ZK Rollups: Validity Proofs & Prover Networks
How zkSync Era, StarkNet, Linea, Scroll, Polygon zkEVM, and Taiko prove L2 execution correct in a single L1 transaction — and what prover hardware actually costs to run.
35 min · expert · part of Rollup Architecture & L2 Engineering
The Validity-Proof Bet
ZK rollups replace the 7-day "wait and see if anyone challenges" model with cryptographic proof. Every batch of L2 transactions is accompanied by a zero-knowledge proof — a SNARK or STARK — that attests "executing these transactions against the prior state root yields this new state root." The L1 verifier contract checks the proof in a single transaction. If the proof is valid, the new state root finalizes immediately; no challenge window.
This is a deeper kind of trust minimization than optimistic rollups achieve. With ZK, you do not need an honest verifier online. You do not need the assumption that censorship is bounded. The L1 contract verifies the math itself. The only trust assumptions remaining are: (1) the SNARK/STARK proof system is sound, (2) the verifier contract is implemented correctly, (3) the data is available (covered in Lesson 4).
The tradeoff has historically been complexity. Building a zkEVM — a ZK proof system that supports the full Ethereum Virtual Machine bytecode — is one of the hardest problems in cryptography. Until 2023, ZK rollups either restricted what code could run (StarkNet's Cairo, zkSync's custom VM) or accepted Ethereum partial-equivalence at the IR level (Type 3 in Vitalik's taxonomy). By 2025-2026, multiple production zkEVMs achieve Type 2 (bytecode equivalent to Ethereum) at acceptable proof times.
As of 2026 ZK rollups still trail optimistic rollups in TVL — StarkNet hovers around $500M, zkSync Era and Linea sit in the few-hundred-million range, Scroll and Taiko smaller still — but they are catching up on tooling and have unique advantages in fast finality, lower bridge-exit times (hours not days), and structural alignment with cryptographic security rather than economic security.
Also in this lesson
- zkSync Era and the ZK Stack
- StarkNet: STARKs and Cairo
- The zkEVM Lineup: Polygon, Linea, Scroll, Taiko
- Prover Networks and Hardware Reality
Key terms
- ZK rollup (validity rollup)
- A rollup that submits a zero-knowledge proof with every batch, allowing the L1 contract to verify state correctness in a single transaction rather than waiting for a fraud-proof window.
- SNARK
- Succinct Non-interactive ARgument of Knowledge. A small (~200-500 byte) cryptographic proof of computation correctness, often based on elliptic-curve pairings. Requires a trusted setup ceremony.
- STARK
- Scalable Transparent ARgument of Knowledge. A hash-based proof system with no trusted setup, post-quantum security, but larger proofs (~100-200 KB). Used by StarkNet and StarkEx.
- zkEVM
- A ZK proof system that supports the full Ethereum Virtual Machine bytecode. Vitalik's taxonomy: Type 1 (fully Ethereum-equivalent), Type 2 (bytecode-equivalent), Type 3 (mostly equivalent), Type 4 (high-level language equivalent).
- Cairo
- StarkNet's native language. Rust-flavored, designed for STARK proving efficiency. Enables optimizations impossible in EVM bytecode but creates a barrier for Solidity developers.
- Based rollup
- A rollup where transaction sequencing is performed by L1 Ethereum validators rather than a separate L2 sequencer. Taiko is the leading example.
- Boojum
- zkSync Era's current proving system. SNARK-based, PLONK-style arithmetization, designed for incremental aggregation and acceptable proof times.
- AggLayer
- Polygon's cross-chain settlement layer that aggregates state proofs from multiple Polygon CDK chains. Part of the strategic pivot away from monolithic Polygon zkEVM (sunsetting July 1, 2026).
- Prover network
- Decentralized infrastructure (Risc Zero Boundless, Aligned Layer, Succinct, =nil; Foundation) for submitting and verifying ZK proofs. Aims to decentralize the prover bottleneck in production ZK rollups.
Continue this lesson — 4 more sections in the CryptoBipto app.
Open lessonEducational only — not financial advice.
