Technician using a laptop beside server racks

Aztec Cut Private Ethereum Proving to 2.5 Seconds. Its Own Warning Still Matters

July 21, 2026 10:00 am Comments

Privacy on Ethereum has always carried a hidden invoice.

You can conceal a transaction, but the proof may take too long. You can make the proof faster, but only by handing private data to a server.

You can push the work onto a user’s device, but then the experience starts to feel like cryptography homework.

Aztec says its newest network upgrade has cut through that tradeoff.

Alpha V5 went live on mainnet July 21 with a fully private token transfer taking approximately 2.5 seconds to prove natively on a consumer laptop. The same flow took 5.2 seconds under V4.

That is the number built to travel. The more important detail is where the work happens: on the user’s own machine, without sending the private inputs to an operator that can read them.

Aztec is an Ethereum Layer 2 designed to let one smart contract combine private and public logic. Private functions run inside a private execution environment on the user’s device, which produces a zero-knowledge proof that the action followed the rules without revealing the underlying data.

The network receives the proof, not the user’s financial life.

That distinction matters. A privacy system that decrypts a transaction on somebody else’s server may hide activity from the public while creating a privileged observer behind the curtain.

Aztec’s pitch is that the app, sequencer and network operator never need to see the private inputs in the first place.

Aztec measured the 2.5-second result on an M2 MacBook with 12 cores throttled to eight, using its native C++ proving binary. Browser proving fell from 12.5 seconds under V4 to about 6.8 seconds under V5.

Those are controlled benchmarks on specific hardware. Results will vary across older laptops and phones, while the roughly twofold improvement in both native and browser tests still attacks one of private crypto’s most stubborn product problems.

A proof that finishes while a user is still focused on the transaction feels like part of checkout. One that leaves the screen spinning becomes a reason to abandon the app.

V5 also lowers the average cost of a fully private token transfer to under five cents, according to Aztec. The network halved its proving-cost parameter and reduced the Ethereum gas needed to verify a rollup proof by approximately 40%.

Because that Ethereum verification cost is shared across a batch, the saving can reach every transaction inside it. The five-cent figure is still an average under current assumptions, not a fixed price immune to network conditions.

Block production has dropped from roughly 14 seconds to six seconds. Behind that improvement is a pipelined design in which the proposer for a slot builds and gathers attestations during the prior slot, reducing the dead period that previously left fresh transactions waiting.

The upgrade also begins proving checkpoint work during an epoch rather than waiting for the entire epoch to finish.

Aztec documented a wider set of V5 changes before the governance proposal executed, including active penalties for validator equivocation, a smaller protocol-contract surface and safeguards intended to stop governance actions from rendering the rollup inoperable.

This is more than a laboratory release. The first applications are arriving with it.

Nyx lets users connect an Ethereum wallet, send and receive privately, and access yield on Aave through a private account. Aztec also lists the Azguard and Nethermind wallets, Shield and TRAIN bridges, RavenHouse collectibles, block explorers and a private-state version of the strategy game Dark Forest among the early V5 ecosystem.

The apps turn a benchmark into a testable proposition. Private transfers are interesting; private balances, lending positions, games and business logic that can interact inside one shared system are the larger bet.

But the word attached to V5 is not decorative.

This is still Alpha, and Aztec’s own security history explains why that label deserves more attention than the performance chart.

In March, the team disclosed that it had found a critical vulnerability in Alpha V4. The flaw affected the proving system as a whole and was not neutralized by the validator committee’s public re-execution safeguard.

Aztec warned that exploitation could severely disrupt the protocol and lead to theft of user funds. It deliberately withheld the bug and patch details until V5 so the fix could be distributed with the new network release.

The severity came from the part of the system the flaw touched. Validator re-execution could provide a training wheel for some public execution bugs, but Aztec said that protection did not neutralize this proving-system vulnerability.

The same disclosure identified another flaw in the Barretenberg proving library that could let incorrect proofs enter the Aztec mempool. Node operators were told to move to version 4.1.2 or later, and Aztec separately credited outside researchers for another issue that was considered critical for Noir.

V5 resolves the critical V4 issues, Aztec says, along with additional bugs found after launch. More than 234 researchers participated in its Cantina bug-bounty program.

That is meaningful work. It is not the same thing as proof that no critical vulnerability remains.

Bug bounties, audits and a larger researcher pool improve the odds of finding dangerous code before an attacker does. They cannot turn novel proving software into mature financial infrastructure by announcement.

Aztec says that plainly in its standing Alpha guidance.

Aztec describes the software as experimental, says confidence is still developing and advises users not to deposit more than they are willing to lose. The team considers Alpha suitable for early use cases and small amounts of value while audits and bounties continue.

Each Alpha deployment is also distinct, with state not migrated between Alpha releases. Builders can deploy real apps on a network that expects more major versions and treats continuity as a later-stage property.

Aztec’s own bar for Beta shows the distance still to travel: more than 10 transactions per second, shorter block times, 99.9% uptime and three months without a critical bug disclosed through the bounty program.

Alpha V5 currently targets one transaction per second and roughly six-second blocks. It may be ready for serious experimentation without being ready for serious savings.

The risk does not stop at the core protocol. A user entering through a bridge, wallet or DeFi app is also trusting that product’s contracts, interface and operational controls.

A private transaction can still leak information through wallet behavior, timing, deposits, withdrawals or a poorly designed application. Cryptographic privacy is a system property, not a magic word that blesses every action around it.

The Block notes that the upgrade followed an onchain governance process and frames V5 as the point where Aztec’s private execution environment reaches production applications. That makes the release consequential even if its safest users remain developers, testers and early adopters.

The strategic opportunity is real.

Transparent ledgers force traders to expose positions, companies to reveal payment relationships and individuals to publish a permanent map of their finances. Institutions cannot move every sensitive financial operation onto a chain where competitors can watch it unfold in real time.

A programmable privacy layer could let an application prove that a rule was followed without broadcasting every input. Payroll, credit, trading, governance and identity become more plausible onchain when confidentiality and verification can coexist.

For that vision to matter, privacy cannot feel like a specialist feature with a specialist wait time and a specialist fee. Aztec’s 2.5-second result brings it much closer to ordinary software.

Now comes the harder benchmark.

V5 must run long enough to earn confidence, keep its app ecosystem usable and survive the attention that arrives when real money discovers cheap private transactions. Its predecessor proved that a network can improve quickly because researchers find serious flaws.

It also proved that serious flaws can exist before they are found.

Aztec has made private Ethereum faster, cheaper and tangible enough to try. The Alpha label tells users exactly how much trust that achievement has not yet earned.

Join the conversation!

We have no tolerance for comments containing violence, racism, profanity, vulgarity, doxing, or discourteous behavior. If a comment is spam, instead of replying to it please click the icon below and to the right of that comment. Thank you for partnering with us to maintain fruitful conversation.