What Is Actually New in ZK Rollups in 2026? A Plain Progress Map

In 2022 the ZK Rollup conversation was “in three years”. In 2024 it was “can we use it this year”. In 2026 it has become “what are people using and what is still missing”. The track went through three clear jumps in those two years: the first public zkEVM mainnets, then EIP-4844 Blob data availability cutting L2 data costs to a tenth, and then recursive proofs plus GPU proving dropping proof time from minutes to seconds. This article maps the real state in May 2026, no jargon piles, each item answered by “where it is now, what is still missing, what it means for regular users”.

Quick refresher: what ZK Rollups solve

A ZK Rollup compresses thousands of L2 transactions into a single zero-knowledge proof sent back to L1, where the contract only verifies the proof and does not re-execute each transaction. The core difference vs Optimistic Rollup: ZK uses cryptography to directly prove validity, no 7-day challenge window; Optimistic assumes “you’re right unless challenged”. One line: ZK trades compute for time.

If the two routes still feel fuzzy, read the L2 guide and ZK Rollup intro first, then the 2026 progress below will read more cleanly.

Progress one: most zkEVMs approaching Type 1

Vitalik defined an early zkEVM taxonomy — Type 1 to Type 4 — based on how closely they match the EVM:

The biggest 2026 change: Type 1 is no longer just a paper goal. Taiko shipped a Type-1-equivalent mainnet in late 2025; Scroll completed phase-two Type 1 work in Q1 2026. The same contract bytecode now executes identically on Ethereum and on those zkEVMs — adapters are nearly unnecessary for developers.

zkSync Era and Starknet stay on their own VMs (Type 4) but both shipped EVM compatibility layers during 2025–2026, letting Solidity contracts deploy directly and dropping migration cost further.

Progress two: prover time from minutes to seconds

In 2023 most zkEVMs needed 3–15 minutes per proof, meaning “L2 feels instant but cross-chain withdrawal takes hours”. In 2026:

• Linea: proofs in 30–90 seconds, L1 finality typically 8–15 minutes.
• Scroll: proofs in 20–60 seconds, finality 6–12 minutes.
• zkSync Era: proofs in 40–120 seconds, finality 10–20 minutes.
• Starknet: proofs in 60–180 seconds, finality 15–30 minutes.
• Polygon zkEVM: proofs in 60–150 seconds, finality 10–25 minutes.

Three things drove the speedup. First, recursive proofs — proofs embedded inside proofs, merging many small proofs into one, multiplying throughput. Second, GPU proving clusters — ZK algorithms no longer rely on single-core CPU, and major L2s built dedicated proving farms. Third, new schemes like Plonky2/Plonky3 — unifying curves and commitment schemes, smaller proofs and faster verification.

For regular users the most visible effect is withdrawal time. L2 to L1 used to take 7 days (Optimistic) or roughly an hour (early ZK); ZK Rollups now usually settle withdrawals in under 30 minutes.

Progress three: data availability after EIP-4844

After EIP-4844 Blobs shipped in 2024, every Rollup swapped Calldata for Blob storage and data costs dropped to a tenth or twentieth overnight. The L2 economic model was reshaped:

• L2 per-transaction gas fell from $0.01–0.05 to $0.001–0.005.
• L1’s Blob queue started showing congestion in late 2025; Blob pricing now has its own “base fee + priority” secondary-market dynamic.
• Some projects use Celestia, EigenDA, Avail as off-chain DA layers — the Validium / Volition route, pushing data off L1 to cut costs further; see Celestia and modular blockchains.

For the rest of 2026 watch Blob capacity expansion after Pectra; see Pectra upgrade explained. If Blob capacity doubles, L2 per-tx cost slides further, which matters most for SocialFi and on-chain games — high-frequency, small-amount use cases.

What is still missing and what it means for you

Three things remain unsolved despite the progress:

1. Proving system diversity is thin — most zkEVMs depend on a single proving system, so a bug there hits many chains at once.
2. Exit windows are still trust-tiered — many zkEVMs’ “emergency exit” paths are not fully permissionless yet.
3. Cross-Rollup interop — L2-to-L2 messaging still depends on third-party bridges; native ZK interop protocols remain in pilot.

For regular users, 2026 is the moment to treat ZK L2s as everyday chains: transfers, swaps, NFT mints, SocialFi, airdrops — UX and cost are nearly free. For large cross-chain moves, still batch them and keep a rollback plan. The ZK Rollup track has graduated from “future” to “daily”, and the next phase will be decided by whoever lands cross-Rollup interop and proving-system diversity first.