With the current block gas limit the chaindata grows 70 GB per year. The growth rate is limited by the block gas limit, which itself can grow and its growth is limited by the protocol (formula 45 in the yellow paper).
Supporting 1 billion of users will require terabytes of storage and it will take years to grow to this scale.
Detailed version below
First of all the chaindata can be broken down into the following pieces:
- block headers
- account nonces (both external and contracts)
- account balances (both external and contracts)
- contract bytecode
- contract storage
According to this article about 85% is taken by the blockchain and only 15% is state https://dev.to/5chdn/the-ethereum-blockchain-size-will-not-exceed-1tb-anytime-soon-58a
This graph shows a steady linear growth of the chaindata size https://etherscan.io/chart2/chaindatasizefast. Around 6 GB per month ~= 70 GB per year.
Here is my rough upper-bound estimation for the contract storage growth rate. I base my estimation on the maximum number of storage slots that can be updated in a block given the block gas limit.
- ~8,000,000 block gas limit
- 20,000 gas per storage slot
- ~8,000,000 / 20,000 = ~400 max slots written in a block
- ~64 bytes per slot (32 for each key and value)
- ~400 * ~64 = ~25,600 bytes per block
- ~6,000 blocks per day
- ~6,000 * ~25,600 = ~154 MB per day
- ~154 * ~365 = ~56 GB per year
This assumes block gas limit is constant. In fact it can change. Looking at this graph https://etherscan.io/chart/gaslimit I think we shouldn't expect it to grow quickly (the growth rate is limited by formula 45 in the yellowpaper).
The above estimation doesn't account for overheads associated with data storage, merkling and indexing.
As for 1 billion users - we can assume most of them having 10 external accounts:
- 32 bytes for nonce
- 32 bytes for balance
- 20 bytes for address itself
- 84 bytes total
- 1,000,000,000 * 10 * 84 = 840 GB
Now if we assume 100 accounts instead of 10 then it becomes ~8 TB. If all state is stored on every node it looks problematic.
You can look at it from another angle - assume 10 transactions per second limit (this is how many transactions the network can handle on average right now). Even if every transaction were to create a new external account in the state it would take ~1157 days to create 1 billion accounts. In reality only a portion of all transactions would create new accounts and each user needs more than 1 account so it will take years to scale to 1 billion users. So the transaction throughput is likely to be the first bottleneck to be solved rather than the storage bottleneck.
The scalability problem is a very important and tough one. The Ethereum team is working on sharding, which will allow nodes to maintain only a portion of the state.
What are the Ethereum disk space needs?