A node is a computer running the software for a particular blockchain, participating in the network by receiving transactions and blocks from peers, verifying that everything follows the protocol rules, and maintaining its own local copy of the chain’s state. The basic unit of a decentralised network is the node. When people say a blockchain is “decentralised”, what they specifically mean is that the network consists of many independent nodes, each running the rules independently, such that no single party can unilaterally change what the network considers valid.
Not all nodes do the same things. A full node downloads and validates every block and transaction, enforcing the consensus rules completely. A light client or SPV client downloads only block headers and verifies specific transactions on demand, trusting full nodes for most of the data. A mining node or validator node is a full node that also participates in producing new blocks — these are a subset of full nodes, not a separate category. An archive node is a full node that keeps the entire historical state, not just the current one, and is needed for block explorers and historical queries. Most ordinary users who run nodes at all run full nodes; archive nodes are almost exclusively run by infrastructure providers because of the disk space required.
Why Running a Node Matters
The pitch for running your own node is that you do not have to trust anyone else’s view of the chain. Your node verifies every block against the rules, so you know with mathematical certainty that the UTXOs you see, the balances you track, and the transactions you broadcast are consistent with what the network considers valid. If the rules change and you disagree with the change, your node will simply reject the new blocks, and you will be on your own fork. This is the core of Bitcoin’s “don’t trust, verify” ethos, and it is the reason the Bitcoin community places so much weight on keeping the node software light enough that individuals can still run full nodes on consumer hardware.
For Ethereum, the story is more nuanced. An Ethereum full node requires substantially more resources than a Bitcoin full node — around 1 TB of disk for a full node, more for an archive node, plus meaningful bandwidth. Running one is still within reach of a consumer, but it is more of a project than plugging in a Raspberry Pi. The Ethereum community has been working on reducing the state bloat through approaches like Verkle trees and statelessness, which would eventually let nodes verify blocks without storing the full state, but those changes are still in progress as of 2026.
The “How Many Nodes Are There” Question
Bitcoin has several thousand reachable nodes at any given time — the exact count depends on the methodology, with bitnodes.io typically showing somewhere around 15,000-20,000 and Coin Dance’s figures being similar. The true number including unreachable nodes (those behind NAT or running privately) is probably several times that. The number has been roughly stable for several years, fluctuating with price cycles but not trending sharply in either direction.
Ethereum has a few thousand reachable full nodes and a much larger number of validators (around a million as of 2026, though many are operated by the same entities). The number of independent entities actually running the node infrastructure is smaller than the raw counts suggest, and a substantial fraction of Ethereum RPC traffic goes through a small number of providers (Infura, Alchemy, QuickNode) rather than through independently-run nodes. This is one of the legitimate critiques of Ethereum’s decentralisation — if most applications rely on a handful of RPC providers, the practical decentralisation is smaller than the node count implies, even if the protocol itself is decentralisation-friendly.
Running One at Home
If you want to run a Bitcoin node, the easiest path is a Raspberry Pi 5 with a 2 TB SSD running Umbrel, Start9, or RaspiBlitz. The initial sync takes a day or two depending on your internet speed, and after that the node uses trivial resources. You can point your wallet at your own node instead of relying on a third-party server, which is the main practical benefit — and it has the side effect of contributing a small amount of network capacity to the Bitcoin network, which is a public good.
For Ethereum, the parallel project is running a local client like Geth, Nethermind, or Erigon paired with a consensus client like Lighthouse or Prysm. The storage and bandwidth are higher, the operational overhead is greater, and the practical benefit for most users is smaller because most Ethereum activity happens through contracts that are hard to interact with directly through a node without substantial tooling. Most Ethereum users who care about running nodes do so via staking rather than pure validation, because staking both runs a node and earns yield for the trouble.