Ethereum co-founder Vitalik Buterin said the network is approaching an inflection point as two major upgrades, PeerDAS and zkEVMs, move from research to working code.
In a post on X, Buterin argued that the combination could move Ethereum to “a fundamentally new and more powerful form of decentralized network” because it addresses the central trade-off that has historically limited blockchains, where a system can be decentralized and have consensus, but bandwidth and throughput remain low.
Loading…
He framed the problem through two models from the Internet age. Systems like BitTorrent can move huge amounts of data in a decentralized way, but they don’t need consensus. Bitcoin has strong decentralization and consensus, but it remains low bandwidth because each node effectively rechecks the same work instead of splitting it up.
Ethereum’s next phase, he said, is about getting all three at once.
The first stage is already underway. PeerDAS (data availability sampling) is now on the Ethereum mainnet, letting nodes verify that data is available without downloading the full dataset.
PeerDAS is a prototype for Data Availability Sampling (DAS), which is essential for Ethereum’s scaling via sharding. It allows light clients to check whether all shard data has been published by extracting small portions, greatly improving scalability while maintaining decentralization and security.
The other leg, zkEVMs, is now “production grade performance,” Buterin said, meaning the remaining work is security and proving robustness at scale.
Buterin described this as a practical step toward solving the so-called “blockchain trilemma,” not as a theory, but through “live running code,” adding that zkEVM nodes could start appearing in limited form by 2026.
A long-term goal is “distributed block building,” Buterin added, where no single party collects the entire block in one place, reducing censorship risks and improving geographic fairness.
The message is that Ethereum’s scaling plan is increasingly about splitting the verification work across the network, rather than asking each node to replicate everything.
