I just realized something quite interesting: the way Ethereum perceives the role of L2s is changing. It's not that L2s are weakening, but that Ethereum itself is "regaining control" over its core capabilities. Vitalik has just proposed a rather bold idea—changing the entire way Ethereum processes transactions at the fundamental layer.

Specifically, two things: first, changing the data structure of the current (state tree) because it’s too "bulky" with the Keccak Merkle Patricia Tree(. Switching to a simpler binary tree, the Merkle branch length would be reduced to a quarter, meaning light clients would need significantly less bandwidth to verify data. This is EIP-7864 that Vitalik is pushing.

Second, and even bolder: replacing the EVM with a RISC-V architecture. It sounds crazy, but the logic is quite straightforward—if ZK proof systems are already using RISC-V, why should the virtual machine use a different language and add a translation layer? Removing that translation layer automatically boosts performance. A RISC-V interpreter only needs a few hundred lines of code.

But this idea isn’t universally supported. Offchain Labs’ )development team of Arbitrum( published a detailed rebuttal in November. They say RISC-V is good for generating ZK proofs, but not the best for transaction formats. Instead, they propose using WebAssembly )WASM( for the contract layer, then compiling down to RISC-V to generate proofs. The two layers operate independently. The reasoning is also sound: WASM runs efficiently on standard hardware, while most Ethereum nodes don’t run RISC-V chips.

A bigger issue Vitalik points out is that the current state tree and virtual machine account for over 80% of Ethereum’s proof bottlenecks. In other words, if these two parts aren’t addressed, scaling in the ZK era will be locked in place. Vitalik even says that Ethereum has already changed its engine once )The Merge, and might need to change it about four more times—including reforms to the state tree, replacing EVM, and other things.

What’s interesting is that L2s aren’t panicking. Instead, they’re looking for reasons to justify their independent existence. Jing Wang from OP Labs says L2s are like independent websites, while Ethereum remains the open standard for payments at the base layer. Polygon’s CEO is even more direct: the real challenge isn’t scaling, but creating a unique block space for real-world use cases.

Vitalik also admits that replacing the EVM still lacks broad community consensus. The reform of the state tree is more mature, but replacing EVM with RISC-V is still in the roadmap stage. Ethereum Glamsterdam is expected to deploy in the first half of 2026, followed by Hegota, but the exact details haven’t been finalized yet.

Looking back at history, Ethereum has demonstrated the ability to shift from PoW to PoS, from L1 all-in to a focus on Rollups. This time, it’s not about adding new features but about reworking the old foundation. Is this a long-term overhaul or an ever-deepening pit of complexity? Perhaps the answer will come in 2027. But at least Ethereum isn’t aiming to become a "legacy system that needs patching" in the ZK era. This debate might be more valuable than the conclusion itself.