Fermah x ZKsync: Proof Generation with the Universal Proof Market

Zero-knowledge proofs have unlocked an entirely new design space for scaling and privacy, and ZKsync has been at the forefront of making that a reality. Their commitment to decentralization, performance, and developer experience has made them one of the most technically advanced projects in the space. That’s why this collaboration between Fermah and ZKsync is such a big deal.

This isn’t just another technical integration—it’s a real milestone for us at Fermah, and for the ZK space more broadly. ZKsync, a leader in the industry, is setting the standard for how proving should be done. It’s a bit surreal, honestly—when we started Fermah, we had this vision of making ZK proof generation seamless, scalable, and accessible. And now, having a project like ZKsync adopt Fermah’s Universal Proof Market is a moment worth celebrating.

Check out the conversation between Anthony Rose, CTO at Matter Labs, and myself from ETH Denver, in which we discuss the integration and its significance:

Proving is Complex

Proving is a computationally intensive process—generating a single batch requires executing tens of thousands of circuits across multiple layers of recursion. This demands significant hardware resources and a high degree of orchestration to maintain efficiency.

Relying on centralized cloud infrastructure comes with inherent drawbacks: high markups inflate costs, and the absence of a competitive, open market prevents price optimization. A cost-effective proving system must not only leverage distributed compute resources but also introduce market dynamics to drive prices down through competition. 

However, achieving this is far from trivial. Proving at scale demands significant computational resources, intelligent orchestration, and a strong incentive mechanism to balance cost, speed, and security.

The key to making proof generation sustainable is building a fault-tolerant, high-performance network that efficiently manages massive computational loads at the lowest possible cost. 

Achieving this requires maximizing hardware utilization, aggregating demand from multiple sources, and ensuring that compute resources remain optimally engaged at all times. Without an optimized system in place, developers must either rely on centralized provers or build custom solutions, both of which introduce trade-offs in cost, decentralization, and scalability.

This is where Fermah steps in.

Fermah is the Endgame for ZK

Fermah leverages a distributed network of EigenLayer Operators, rather than a single source, to generate proofs. This enhances fault tolerance, censorship resistance, and ensures optimal resource utilization. Fermah’s lean mechanism design dynamically orchestrates proof generation, intelligently matching requests with the most cost-effective hardware while maximizing efficiency.

Beyond simplifying access to proof generation, Fermah optimizes the entire proving lifecycle:

• Optimized Proof Generation – High-performance bare-metal infrastructure ensures that proofs are generated efficiently while minimizing costs. By introducing market dynamics and competition, Fermah reduces pricing inefficiencies found in centralized cloud-based solutions.
• Flexible Compute Options – Fermah supports a wide range of machines with varying CPU, RAM, and GPU configurations, enabling users to balance cost and performance dynamically.
• Adaptive Scaling – Fermah aggregates demand across various sources to maximize hardware utilization, dynamically scaling resources to maintain efficiency and keep proving costs low.
• Fault Tolerance & Recovery – The system proactively handles hardware failures, network issues, and computation errors, guaranteeing uninterrupted proof generation.

ZKsync and Fermah: A Powerful Integration

By integrating with Fermah, ZKsync and the Elastic Network now generate proofs for their blocks on Fermah’s decentralized proving network. This setup significantly strengthens their proving stack by:

• Increasing Resilience – Shifting away from reliance on a single proving entity enhances security and censorship resistance.
• Lowering Costs – Efficient resource allocation and competitive pricing within Fermah’s universal proof market drive down proof generation expenses.
• Enhancing Scalability – A dynamic and decentralized proving market allows ZKsync to scale efficiently without bottlenecks.

This integration is a major step toward a fully decentralized proving ecosystem, ensuring that proof generation remains both economically viable and technically robust.

Let’s dive into how this works. 

Under the Hood: How Proof Generation on Fermah Works

To understand how Fermah optimizes proof generation, let’s first explore its core components:

• Seeker – The entity that submits proof generation requests.
• Operator – A network participant providing compute power in exchange for rewards.
• Matchmaker – The system that intelligently pairs proof requests with the most cost-effective operator.
• Workflow Engine – The backbone of Fermah, orchestrating proof computations across multiple operators.
• Storage Layer – A decentralized storage system that ensures checkpointing and fault recovery.

Proof Generation Flow

And here’s how these core components work together to generate a proof: 

1. Seeker submits a proof request – This request includes a workflow file defining the entire proof generation process.
2. Workflow Engine processes the request – It verifies inputs, configures execution parameters, and assigns resources.
3. Matchmaker selects an operator – The system scans the network to find the most cost-effective and optimal proving node.
4. Workflow execution begins – The assigned operator executes computations according to the defined workflow.
5. Storage Layer ensures reliability – If an issue arises, Fermah resumes from the last successful checkpoint rather than restarting the process.

This intelligent design ensures that proof generation remains efficient, decentralized, and scalable, while continuously optimizing for cost and performance.

The diagram above provides a simplified, high-level overview of what happens when a seeker submits a proof request. While the actual process is more involved, this should serve as a concise illustration for our discussion.

When a new ZK Stack batch is ready to be proven, the seeker picks it up and submits it to Fermah along with a workflow file. This workflow file contains the entire end-to-end proof generation process, including setup configs, input data, etc. This request is picked up by Fermah Core, which then starts running a workflow.

Think of a workflow as an isolated, self-contained execution plan that defines how the proof should be generated, step by step. It dictates which computations need to be performed; how data flows through the process; and which hardware resources are required to complete the proof efficiently.

Whenever it needs to perform a larger computation, the workflow issues a delegation request with the required hardware specifications. The Matchmaker then steps in to scan the network and find the cheapest operator nodes that meet the seeker’s requirements. Once a suitable operator is selected, the computation is offloaded to that node, ensuring efficient resource allocation and the optimal proving speed. 

With a strong focus on reliability and fault tolerance, Fermah ensures that proof generation remains efficient even in extreme scenarios. Instead of starting from scratch, we store intermediate outputs in the Storage Layer—Fermah’s decentralized storage system. This allows workflows to resume from the last successful checkpoint, minimizing wasted computation and keeping the process seamless.

Fermah Gives You Control 

When defining a workflow, you’re not limited to just specifying hardware requirements—you can also control how the computation is distributed. You can choose to run on a few high-powered machines or spread the workload across multiple smaller nodes, depending on your needs. 

Additionally, you can implement custom logic to dynamically scale resources, optimize costs, and distribute tasks efficiently. Whether you need raw power or want to maximize parallelism, Fermah gives you the flexibility to tailor proof generation to your exact requirements. 

Fermah is a step function improvement in ZK infrastructure. 

With it, the possibilities are endless.

Come Build The Future with Us

You don’t have to deal with the complexities of proof generation—Fermah lets you focus on building innovative products.

Ready to get started?

Fill out this form, and our team will get in touch to onboard you. We’ll also set up a dedicated Telegram chat, giving you direct access to our support team whenever you need assistance.

For in-depth technical details, check out our documentation.

Let’s bring moon math to the masses!

‍