What Is ZKsync Era? The LLVM zkEVM and Elastic Network, Explained

Most Layer 2s pick a lane: copy the EVM as faithfully as possible, or rebuild it for speed and hope the tooling catches up. ZKsync Era took the second road and never really came back. Built by Matter Labs, it is a general-purpose ZK rollup that proves every batch correct with cryptography instead of trusting an operator and waiting out a challenge window — and it runs on a custom virtual machine, fed by an LLVM compiler, that deliberately breaks bytecode-equivalence to get something most chains can't offer: account abstraction baked into the protocol itself. This is the full picture — how the proving works, why EraVM is different, what the ZK token does (and doesn't) do, and the risks worth knowing before you bridge in.

Who built it, and why

ZKsync is the work of Matter Labs, a research-heavy team led by co-founder Alex Gluchowski that has been building zero-knowledge scaling for Ethereum since 2018. The thesis has always been the maximalist one: ZK proofs are the endgame for Ethereum scaling, so build for that future even if it's harder in the near term.

The first product, ZKsync Lite (originally zkSync 1.0, live in 2020), was a payments-only rollup — you could transfer and swap tokens but not run arbitrary contracts. The real swing came in March 2023, when ZKsync Era launched to mainnet as one of the first zkEVMs in production: a rollup where developers could deploy Solidity contracts and prove their execution with validity proofs. (ZKsync Lite is being sunset in 2026, with a migration plan, now that Era has long since eclipsed it.)

The branding also shifted along the way — the project restyled itself from "zkSync" to ZKsync and reframed the whole thing around a network of chains rather than a single rollup. Which brings us to the architecture.

The architecture: a real ZK rollup

ZKsync Era is a validity rollup, and that word matters. Two families dominate Ethereum L2s:

• Optimistic rollups (Arbitrum, Optimism, Base) assume batches are honest and give anyone a window — roughly seven days — to submit a fraud proof if they aren't. Security is real but withdrawals back to Ethereum inherit that challenge delay.
• ZK (validity) rollups like ZKsync Era prove each batch correct before it settles, by posting a zero-knowledge validity proof to Ethereum. If the proof verifies, the state is final. No challenge window, no seven-day exit.

ZKsync Era's proving stack is Boojum, an implementation of the Redshift protocol built on a PLONK-style arithmetization with a FRI-based polynomial commitment. In plain terms: its STARK/FRI core is transparent, needing no trusted setup of its own — a meaningful security property, since a botched or compromised trusted setup is a classic ZK failure mode. That STARK proof is then wrapped in a small final PLONK+KZG SNARK so it's cheap to verify on Ethereum, and that wrap does reuse an existing trusted-setup ceremony (Aztec's Ignition), rather than one ZKsync ran itself. Ethereum verifies the proofs on L1, so the correctness of your funds ultimately rests on math and Ethereum, not on Matter Labs' honesty.

On data availability, ZKsync Era operates as a true rollup by default: it posts transaction data ("pubdata") to Ethereum, so anyone can reconstruct the chain's state from L1 alone. That's the property that separates a rollup from a validium. The broader ZK Stack framework does let other chains run in validium mode, pushing data to external DA layers such as Avail, Celestia, or EigenDA — but that's a per-chain choice, and Era itself is the rollup configuration.

The honest asterisk is decentralization. On L2BEAT, ZKsync Era is a Stage 0 rollup — the training wheels are still firmly on. The sequencer that orders transactions is operated by Matter Labs, proving is not yet permissionless, and upgrade keys exist. Your funds are protected by validity proofs regardless, but liveness and censorship resistance depend on the operator until the roadmap's permissionless-proving work (the Boojum 2.0 effort, targeted around late 2026) actually lands.

EraVM, the LLVM compiler, and native account abstraction

Here's what makes ZKsync Era genuinely distinctive, and also its most-argued-about design choice.

Most zkEVMs chase bytecode-equivalence — they execute the exact same EVM bytecode Ethereum does, so every tool, debugger, and precompile just works. ZKsync Era doesn't. Instead it compiles Solidity, Vyper, and Yul through a custom LLVM-based toolchain (zksolc) into EraVM, its own zk-optimized virtual machine. EraVM is EVM-compatible at the language level but not bytecode-equivalent at the machine level.

The cost of that decision is friction: historically some Ethereum tooling, certain precompiles, and low-level assembly tricks behaved differently, which could bite during debugging and audits. The payoff is what the extra freedom buys — chiefly native account abstraction.

On Ethereum mainnet, account abstraction is bolted on via ERC-4337 and a separate mempool. On ZKsync Era it's native: at the protocol level, every account can be a smart contract. In practice that means:

• Smart-contract wallets by default — social recovery, multisig logic, session keys, spending limits, all first-class.
• Paying gas in ERC-20 tokens, not just ETH, because the account can define how fees are settled.
• Paymasters — a dApp can sponsor your gas so onboarding feels free, or accept a stablecoin for fees.

This is the feature builders keep coming back for, and it's why consumer-facing projects gravitate to the stack. In 2025 Matter Labs also shipped an EVM interpreter that lets unmodified EVM bytecode run on top of EraVM — closing much of the historical compatibility gap while keeping the account-abstraction advantage. You increasingly get both: standard Ethereum tooling and native smart accounts.

Fees, speed, and the Atlas upgrade

ZK rollups amortize a lot of cost by proving thousands of transactions in one batch, and ZKsync Era's fees reflect that — typically fractions of a cent per transaction, far below Ethereum L1.

The big 2026 performance story is the Atlas upgrade to the ZK Stack (rolled out from late 2025). Atlas targets 15,000+ transactions per second with roughly one-second finality and near-zero fees, backed by a next-generation prover (Airbender) built for sub-second confirmations. The headline numbers are aspirational ceilings, as such figures always are, but the direction is clear: ZKsync is optimizing for the throughput and latency that consumer apps and institutional rails actually need, not just for lower fees.

The ZK token and incentives

Yes, there's a token — and its role is commonly misunderstood.

The ZK token launched with an airdrop in June 2024, distributing 17.5% of supply (about 3.675 billion tokens) to roughly 695,000 eligible wallets, with no lockup on the airdropped portion. Maximum supply is 21 billion. Allocations skew toward the community and ecosystem (a large "Token Assembly" governance allocation plus ecosystem initiatives), with team and investor tranches vesting over four years (a one-year cliff that passed in June 2024, then gradual monthly unlocks through 2028).

The key thing to internalize: ZK is not the gas token. You still pay transaction fees in ETH. ZK is a governance and staking asset — it votes in ZKsync's on-chain governance (structured as ZK Nation and its Token Assembly) and, since a staking pilot that began in February 2026, can be staked for rewards streamed over 30-day epochs with no lockups. So the token's value case rests on governance weight and future fee/value accrual, not on being a mandatory gas currency.

The Elastic Network and the ecosystem

ZKsync's ambition isn't a single chain — it's a network of ZK chains. The ZK Stack is Matter Labs' open framework for deploying your own ZK-powered rollup or validium, and all such chains form the Elastic Network: a mesh that shares a bridge contract on Ethereum L1 and native bridging between chains, so liquidity and messages move without wrapped assets or trusted third-party bridges. (This is the idea earlier pitched as "Hyperchains.")

The most visible proof point is Abstract — a consumer-focused chain built on the ZK Stack, associated with the Pudgy Penguins ecosystem, that runs as its own Elastic Network chain rather than as apps on Era. That's the model: ZKsync Era is the flagship general-purpose rollup, and app- or brand-specific chains spin up alongside it while staying interoperable.

On Era itself, the ecosystem leans DeFi and consumer: SyncSwap and other DEXs anchor liquidity, LayerZero provides cross-chain messaging, and gaming/NFT projects use the cheap-and-abstracted UX. In total-value terms, ZKsync Era has consistently ranked among the largest ZK rollups, with multi-billion-dollar value secured (in the roughly $4–5 billion range through late 2025 into 2026 per L2BEAT). Treat any single TVL figure as a moving snapshot, not a fixed fact.

How to get on: bridging and wallets

Getting onto ZKsync Era is straightforward:

1. Add the network. Any EVM wallet — MetaMask, Rabby, and others — works once ZKsync Era's RPC is added. Thanks to native account abstraction, you'll also see smart-account and SSO-style login options in ZKsync-native apps.
2. Bridge assets in. The official ZKsync bridge (Portal) moves ETH and tokens from Ethereum L1 to Era via the canonical bridge. It's the most trust-minimized route but settles on Ethereum's schedule.
3. Or withdraw from an exchange. Many centralized exchanges support direct withdrawals to ZKsync Era, which is often the cheapest way to arrive with fresh funds. Third-party bridges offer speed for a fee.

One rule matters more than the rest: verify the bridge URL yourself, from official docs or your own bookmarks. Fake "ZKsync bridge" sites that drain wallets are one of the most common scams in the entire L2 space.

Prividium and the institutional bet

A notable 2026 thread is ZKsync's push into regulated finance via Prividium — a privacy-preserving deployment of the stack pitched as a full banking/enterprise system, where a bank or market-infrastructure provider runs a permissioned ZK chain that still settles proofs to Ethereum. ZKsync has publicized institutional interest (names like Deutsche Bank, UBS, Sygnum, and Chainlink have appeared in its materials), and initiatives such as the Cari Network onboarding U.S. regional banks target pilots later in 2026. This is early and largely pilot-stage — worth watching, not yet a proven revenue engine — but it explains where Matter Labs is steering: real-world assets and regulated rails, not just retail DeFi.

Risks and what to watch

A clear-eyed list, because the sober version is more useful than the pitch:

• Centralization (Stage 0). The Matter Labs–run sequencer and non-permissionless proving are the honest weak point. Validity proofs protect your funds' correctness, but the operator controls ordering and liveness until decentralization ships. Watch the Boojum 2.0 / permissionless-proving milestones.
• Compatibility edge cases. The non-bytecode-equivalent EraVM has closed much of its gap with the EVM interpreter, but subtle differences in tooling, gas, and precompiles can still surprise developers. Audit assumptions carefully.
• Token value accrual is unproven. ZK is governance/staking, not gas, so its long-term value depends on governance demand and future fee-capture mechanisms rather than mandatory usage. Ongoing team/investor unlocks add supply pressure through 2028.
• Competition. The zkEVM field is crowded — Linea, Scroll, Polygon zkEVM, StarkNet — all chasing the same liquidity and developers. Leadership isn't guaranteed.
• Bridge phishing. As above: the biggest immediate risk to you personally is a fake bridge or airdrop-claim site, not a protocol failure.

Bottom line

ZKsync Era is one of the most opinionated L2s in the market. It committed early to validity proofs built on a transparent STARK/FRI core (needing only a small final SNARK wrap, for cheap on-chain verification), and it paid a real compatibility cost — a custom LLVM compiler and EraVM that isn't bytecode-equivalent — to win something durable: account abstraction as a native, protocol-level feature, plus a whole Elastic Network of interoperable ZK chains built on the same stack. The trade-offs are equally real: it's still Stage 0, the sequencer is Matter Labs', and the ZK token is governance, not gas. If you value the ZK-rollup security model and want the smoothest smart-account UX in the L2 space, Era is a serious home. Just go in understanding that the decentralization story is a roadmap, not a finished product.

Related guides

• Starknet
• Status Network
• Scroll
• Linea

For related reading: what is DeFi, best Ethereum L2s, account abstraction explained, and how to bridge to Layer 2.

Not financial advice. ZK is volatile, ZKsync Era is a Stage 0 rollup, and bridge and token details change — always verify on official ZKsync channels and your own bookmarks.