Vitalik Buterin calls for hard cost ceiling on Ethereum block proofs

Ethereum developers are wrestling with how to guarantee that zero-knowledge proofs of blocks will complete within predictable time and cost limits as the network scales. This matters to operators and institutions evaluating the infrastructure burden of Ethereum’s path to higher throughput.

• Vitalik Buterin called for a hard ceiling on proving costs per unit of block gas in an October 15 post.
• In its October 13 announcement, ZKsync reported that one NVIDIA RTX PRO 6000 Blackwell GPU proved 99% of 7,200 consecutive Ethereum mainnet blocks in under 10 seconds each, including witness preparation.
• EIP-7904, a key compute-cost analysis, faces a November 1 deadline review but explicitly does not address ZK-SNARK proof generation workloads.

• 99% of consecutive mainnet blocks proved under 10 seconds per block
• 100M gas per second, the target throughput for EIP-7904’s bare CPU analysis
• Nov 1 deadline for EIP-7904 review as an Informational proposal in Last Call
• 12 seconds maximum for all sampled blocks in ZKsync’s GPU evaluation

Ethereum’s push toward faster zero-knowledge proofs has collided with a fundamental question: how do you guarantee that the hardest permitted block will still be provable in reasonable time and cost? CryptoSlate reported that Vitalik Buterin framed the problem in October 15, calling for a hard resource ceiling tied to block gas, the unit Ethereum uses to meter computational work. The issue is not average performance but worst-case burden: a block deliberately assembled to be expensive to prove could demand far more resources than a typical block, even if ordinary blocks finish quickly. For independent prover operators budgeting hardware, that worst-case requirement determines how many GPUs or specialized circuits they must maintain on standby.

ZKsync demonstrates single-GPU proving at scale

ZKsync’s Airbender team published results on October 13 showing that one NVIDIA RTX PRO 6000 Blackwell GPU proved 99% of 7,200 consecutive Ethereum mainnet blocks in under 10 seconds each, including witness preparation, with every sampled block finishing in under 12 seconds. The company also published separate configurations using two and four GPUs. This performance marks tangible progress on the proving infrastructure side, but it does not automatically establish a reliable bound for every workload an Ethereum block validator might encounter.

Proof generation and CPU execution impose fundamentally different constraints on hardware and algorithms.

Gas limits do not yet account for proving work

Ethereum’s gas model accounts for execution cost but does not currently tie proving cost to any explicit limit. EIP-7904, a compute-cost analysis now in Last Call before a November 1 review deadline, targets 100 million gas per second on optimized native execution clients but explicitly does not address ZK-SNARK proof generation. The document recommends no changes to Ethereum’s gas schedule because its analysis focuses on CPU execution measured on bare hardware, not the circuit costs of proof systems.

Nethermind Research’s Lin Oshitani found in January 27 measurements that not all gas units impose equal proving burden. Cryptographic precompiles and several modular arithmetic or division operations were relatively expensive to prove per gas unit, and costs varied by prover type. The same operation could require different elapsed time depending on whether it ran on SP1, RISC0, or another proving system. This asymmetry means a simple gas ceiling on execution does not guarantee a predictable proving deadline.

A stronger bound requires a specified performance envelope that accounts for fixed overhead, nonlinear aggregation costs, and hardware configuration assumptions.

Next step: workload-specific proving limits

The Ethereum Foundation outlined a worst-case benchmarking framework in January that ties required proving throughput to a chosen block gas limit and proving deadline. Under that approach, a deliberately expensive block must still fit the intended hardware and time budget. The framework explains why operation-by-operation cost estimates need end-to-end validation, since proof generation has fixed overhead and aggregation can be nonlinear.

For independent operators, the consequential next step is a bound tied to explicit workloads, implementations, security settings and hardware assumptions. That would turn an abstract gas allowance into a dependable limit on the proving resources an operator must supply. Alternative approaches include weighted ZK-gas metering with a per-block ceiling, which would count proof-heavy operations against a separate allowance and potentially constrain blocks before their ordinary execution-gas limit is exhausted. The choice of how to price proving work separately from execution, or whether to reprice expensive opcodes within the existing gas model, remains open.

The BlockWest read. Operators and infrastructure providers building for Ethereum’s proving layer need a guarantee, not an aspiration. ZKsync’s results show the technology works in practice, but proving costs will become a real constraint when blocks scale to the network’s limits. A hard ceiling tied to measurable workloads would shift the burden from guesswork to engineering: either improving circuits to reduce cost, or capping which operations can enter a block.

Buterin’s call for a cost ceiling remains a framework, not yet encoded in a formal EIP or the protocol. His October 15 post set three priorities for ZK-EVM: matching mathematical specifications to Ethereum’s execution, bounding prover cost relative to block gas, and securing the proof system itself. The burden now falls on protocol developers to translate that second priority into measurable performance guarantees before Ethereum’s proving infrastructure becomes the throughput bottleneck.

BlockWest is a news publication. Nothing here is investment advice. Read our disclaimer and editorial policy.