I just saw Vitalik’s post about Ethereum’s major upgrade plan, and I have to say it’s quite bold.

Roughly speaking, he wants to make two fundamental changes that will deeply affect how Ethereum operates at a deep level. The first is the state tree—if you think of it simply, it’s Ethereum’s system for storing and retrieving data. The problem is that it’s currently too large and too complex. Vitalik proposes switching from the Hexadecimal Keccak Merkle Patricia Tree structure to a simpler binary tree. The real-world impact is that the bandwidth required to verify data would drop to about one-fourth of the current amount.

But that still isn’t enough. He also wants to change the hash function, suggesting Blake3 or Poseidon Blake3. Blake3 is faster, but Poseidon seems especially impressive. In theory, it could improve efficiency by dozens of times.

The second change is more intense. He wants to replace the EVM with RISC-V in the long term. The reason is fairly understandable: if all zero-knowledge proof (ZK) systems already use RISC-V, then why have the EVM speak a different language? And why do you need translation between them? If you remove the translation layer, performance would improve automatically.

His plan has three steps. First, use a new virtual machine to run contracts that have already been compiled. Second, allow developers to deploy contracts on the new virtual machine right away. Third, deprecate the EVM, but rewrite it as smart contracts on the new virtual machine for backward compatibility.

Vitalik says that the combined improvements to the state tree and the virtual machine account for more than 80% of Ethereum’s proof limitations. If you don’t touch these two parts, scaling in the ZK era will get stuck right here.

But not everyone agrees. The Offchain Labs team from Arbitrum responded with fairly detailed pushback. They say that RISC-V is suitable for building ZK proofs, but it’s not suitable as a contract delivery/execution format. They recommend using WebAssembly instead, and then compiling to RISC-V to generate proofs. What’s notable is that they’re not just talking—Arbitrum already has a prototype using this approach that works.

They also point out the risk that ZK proof technology changes rapidly. If you lock Ethereum L1 tightly to RISC-V and then, two years later, a better architecture appears, that will be a problem.

What’s interesting is that this ties into a broader trend. Vitalik proposes these improvements while various L2s are gradually “diverging” from Ethereum. A Polygon executive was blunt: the real challenge isn’t scaling, but creating a dedicated space for real-world use cases.

Vitalik himself admits that there still isn’t broad consensus about replacing the EVM, but the state tree upgrade seems more mature. He said that Ethereum has already changed its engine in mid-air once during The Merge, and it can change about four more times.

The Glamsterdam upgrade plan is expected to begin in the first half of 2026, followed by Hegota. In truth, this isn’t a question of “can it be done or not.” Ethereum has proven that it’s willing to make bold improvements at a fundamental level. What will be clear is likely in 2027, but at least Ethereum isn’t planning to become an outdated system that has to be constantly fixed.