As ZK proofs become central to scaling solutions, privacy infrastructure, and decentralized verification systems, the demand for scalable, production-grade proving infrastructure increases swiftly. Yet while the cryptographic foundations of ZK continue to advance, deployment remains a bottleneck. There is no industry standard for building, scaling, and managing ZK infrastructure—especially for organizations aiming to run provers efficiently at scale.
This lack of standardization creates friction across the ZK ecosystem. Whether a team is developing a new protocol or contributing prover capacity to existing networks like Scroll or Aleo, the operational challenges are significant.
GPU resources are fragmented. DevOps workflows are inconsistent. Hardware provisioning and configuration remain largely bespoke, increasing time-to-market and cost.
For ZK to fulfill its potential across mainstream use cases, from scalable L2s to proof networks and verifiable AI, there must be a shift toward unified tooling and production-ready infrastructure.
A Standardized Solution: ZKContainers and ICICLE
To address these challenges, Ingonyama introduced ZKContainers, a new packaging framework purpose-built for zero-knowledge workloads. At the heart of this system is their flagship product, ICICLE, a CUDA-based framework that brings the same accessibility and modularity to ZK proof generation as PyTorch brought to AI development.
Together, ZKContainers and ICICLE establish a new paradigm for deploying ZK applications efficiently and reliably on GPU infrastructure.
A ZKContainer is a Docker image that includes ICICLE and all the dependencies needed to run a high-performance prover. With ZKContainers, developers can rapidly prototype new applications, configure storage and caching strategies, switch between different proving systems (like Groth16 and Plonk), and scale from a single machine to a full data center with minimal operational overhead.
Importantly, these containers are hardware-aware. They are pre-optimized for Nvidia GPUs and designed to work out-of-the-box with GPU data center tooling like the NVIDIA GPU Operator. They support multiple container runtimes (including Docker, containerd, and Singularity), and can be deployed across bare metal, virtual machines, Kubernetes clusters, or any environment that supports containerized workloads. Built-in logging, monitoring, and security scanning make them robust for enterprise-grade usage.
RockawayX: Infrastructure to Accelerate ZK at Scale
As part of our mission to support foundational blockchain infrastructure, RockawayX partnered with Ingonyama to demonstrate how ZKContainers can be deployed across a high-density GPU environment.
We provided access to one of our GPU-optimized data centers, capable of supporting over 6,000 GPUs in total. Each node in the cluster consists of 8 Nvidia GPUs, forming a powerful platform for batch testing, live deployment, and performance benchmarking of containerized ZK applications.
Together with Ingonyama, we focused this collaboration on deploying a custom ZKContainer running an Aleo consensus prover. The container was built using Ingonyama’s CLI tooling and packaged with all necessary dependencies, enabling seamless deployment across our infrastructure. To orchestrate the deployment, we used Ansible for automated provisioning and configuration. Once the nodes were prepared, the prover containers launched automatically—each consuming the correct number of GPUs per machine and beginning to generate proofs immediately.
The results validated the scalability and efficiency of ZKContainers. Each node reached approximately 56,000 proofs per second (PPS), with the total performance of the data center scaling linearly with the number of GPUs. Configuration errors were virtually nonexistent. What would have been a daunting manual task—ensuring compatibility across thousands of GPUs—was streamlined thanks to containerized deployment and native monitoring tools.
Most importantly, this architecture is future-proof. As hardware requirements change or new proving systems are adopted, updates can be made to the container image and redeployed across the data center instantly. ZKContainers decouple the complexity of ZK workloads from the underlying infrastructure, giving teams the flexibility to adapt and evolve.
Handling High-Performance ZK Workloads With Ease
At RockawayX, our infrastructure is designed from the ground up to meet the rigorous demands of next-generation decentralized networks. Our data centers, located across Central Europe with a primary facility in Prague, provide a secure, high-availability environment tailored for high-performance computing and zero-downtime operations.
We operate entirely on enterprise-grade, owned hardware—eschewing public cloud for production blockchain servers to deliver maximum reliability, control, and security. All hardware components are dualized to eliminate single points of failure: dual power supplies, dual networking paths, and redundant inter-data center links ensure uninterrupted operation. Our stack currently includes 1000+ physical CPU cores, nearly 30+TB of RAM, and over 400TB of storage capacity, optimized for compute-intensive applications like ZK proof generation.
This robust infrastructure makes RockawayX an ideal partner for teams building and scaling ZK technologies. We provide a reliable, secure, and performance-optimized environment for deploying prover networks, running containerized workloads, and executing cryptographic operations at scale.
Pushing the Industry Forward
This collaboration between RockawayX and Ingonyama showcases what’s possible when specialized ZK software is matched with industrial-grade infrastructure.
At RockawayX, we believe that solving critical infrastructure gaps is key to unlocking the next wave of blockchain innovation. We’re proud to support builders like Ingonyama who are laying the groundwork for scalable ZK systems, and we’re actively seeking to collaborate with more teams that share this vision.
If you're developing ZK applications or need GPU infrastructure support to scale your prover systems, we invite you to reach out to our team.
Let’s build the future of cryptographic infrastructure together.
