Ethereum rollups have spent the last few years trying to prove that the EVM itself can be turned into a zero-knowledge system without compromise. Linea just spent three years doing exactly that & now it is walking away from it.

On March 29, 2026, Linea the Consensys incubated zkEVM, announced it is pivoting its proving architecture away from direct EVM arithmetization to a RISC-V–based zkVM. The reasoning is not performance alone. It is alignment with Ethereum’s roadmap, with where ZK proving is heading, & with what an eventual “enshrined rollup” might require.

The pitch, put bluntly: proving the EVM directly works, but it may no longer be the right thing to optimize for.

• Why Linea Is Moving Away From Direct EVM Proving
• What RISC-V Actually Changes in Practice
• What Linea Keeps (& Why This Isn’t a Rewrite)
• The Bigger Signal: Ethereum’s Shift Toward RISC-V

Why Linea Is Moving Away From Direct EVM Proving

Linea’s original approach: direct EVM arithmetization was one of the most ambitious attempts in the zkEVM design space. It meant translating Ethereum’s execution model into constraint systems as faithfully as possible, avoiding abstraction layers or virtual machines in between.

Our cryptographic researcher @alexand_belling revealed yesterday at @eth_proofs that Linea is moving to RISC-V.

After 3 years of directly arithmetizing the EVM, producing a 1000+ page spec and one of the most rigorous proving system in production, we’re changing course.

Here’s… pic.twitter.com/jXIF5mZaPT— Linea.eth (@LineaBuild) March 29, 2026

Over three years, the effort produced a 1,000+ page specification & one of the most rigorous proving systems in production. The problem is every Ethereum hard fork required rewriting constraint modules. Every protocol change cascaded into proving complexity & the system kept resetting itself.

The Linea team described it plainly: they were “fighting complexity instead of pushing the frontier”

At some point, the question stops being “can this be done?” & becomes “should this be the thing we keep doing?”

What RISC-V Actually Changes in Practice

The shift to RISC-V is not cosmetic. It changes the foundation of how proofs are generated.

RISC-V is a minimal, standardized instruction set roughly ~40 instructions, compared to the layered complexity of the EVM. That simplicity translates directly into cleaner traces, faster proving pipelines & more predictable system behavior.

Two practical consequences:

• Proving becomes more viable
  Narrower execution traces allow provers to start earlier & operate on smaller chunks, instead of waiting for large batches.
• Type-1 compatibility without reimplementation
  Instead of recreating Ethereum internals (Keccak, RLP, Merkle Patricia Trie) inside constraints, a standard EVM client can compile directly into RISC-V.

This removes one of the biggest hidden costs in zkEVM design: maintaining two versions of Ethereum, one for execution, one for proving.

There is also a more subtle shift here. RISC-V is not just simpler it is legible. It is widely taught, well-documented & already has a growing ecosystem. That lowers the barrier for contributors in a way custom EVM constraint systems never could.