Archive node vs full node: what it is, and when you need one

An archive node is a full node that keeps every past state of the blockchain. Ethereum's own docs describe archive nodes as full nodes that verify every block from genesis and never delete any of the downloaded data.

To see why that matters, you need one word: state. State is the snapshot of who owns what at one block: every account balance and every value stored inside every smart contract.

Each new block changes the state a little. Your node checks the block, applies the changes, and now has a new snapshot. The question is what happens to the old one.

A full node throws old snapshots away after a short while. An archive node keeps all of them, back to the genesis block (the very first block).

Both node types check every block and both can send your transactions. The archive node also answers questions about the past, and pays for it with several times more disk and a longer sync.

Here is the whole difference in one table. The Ethereum numbers come from Erigon's own docs, measured in September 2026.

The quickest way to feel the difference is to ask both nodes the same old question. Below, an app asks for an address's balance at block 10,000,000 using the standard eth_getBalance call.

A full node keeps state for a short, moving window of recent blocks and deletes anything older. This is called pruning. On Ethereum the window is typically the most recent 128 blocks.

Ethereum makes a block every 12 seconds when all validators are online. So 128 blocks is about 26 minutes of history. Ask for state older than that and a pruned node has nothing to read.

Every time a block arrives, the window slides forward by one. The newest state joins, the oldest one drops out.

Sometimes, slowly. Geth's tracing docs say a full node can rebuild older states by replaying blocks from a saved point, which can take a long time for distant transactions.

That is fine once. It is not fine for an indexer that asks a million old questions a day. That workload is exactly why archive nodes exist.

On Ethereum, an archive node takes from about 2 TB to more than 20 TB, depending on the client. Erigon measured 2.25 TB on 17 September 2026. Geth's older hash-based archive can exceed 20 TB.

Most articles quote one number with no client and no date. That number is usually wrong for you. Archive data grows every day, and each client stores it in its own format.

So here is every figure we could confirm in official docs, with its client, its source and the date it describes.

Look at Ethereum in the table: from about 2 TB to over 20 TB for the same history. The difference is the client, not the chain.

Geth's older hash-based archive stores every historical state as complete Merkle Patricia tries, which is why it can pass 20 TB. Its newer path-based archive, added in v1.16.0, needs around 2 TB. Ethereum's own docs make the same point: most clients need more than 12 TB, while Erigon stores the same data in under 3 TB.

Reth shows the effect of storage format too. Its storage page lists an Ethereum archive at 2.99 TB in its old format and 2.31 TB in its new one, measured at block 24,396,823.

If a node says state is missing, pruned or unavailable for an old block, you asked a full node about the past. Querying a newer block will work; for old blocks you need an archive node.

These are the exact messages from the client source code and docs. Your provider may wrap them in its own text, so search for the key words.

You need an archive node when your code reads state at old blocks, regularly. If you only read what is true now and send transactions, a full node does the job for less than half the price.

Most people need a full node. Here is how to tell which side you are on.

One grey zone: you need a few old values, once in a while. Then a public block explorer can show you past transactions without any node of your own.

Answer three questions in order. If any answer sends you to "full node", stop there.

Not sure yet? Start with a full node. On RocketRPC you can switch to archive later from your dashboard and pay only the difference for the rest of your term.

A dedicated archive node on RocketRPC costs $289 a month for Ethereum, Base, Arbitrum, Avalanche and Optimism, and $869 a month for Polygon. A full node on the same chains starts at $139.

Archive costs more for one reason: disk. Chapter 4 showed archives are several times the size of a full node, and they need fast drives. Reth's docs, for example, recommend TLC NVMe for both.

Polygon is the exception on price because its data is far larger than the others.

You can run an archive node on your own server. You then pay for the machine and its disk, and you spend the time: the first sync, every client upgrade, and watching that it stays in step with the chain.

If you want the endpoint without the upkeep, every price is on our dedicated node pricing page, and every chain we run is on the dedicated nodes page.