Ethereum Co-founder Vitalik Buterin: Node Synchronization Takes Half a Day
Ethereum co-founder Vitalik Buterin stated that synchronizing an Ethereum node currently takes about half a day at the fastest, with aggressive configurations allowing disk usage to be below 0.5 TB, primarily due to EIP-4444 and ongoing optimizations for snapshot synchronization across various clients.
He mentioned that the upcoming Glamsterdam upgrade will continue to improve synchronization, and Nimbus's new synchronization protocol has adopted relevant mechanisms. Users can point their wallet's remote procedure calls to local nodes, but must correctly enable the node's RPC; some decentralized applications in browsers cannot effectively utilize local RPC, while others hard-code RPC to their own servers. He expressed a preference for avoiding browser applications in favor of command line usage, and has updated ENS records using a Python script connected to the local node. The next step for the Kohaku command line tool will be to integrate privacy protocols, with early development already underway.
EIP-4444 aims to relieve the execution layer from permanently storing all history, separating historical retrieval from "each full node maintaining a complete disk," while snapshot synchronization allows new nodes to pull state first and then catch up, rather than replaying from scratch. Nimbus has long pursued lightweight and beacon-side synchronization, and connecting the new protocol to Glamsterdam effectively turns client differences into a single item on the upgrade list. Kohaku is being advanced by the Ethereum Foundation's access layer initiative, aiming to make security, trustlessness, and read-write privacy the default, including assigning independent addresses for each new application and encapsulating privacy protocols like Railgun for wallets and agents to call.
Home-grade hard drives and the half-day window lower the barrier for building nodes, but browser wallets and hosted RPCs remain the default entry points. Infura, Alchemy, self-built RPCs, and wallet vendors sell usability; local nodes sell verification and privacy. Funds are flowing back from pay-per-request centralized RPCs to operators willing to allocate disk space and bandwidth, provided that the application layer no longer locks remote addresses. Beneficiaries are client teams and command line users who can point to 127.0.0.1; those under pressure are applications that hard-code RPC and rely on browser injection.
Source: Public Information
ABAB AI Insight
Vitalik has discussed transforming nodes from "archive managers" to "current state validators" for many years: statelessness, historical expiration, portal networks, lightweight clients. EIP-4444 offloads historical responsibility from default full nodes, and snapshot synchronization compresses initial startup from weeks to hours. His use of Python to modify ENS demonstrates that the access layer should reside in local processes rather than browser plugins plus third-party gateways.
The capital path lies at the infrastructure layer: client teams invest in synchronization protocols, the foundation develops Kohaku into an access layer SDK, and privacy protocols shift from independent frontends to command line defaults. The motivation is that after L2 takes away execution and user interfaces, if the mainnet only has expensive full nodes and hosted RPCs, verification rights will concentrate in a few endpoints. Half a day and 0.5 TB bring verification costs back to consumer hardware, making "self-built nodes" an option again rather than a mining farm specification.
In contrast to Bitcoin's UTxO pruning, Solana's continuous hardware demands for large ledgers, and the Cosmos ecosystem's general use of snapshot startups: Ethereum is pursuing protocol-level expiration and client snapshots, rather than defaulting trust to a specific snapshot station. The industry phase is that execution has been modularized while the access layer remains centralized; Glamsterdam and Kohaku aim to pull the second segment back.
Structural changes represent a transfer of pricing power. Whoever controls RPC sees user addresses, balances, and transaction forecasts. Local nodes reclaim this pricing power from custodians for validators; if browser applications still hard-code servers, the transfer only occurs among a minority of command line users. Historical storage pricing power shifts from full node disks to dedicated historical providers. Synchronization becoming cheaper does not mean that default RPCs have moved.
ABAB News · Cognitive Laws
Synchronization becomes cheaper, hard-coded RPCs render it ineffective
Not everyone needs to store history, but everyone must verify state
If privacy exists only in browsers, local nodes are merely hard drives