Here we report on the progress of the leading builders in the Proving Service ZKonomy, documenting recent significant releases, technical breakthroughs and general updates 👇

ZkCloud @thezkcloud has received a grant from the @ethereumfndn to continuously generate ZKPs for @ethereum mainnet blocks throughout Q2 2025. All proofs are publicly posted via @eth_proofs. The goal is to demonstrate scalable infrastructure for real-time proving. ZkCloud uses a 24-GPU cluster and modified open-source tools, with plans to adopt the fastest zkVMs available and reduce latency for sub-12-second block proofs: In addition, the project has added four new validators to its Testnet: @kintsugi_tech, @DeSpreadTeam, @CalBlockchain, and @AhnLab_ABC. These groups bring diverse strengths - from localized infrastructure and cybersecurity to research and validator operations. Their participation strengthens decentralized proving as ZkCloud moves closer to Mainnet and ramps up real-time proof capabilities within its network: Moreover, ZkCloud was one of the major sponsors of House of ZK's Ethproofs Summit during @BerBlockWeek. Head of Product @NorbertVadas participated in three sessions at the event: 1) First, he gave a keynote where he talked about the architectural and technical bottlenecks of scaling ZK proving - from optimizing RPC data fetching to handling unexpected variance in block complexity due to EVM opcodes with disproportionate proving overhead: 2) Later, Norbert joined @Khovr of Ethereum Foundation, @0xAlecJames of @NexusLabs, @MayaD09111862 of @StarkWareLtd, and @alicelingl of @HouseofZK to discuss how standardized zkVM benchmarks can mislead when optimized for narrow metrics, the importance of hardware-aware performance evaluation, and the need for transparent, multi-dimensional frameworks that reflect real-world workloads and build trust in verifiable compute: 3) Finally, along with @carlo_cretino50 of @0xLita, @DacEconomy of @ProjectZKM, @schmiddominik1 of @0xMiden, and Alice Liu of House of ZK, he explored the evolution of zkVM interoperability, highlighting the potential to move beyond traditional bridges, the role of decentralized proof providers, and the importance of shared standards for scalable and secure blockchain applications:

Ingonyama @Ingo_zk has released an accelerated implementation of ML-KEM, a lattice-based key encapsulation protocol selected by NIST for post-quantum cryptography. The CUDA-based code is open-source under the MIT license and achieves up to 30% higher throughput compared to current implementations. ML-KEM supports three security levels aligned with Kyber512, Kyber768, and Kyber1024: In addition, @BagadSuyash and @yuval_domb of Ingonyama, @QuangVDao of @CarnegieMellon, and @SuccinctJT of @a16zcrypto have released a paper describing the integration of two key optimizations that significantly speed up the sum-check protocol, a fundamental component of SNARKs. The techniques target small-value polynomial inputs and equality-polynomial forms, enabling 2-4× proving time reductions in the Jolt zkVM, with even greater improvements when memory is constrained. These results directly address prover bottlenecks in systems like Spartan within Jolt:

Fermah @fermah_xyz has launched Phase 2 of its testnet, which includes integrating external Prover Nodes, improving developer tools, and running custom benchmarks. The phase is named after Professor Amit Sahai, whose work significantly shaped the foundations of ZK cryptography. Fermah notes that future phases will expand participation beyond the current infrastructure network, signaling continued growth and broader community involvement: Additionally, @vanishree_rao, Founder of Fermah, during @eth_proofs Call #2 shared that Fermah has integrated with @RiscZero and @openvm_org to locally generate proofs, aiming to support all zkVMs and submit the most efficient ones to Ethproofs. She also shared a proposal for a new security definition - chosen proof extractability - highlighting gaps in current standards. [Timestamp - 02:04:50]:

Cysic @cysic_xyz introduces ComputeFi, aiming to decentralize compute power, which is critical to both AI and ZK technologies. The platform enables users to monetize hardware and gives developers affordable, transparent access to computing resources. By reducing reliance on centralized providers, Cysic says builders can innovate freely, securely, and cost-effectively - paving the way for broader digital participation and privacy-focused apps: In addition, the project has launched Phase III: Ignition of its testnet - building on earlier stages that introduced a dual-token system, GPU provers, and mobile verifiers. Ignition is a public beta offering users a chance to participate in decentralized compute and earn testnet rewards. Access is currently invite-only, with mobile support coming soon: Finally, at the recent @eth_proofs Call #2, Cysic’s co-founder @leofanxiong shared updates on their custom ASICs built for ZKPs. The chips support key ZK fields and deliver over 1.3 million Keccak hashes/sec. They use a custom network-on-chip for efficient data reordering and pair fixed-function hardware with CPUs for flexibility. Though still on FPGA, they’re outperforming typical GPU throughput:

Snarkify During the latest @eth_proofs Call, @drouyang, Co-founder & CEO of @Snarkify_ZKP, told attendees that Snarkify has built the fastest H2 GPU prover on @Scroll_ZKP mainnet and surpassed 750 MHz in RISC-V emulation speed. The team moved full trace generation to GPUs, significantly boosting throughput. Their product, enabling real-time proving for any zkVM, will launch by the end of June. [Timestamp - 01:12:30]:

Marlin @MarlinProtocol announced that developers building on the @SuiNetwork will have untethered access to its TEE compute marketplace via Nautilus. They will be able to choose between dedicated confidential VMs or serverless execution, payable in USDC. Tooling includes Docker support, key management, and attestation: Moreover, Marlin announced that its Trusted Execution Environments now support @cyclenetwork_GO’s settlement layer and bridgeless liquidity infrastructure, enhancing secure offchain data processing. This upgrade meets institutional confidentiality needs while enabling high-throughput settlement:

ZAN During the @eth_proofs Call #2, @zan_team uncovered that they aim to lower the barrier for GPU use in ZKPs. They’ve accelerated cryptographic operators and support multiple proof systems like Groth16 and Plonk. GPU optimization led to ~121× speedups over CPUs. Their focus now is real-time proving with fewer GPUs, ideally under 50. [Timestamp - 01:27:50]: