-purpose smart contracts. Both have been improving rapidly. zkSync, Scroll, Linea, and StarkNet are major zk-rollups.

The major L2s worth knowing

By total value locked and unique users in 2026, the dominant L2s are:

Arbitrum is the largest by both metrics. Optimistic rollup architecture. Full EVM compatibility (most Ethereum applications work without modification). Mature DeFi ecosystem.

Optimism is the second-largest optimistic rollup. Also the foundation for the Superchain — a vision of multiple chains sharing the same security and infrastructure. Coinbase's Base network is built on Optimism's tech stack.

Base is Coinbase-backed and grew rapidly through 2024-2025. Particularly strong adoption in consumer-facing crypto applications. Uses the OP Stack from Optimism.

zkSync Era is the largest zk-rollup by activity. Full EVM compatibility. Matter Labs is the operator.

Scroll, Linea, Polygon zkEVM are newer zk-rollups with various tradeoffs in compatibility and performance.

StarkNet uses a different programming model (Cairo) but offers strong performance for specific applications.

For most users, the practical answer is: use whichever L2 the application you want to use is deployed on. For general DeFi, Arbitrum has the deepest liquidity. For Coinbase-related and consumer-facing applications, Base. For faster withdrawals, a zk-rollup.

The real tradeoffs

The performance gains of L2s come with real tradeoffs. These are not deal-breakers but they are worth understanding.

Sequencer centralization. Most L2s currently rely on a single "sequencer" — a centralized operator that orders transactions before they get committed to the L1. This is an operational centralization point that could theoretically censor transactions or delay them, though it cannot steal funds (the L1 fraud-proof mechanism prevents that). Decentralizing the sequencer is active work but has not yet been achieved on most major L2s.

Bridge risk. Moving assets from Ethereum to an L2 (or back) involves a bridge — a smart contract on each side that locks and mints the corresponding asset. L2 bridges that are part of the rollup's own security model are much safer than third-party bridges, which have been the largest single source of crypto hacks. Use the official bridge for whatever L2 you are targeting.

Withdrawal delays. Optimistic rollups have the seven-day withdrawal delay. Third-party "fast bridges" offer immediate withdrawals in exchange for a fee, but they introduce their own counterparty risk.

Fragmented liquidity. The same asset (USDC, for instance) exists in different bridged forms on different L2s. Aggregators have largely solved this for users in practice, but it is worth knowing the friction exists.

The practical takeaway

For users entering the modern Ethereum ecosystem in 2026, the practical guidance is simple. Start on Base or Arbitrum for low-cost everyday DeFi and consumer apps. Use a zk-rollup like zkSync if fast withdrawals matter for your specific use case. Use Ethereum mainnet only for high-value long-term holds or applications that specifically require it.

The shift from "Ethereum mainnet" to "Ethereum plus L2s" as the relevant architectural unit is one of the most important changes in crypto over the past four years. The protocols, the developers, and the user experience are all moving onto L2 rails. You will use them whether or not you realize you are using them.

The next module looks at the most-used product in all of crypto: stablecoins. Whatever your relationship with crypto, stablecoins are the part of the on-chain economy you are most likely to encounter directly, and understanding what they are — and what they are not — is essential.