One Sentence Summary
Ethereum is preparing the most radical overhaul in its history: swapping out the Ethereum Virtual Machine for the open RISC-V instruction-set architecture to escape the 50–800× performance penalty of zero-knowledge proving and to “snarkify everything” at Layer-1.

Why the EVM Has Become a Dead End

• ZK Bottleneck
  Current zkEVMs don’t prove the EVM itself; they prove an interpreter that is itself compiled down to RISC-V. Vitalik calls this “double taxation”: the extra layer costs 50–800× in proving time and still eats 80–90 % of total block-proving cycles after every other optimization.

• Pre-compile Hell
  Each new cryptographic primitive has historically been stuffed into the protocol as a hard-coded pre-compile. The wrapper code for modexp alone is now larger than an entire RISC-V interpreter, and every pre-compile bloats consensus-critical code that has nearly triggered chain splits twice.

• 256-Bit Baggage
  The EVM’s 256-bit word size is great for big-integer cryptography but disastrous for everyday 32/64-bit math inside ZK circuits, inflating constraints 2–4× and complicating compilers.

Enter RISC-V: A CPU, Not a VM

• Minimal Core – 47 base instructions that map 1-to-1 to silicon.

• LLVM Maturity – out-of-the-box compilers for Rust, C++, Go, Python.

• Formality – the SAIL spec gives a machine-checkable definition that the Yellow Paper never had.

• Hardware Path – ASIC/FPGA provers (SP1, Nervos, Cartesi) already in test nets.

The Three-Stage Migration Roadmap

1. Pre-compile Probe
   Ship RISC-V as a pre-compiled module inside the current EVM—low-risk, opt-in.

2. Dual-VM Era
   Run EVM and RISC-V side-by-side with full bidirectional calls; dApps can migrate gradually.

3. Rosetta Flip
   Re-implement the EVM inside RISC-V (à-la Rosetta) and sunset the legacy interpreter.

Fallout Across the Stack

• Optimistic Rollups (Arbitrum, Optimism) – must rebuild fraud provers for RISC-V.

• ZK Rollups (Polygon, zkSync, Scroll) – get an overnight 100× cost cut; proof generation collapses from minutes to seconds.

• Developers – write L1 contracts in safe, mainstream languages; no more Solidity-only ghetto.

• Users – Gigagas L1: ~10 k TPS at base-layer fees that feel like side-chains.

Lean Ethereum: The Endgame

The foundation’s new north star is “Lean Ethereum”:

• Lean Consensus – single-slot finality

• Lean Data – Danksharding + EIP-4444

• Lean Execution – RISC-V replacing bespoke EVM

Justin Drake sums it up: “We’re not upgrading the EVM; we’re replacing it so that everything—consensus, data, execution—can be snarkified end-to-end.”

The decade-long detour through custom opcodes, pre-compiles, and 256-bit math is being rolled back. Ethereum is rebooting on an ISA that was designed in a university lab, battle-tested in billions of phones, and now ready to become the world’s most auditable, provable, and future-proof computer.

One Sentence Summary
Ethereum is preparing the most radical overhaul in its history: swapping out the Ethereum Virtual Machine for the open RISC-V instruction-set architecture to escape the 50–800× performance penalty of zero-knowledge proving and to “snarkify everything” at Layer-1.

Why the EVM Has Become a Dead End

• ZK Bottleneck
  Current zkEVMs don’t prove the EVM itself; they prove an interpreter that is itself compiled down to RISC-V. Vitalik calls this “double taxation”: the extra layer costs 50–800× in proving time and still eats 80–90 % of total block-proving cycles after every other optimization.

• Pre-compile Hell
  Each new cryptographic primitive has historically been stuffed into the protocol as a hard-coded pre-compile. The wrapper code for modexp alone is now larger than an entire RISC-V interpreter, and every pre-compile bloats consensus-critical code that has nearly triggered chain splits twice.

• 256-Bit Baggage
  The EVM’s 256-bit word size is great for big-integer cryptography but disastrous for everyday 32/64-bit math inside ZK circuits, inflating constraints 2–4× and complicating compilers.

Enter RISC-V: A CPU, Not a VM

• Minimal Core – 47 base instructions that map 1-to-1 to silicon.

• LLVM Maturity – out-of-the-box compilers for Rust, C++, Go, Python.

• Formality – the SAIL spec gives a machine-checkable definition that the Yellow Paper never had.

• Hardware Path – ASIC/FPGA provers (SP1, Nervos, Cartesi) already in test nets.

The Three-Stage Migration Roadmap

1. Pre-compile Probe
   Ship RISC-V as a pre-compiled module inside the current EVM—low-risk, opt-in.

2. Dual-VM Era
   Run EVM and RISC-V side-by-side with full bidirectional calls; dApps can migrate gradually.

3. Rosetta Flip
   Re-implement the EVM inside RISC-V (à-la Rosetta) and sunset the legacy interpreter.

Fallout Across the Stack

• Optimistic Rollups (Arbitrum, Optimism) – must rebuild fraud provers for RISC-V.

• ZK Rollups (Polygon, zkSync, Scroll) – get an overnight 100× cost cut; proof generation collapses from minutes to seconds.

• Developers – write L1 contracts in safe, mainstream languages; no more Solidity-only ghetto.

• Users – Gigagas L1: ~10 k TPS at base-layer fees that feel like side-chains.

Lean Ethereum: The Endgame

The foundation’s new north star is “Lean Ethereum”:

• Lean Consensus – single-slot finality

• Lean Data – Danksharding + EIP-4444

• Lean Execution – RISC-V replacing bespoke EVM

Justin Drake sums it up: “We’re not upgrading the EVM; we’re replacing it so that everything—consensus, data, execution—can be snarkified end-to-end.”

The decade-long detour through custom opcodes, pre-compiles, and 256-bit math is being rolled back. Ethereum is rebooting on an ISA that was designed in a university lab, battle-tested in billions of phones, and now ready to become the world’s most auditable, provable, and future-proof computer.