Ethereum’s current state storage system, the Merkle Patricia Trie, was never designed to scale to the size the network has reached. As of mid-2026, Ethereum’s state - all accounts, balances, contract code, and storage slots - has grown to hundreds of gigabytes. Full nodes must carry this entire structure to participate honestly in the network. Verkle trees are the proposed replacement, and understanding why requires looking at what Merkle proofs actually cost.
The Witness Problem
When a node wants to prove that a particular piece of state is valid - say, the balance of a specific address - it provides a witness: a chain of hashes from the leaf node up to the root. In a Merkle Patricia Trie, that witness can be several kilobytes per account. Multiply that by every account touched in a block, and you get witness sizes that make stateless clients impractical. Stateless clients are central to Ethereum’s roadmap because they would allow nodes to verify blocks without storing the full state, dramatically lowering hardware requirements.
Verkle trees solve this with a different cryptographic primitive. Instead of hash-based commitments at each node, they use polynomial commitments - specifically, a scheme called IPA (Inner Product Argument). The key property: a single commitment can cover a much wider branching factor (256 children per node, versus the current 16) while still producing witnesses that are far smaller. A Verkle witness for a block touching hundreds of accounts can fit in tens of kilobytes rather than megabytes.
What Actually Changes
The branching factor shift sounds like an implementation detail, but it flattens the tree considerably. Shallower trees mean fewer database reads per lookup, which matters for validator performance at scale. More importantly, the small witness size makes the stateless client model viable - nodes could verify new blocks by downloading only the witness data for that block, not the full state.
This is not a soft fork. Migrating Ethereum’s state from a Merkle Patricia Trie to a Verkle tree requires a conversion process during a hard fork, touching every account in the state simultaneously. The EIP process for this - EIP-6800 and related proposals - has been in development for several years.
Still Pending
Verkle trees have not shipped on mainnet as of this writing. The Ethereum development community has treated the migration as post-Pectra work, and the cryptographic libraries required for IPA commitments are still being optimized for production use.
The underlying argument for Verkle trees is straightforward: Ethereum cannot achieve meaningful decentralization if running a full node requires enterprise-grade storage hardware. Smaller proofs are the mechanism, not the goal.