Inside Zcash’s new node that targets Visa-scale privacy at 50,000 transactions per second

A full node is the program that keeps a complete copy of a blockchain, the Zcash ledger, in this case, and independently checks every transaction against the network’s rules. Zakura is a fork of Zebra, the Zcash Foundation’s node software – meaning it started from the Foundation’s official code and was rebuilt from there.

Consensus rules are the shared rulebook every node enforces, the thing that decides which blocks and transactions the whole network accepts as valid. If a node applies different rules, it forks off and stops following the same chain as everyone else.

Zakura can also prune, a term for deleting old blockchain data a node no longer needs, and cut disk usage substantially. That shrinks the chain enough that the team publishes ready-made copies of it, about 11 gigabytes with the old data stripped, which a new node can download instead of pulling the whole history from other nodes one block at a time.

That takes a node from nothing to running in under two minutes, which the team says is “680 times faster.”

A compatibility mode further reproduces the interface of zcashd, the original client that reaches end of life on July 18, so wallets and exchange integrations built against it will keep working as is.

The reason for building all this is arithmetic.

Mastercard and Visa process more than 50,000 transactions per second, and the team calls that figure ‘“its floor, not its target.” Zcash’s current cryptography would require a node to take in and verify more than 500 megabytes of data every second to keep up, because every private transaction carries a proof, and proofs are large.

That is roughly a full DVD of data arriving every ten seconds, continuously, and no current Zcash software runs anywhere near that. But the missing piece is the reason each bottleneck exists.

Bowe’s Project Tachyon is tackling this by working on recursive proofs, in which one proof attests to the validity of thousands of others, dramatically reducing the amount of data that must be checked at consensus.

Under Tachyon, a node verifies a single proof instead of the thousands, which the team says reduces the requirement for consensus data from 100 megabytes per second to 500 megabytes, a level they claim is technically achievable with careful engineering.

Wallets have a different problem. Because Zcash hides who a transaction is for, a wallet cannot ask a server which transactions belong to it without giving itself away. It pulls down everything and tests each one, which is why wallet software tops out at about one transaction per second.

To remove that bottleneck, Valar Group is working on private information retrieval techniques that let a wallet fetch its own data from a server without the server learning which entries were requested.

Fast block propagation means broadcasting newly mined blocks across a blockchain network as quickly as possible. Zakura is a software layer tasked with that.

It has to move new blocks between nodes fast enough for high‑volume proofs and wallet traffic to matter. It ships with an experimental system aimed at delivering every block to every node in under half a second, which is switched off by default for now.

The near‑term test of these ideas arrives in late July. Ironwood, formally NU6.3, activates on mainnet at block 3,428,143, roughly 8 a.m. Eastern on July 28, and Zakura supports it from release.

Bowe said on July 10 that all major organizations are committed to that height, a week later than originally planned, after exchanges and wallet providers requested preparation time.

Ironwood exists because of a flaw that nearly broke Zcash in June. The so-called shielded pools are the private side of the network, where amounts and participants are hidden, and a zero-knowledge proof stands in as evidence of the math work.

On May 29, Shielded Labs researcher Taylor Hornby found that the proof circuit for Orchard, the newest shielded pool, contained a soundness bug that let an attacker mint counterfeit ZEC with no onchain trace. The flaw had been live since Orchard activated in May 2022.

Developers disabled Orchard through an emergency response completed June 2, then restored it with a corrected circuit via the NU6.2 hard fork at block 3,364,600 on June 3.

The patch could not account for the four years the hole was open. A zero-knowledge proof reveals nothing beyond the fact that it verified, so the chain holds no record of what any Orchard transaction moved, and nobody can prove counterfeit ZEC was never created.

Ironwood is built to settle that. A so-called ‘turnstile’ at the pool’s boundary caps what can leave and what can enter, leveraging the fact that ZEC amounts crossing into or out of shielded pools are public even when the transactions inside are not. Sealing Orchard to new deposits leaves the turnstile as the only exit, and any fake coins inside are stuck there.

In simple terms, honest balances can migrate out over time, while counterfeit coins may be prevented from fully exiting and entering into circulated supply. This setting traps any attempted excess supply at the boundary, restoring reliability of the token’s supply.