CIQ redefines advanced computing with Fuzzball 4.0, democratizing access to HPC for researchers

Native storage integration, a built-in object cache and container registry, and multi-cloud support across every major hyperscaler allow organizations to consume HPC and AI as a capability rather than operate it as an engineering project.

RENO, Nev. — June 11, 2026 — CIQ, the enterprise software company behind Rocky Linux and the Fuzzball AI and HPC orchestration platform, today announced the release of Fuzzball 4.0. The latest version is built to meet the production requirements of national laboratories and HPC centers.

For decades, organizations treated infrastructure complexity as the price of advanced computing. National laboratories and HPC centers built specialized expertise around storage systems, registries, cloud integrations, and operational tooling before researchers could run a single workload. Fuzzball 4.0 removes the infrastructure expertise historically required to run advanced computing, allowing organizations to consume HPC and AI as a capability rather than operate it as an engineering project.

"Every transformative platform in computing history introduced an abstraction layer that made something previously out of reach, consumable. Beowulf democratized access to compute by replacing specialized hardware with commodity systems. Fuzzball democratizes access to outcomes by replacing specialized operational expertise with intelligent infrastructure abstraction. For the first time, organizations can treat HPC, AI training, AI inference, cloud resources, and on-premises infrastructure as a single consumable capability. What we have built is the supersuit for scientific computing: the most powerful HPC and AI platform in the world, finally as easy to deploy, operate, and scale as the science demands."

— Gregory Kurtzer, CEO and founder, CIQ

Deploy on existing infrastructure and file systems

Research organizations have long treated storage migration as an unavoidable prerequisite to adopting new HPC platforms. Fuzzball now eliminates that prerequisite by connecting directly to existing parallel file systems, including the multi-petabyte Lustre, GPFS, and BeeGFS environments common in national laboratories and HPC centers, with no requirement for data migration, duplication, or changes to established storage architecture.

Volumes are externally managed and dynamically imported. Organizations run workloads against data where it already lives. A generic hostpath storage driver extends compatibility to arbitrary file system backends, giving operators the flexibility to use Fuzzball with whatever storage infrastructure their environment runs.

Operate on a complete platform with native object cache

The latest version also adds a native object cache that retains ingressed data for reuse across workloads, eliminating redundant data movement and reducing time-to-compute for iterative and recurring jobs. Researchers can now spend less time waiting for data and more time running experiments. The cache provides a direct staging path for workflow inputs and outputs, accessible through both the CLI and web interface.

An integrated container registry decouples workload execution from external registry availability, removing a common point of failure for production HPC deployments. Organizations push container images directly into Fuzzball without a separate registry, lowering prerequisites for new deployments and improving operational resilience for established ones.

Together, these capabilities collapse a multi-service deployment into a single platform. Organizations no longer need to stand up, maintain, or integrate separate infrastructure before research teams can work.

Run workloads anywhere, from AWS to Google Cloud and beyond

Researchers no longer need separate workflows for on-premises and cloud environments. With the addition of Azure, Fuzzball 4.0 now supports deployment across AWS, Google Cloud, Oracle Cloud, CoreWeave, and Azure. The same workflow runs on an on-premises cluster or any supported cloud without modification, enabling organizations to place workloads wherever capacity exists, data governance requirements allow, or cost models dictate.

A unified command-line deployment interface delivers a consistent installation and management experience regardless of target environment. A redesigned CLI and rebuilt web interface consolidate administrator and user workflows in a single experience, surfacing status, dependencies, and service endpoint visibility in real time.

Availability

Fuzzball 4.0 deploys to AWS, Google Cloud, Oracle Cloud, CoreWeave, and Azure, as well as on-premises clusters built with Warewulf, VMware, or bare metal. To get started, contact CIQ at ciq.com/products/fuzzball.

CIQ will host two live sessions for organizations evaluating modern HPC and AI infrastructure.

On June 18, "How to modernize your on-prem HPC without rebuilding it" brings together CIQ's HPC engineering team for a hands-on walkthrough of Fuzzball running on existing clusters, covering storage integration with parallel file systems, native data caching, container image builds without an external registry, and running the same workflow across on-premises and cloud environments without modification. Built for HPC administrators, infrastructure architects, and research computing leads. Join at 2:00 PM ET / 11:00 AM PT. Register for the on-prem HPC webinar.

On June 25, "Cut through the AI hype: get real infrastructure that delivers" brings Addison Snell of Intersect360 Research and CIQ founder and CEO Gregory Kurtzer together to address where enterprise AI adoption actually stands, where organizations consistently get stuck, and how intelligent infrastructure abstraction changes what is possible. Join at 2:00 PM ET / 11:00 AM PT / 8:00 PM CEST. Register for the AI infrastructure webinar.