Proving services Latest
Share on

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

Fermah

@fermah_xyz has introduced 'Confidential Proving Delegation' - a system allowing ZKP tasks to be securely offloaded to powerful external machines without exposing sensitive data: https://fermah.xyz/blog-posts/confidential-proving-delegation…

By using multiparty computation and trusted hardware, CPD preserves privacy while offering performance on par with centralized systems. This eliminates the long-standing trade-off between privacy and user experience, unlocking practical, scalable applications in identity, DeFi, and beyond.

The introduction of CPD led to multiple new partnerships aiming to utilize the system, including:

@reclaimprotocol: Enabling private zkTLS verification by integrating CPD into Reclaim’s Web2-to-Web3 credentialing flow. The partnership enhances privacy-preserving applications like @3janexyz - a credit-based money market on @base - by significantly reducing proof generation times and offloading compute to Fermah’s infrastructure. As a result, end-users experience near-instant verifiability and seamless UX while retaining full data privacy: https://fermah.xyz/blog-posts/fermah-x-reclaim… 

@Labyrinth_HQ: Powering compliant privacy infrastructure with CPD across Labyrinth’s multi-chain privacy layer. Labyrinth provides modular tooling for private, regulation-friendly transactions using zero knowledge and threshold cryptography. By delegating proof generation to Fermah, it reduces latency, lowers compute costs, and enables privacy-preserving flows that are fast enough for real-world DeFi, DAO governance, and social applications: https://fermah.xyz/blog-posts/fermah-x-labyrinth… 

Finally, @anthonykrose, CTO of @the_matter_labs, and @vanishree_rao, Founder of @fermah_xyz, featured in a House of ZK special interview, explaining how ZKsync’s Elastic Network and Fermah’s universal proof market combining to reshape decentralized proving. 

They discussed its impact on scalability, resilience, and the long-term evolution of ZK systems across multi-chain ecosystems: https://x.com/HouseofZK/status/1922720308072489069

Ingonyama

@OmerShlomovits, CEO of @Ingo_zk, shared that ICICLE 3.9 introduces decision trees to predict optimal MSM parameters for each curve on AMD, Intel, and ARM CPUs. This significantly boosts performance compared to the heuristic method previously used in Arkworks: https://x.com/OmerShlomovits/status/1922962049149104503… 

He also announced that ICICLE v4 will introduce lattice-based cryptography, starting with lattice-based zero-knowledge systems like LaBRADOR, Greyhound, and the LatticeFold family. Future plans include expanding to post-quantum cryptography, private information retrieval, and fully homomorphic encryption: https://x.com/OmerShlomovits/status/1920810288716079185… 

Additionally, Ingonyama’s ICICLE cryptography library has been integrated into @zerobasezk's infrastructure to accelerate real-time ZKP generation. This enabled faster, more scalable performance across apps like @Solanasim, zkLogin, and Zerobase’s staking network. ICICLE’s GPU acceleration reduced latency by up to 40%. Zerobase plans deeper ICICLE integration to support larger circuits, decentralized ZK workflows, and future-proof cryptographic systems: https://medium.com/@ingonyama/zerobase-x-icicle-accelerating-real-time-zk-applications-at-scale-b0b9451b4172… 

They also partnered with @Cornami_Inc to accelerate ZKP cryptography by combining Ingonyama’s ICICLE software stack with Cornami’s FracTLcore® hardware. This integration delivers 10× to 100× improvements in speed and energy efficiency over GPU-based systems. Their joint solution, the ZPU™, addresses rising demands in privacy-focused technologies and post-quantum cryptography, with early validation by partners like @brevis_zk: https://medium.com/@ingonyama/ingonyama-and-cornami-announce-strategic-collaboration-to-accelerate-high-speed-cryptography-af2ddb22d765

Cysic

@cysic_xyz and researchers from @Tsinghua_Uni have unveiled ZKPoG (https://eprint.iacr.org/2025/765), a GPU-based platform that accelerates the entire ZKP process, not just select steps. It enables real-time proving on consumer-grade GPUs like the RTX 4090, showing up to 52x CPU speedups. ZKPoG supports Plonky2-style protocols, custom gates, and reduces GPU memory use by over 60%, making advanced ZKPs accessible and efficient for developers using standard hardware: https://x.com/cysic_xyz/status/1917772291083231530…  

They also presented Hypercube IR, a new intermediate representation designed to improve ZKP systems. It simplifies ZK logic into efficient, portable operations optimized for various hardware. It achieved 1.31 million Keccak proofs per second on their ZK-ASIC. Unlike generic IRs, it’s tailored for predictable ZK workloads, enhancing speed, scalability, and hardware-software integration - potentially beyond just ZK apps: https://hackmd.io/@Cysic/ByY3_L5hkl… 

Finally, the project announced a partnership with @NexusLabs to expand the Nexus network’s compute capacity. Known for its real-time ZKP technology, Cysic will enhance performance, speed, and efficiency. 

The collaboration supports Nexus’s goal of building a distributed, verifiable supercomputer by improving scalability and access to trusted computation. It also opens opportunities for joint work on hardware improvements for secure, transparent systems in AI, blockchain, and privacy: https://blog.nexus.xyz/cysic-partnership/

ZkCloud 

@thezkcloud is retiring its Firestarter network to make way for a new version of its zero-knowledge compute infrastructure. Firestarter, a Cosmos SDK-based chain supporting various ZK provers, will be fully open-sourced. Proof generation for platforms like Ethereum continues as ZkCloud prepares to launch its mainnet and integrate Firestarter’s successor into the Deluge testnet for broader deployment: https://blog.zkcloud.com/p/sunsetting-zkcloud-firestarter… 

Also, the project has integrated @RiscZero’s latest zkVM, R0VM 2.0, into its Firestarter proving network. Developers can now generate verifiable proofs using dual RTX 4090 GPU nodes, with clear, pay-per-proof pricing. A full template with setup instructions is available on GitHub. Benchmarking for performance and cost is ongoing, with results to be shared transparently in future updates: 

https://x.com/thezkcloud/status/1918206408471830832…

Finally, ZkCloud is actively supporting @aztecnetwork’s privacy-first Ethereum L2 network, participating in its public testnet as both a sequencer and prover. 

With no centralized components, Aztec enables encrypted, decentralized execution. ZkCloud contributed to earlier initiatives like ProverNet 1.0 and the S&P Testnet, emphasizing a shared commitment to decentralized infrastructure and transparent governance in blockchain systems.

https://x.com/thezkcloud/status/1930253378715603276

Marlin

@MarlinProtocol has expanded its Serverless platform with delegated staking, allowing broader participation and reinforcing network reliability. 

Executors run off-chain jobs and are selected based on stake weight. Delegators contribute through a flexible Stash system that supports asset management and redelegation. Rewards are tied to job execution and periodic emissions, distributed by performance. The design aligns incentives, encouraging dependable execution and informed delegation to strengthen the network’s integrity: https://x.com/MarlinProtocol/status/1925903830689984946… 

The project also announced new partnerships, including:

@inference_labs: Enabling trustless AI inference by combining TEEs and ZKPs - keeping model inputs private while generating cryptographic proofs of correctness:  https://x.com/MarlinProtocol/status/1920539454147600610… 

@DeepLinkGlobal: Powering decentralized gaming environments with TEE-secured compute - from AI rendering to persistent game logic, maintaining data confidentiality and verifiable execution.: https://x.com/MarlinProtocol/status/1922629832082354643

Snarkify

@Snarkify_ZKP has introduced SRT, a modular, GPU-native proving engine that enables real-time proof generation for any zkVM without requiring architectural changes: https://blog.snarkify.io 

SRT improves proving speed by 10-100x and integrates directly into existing stacks. Launching in June with a priority partner, SRT aims to prove each @ethereum block in under 12 seconds, easing zkVM bottlenecks and supporting scalable, verifiable applications across the Ethereum ecosystem.

Zerobase

In the run-up to its Mainnet launch, @zerobasezk has integrated @Ingo_zk’s ICICLE cryptographic library into its ZK circuits, improving performance and scalability.

ICICLE optimizes key processes like Number Theoretic Transforms and Multi-Scalar Multiplications, while supporting multiple platforms including GPUs and Apple Silicon. 

The collaboration enhances proof generation speed without compromising security, with future plans to expand GPU optimization and cryptographic capabilities: https://x.com/zerobasezk/status/1928334051821101087

ZAN

@zan_team has completed a full security audit of @BlockRock_rwa, a platform that tokenizes real-world financial assets for use in the Web3 ecosystem. 

The audit focused on smart contract integrity, helping to ensure user asset safety. BlockRock’s partnership with ZAN highlights its commitment to security in bringing traditional finance onto the blockchain: https://x.com/zan_team/status/1927180672516411518… 

They also announced a collaboration with the @solana ecosystem to support growth in the Asia-Pacific region. The partnership focuses on RPC services, validator operations, and node hosting. 

ZAN also aims to assist developers by offering tools and resources, contributing to the ongoing expansion of the Solana community and its technical infrastructure: https://x.com/zan_team/status/1922524828738978113

More articles
Interview
From Digital Gold to Digital Cash: Eli Ben-Sasson on Scaling Bitcoin with STARKs
Read More
May 1, 2025
Reports
Highlights from the Cryptology ePrint Archive
Read More
March 10, 2025