CIQ has expanded Fuzzball to support multiple cloud providers, including CoreWeave, AWS, Google Cloud, Oracle Cloud Infrastructure and Microsoft Azure.
Users can define an AI training, inference or high-performance computing workflow once and run it across those environments as well as on-premises systems, with the platform directing each job based on cost, performance and data location.
The approach addresses a common problem for companies running AI and HPC workloads across multiple infrastructure providers. Moving a workflow from one cloud to another often forces teams to rebuild pipelines, rewrite deployment scripts and repeat testing before jobs can run in a new environment.
Fuzzball uses a provider-neutral workflow definition, allowing the same file describing compute jobs, data movement, container images and resource requirements to run across different cloud services. Its orchestration layer then translates that definition for the underlying infrastructure.
According to CIQ, a genomics team that validates a sequencing pipeline on AWS can move it to Azure or OCI without changing the workflow definition. A training job that needs dense H100 GPU access could be sent automatically to CoreWeave, while a data-sensitive simulation could remain on premises under policy controls.
Single control plane
The platform federates across five public cloud environments alongside on-premises clusters. It evaluates available environments at runtime and routes each job to what it determines is the best destination.
CIQ argues that this gives customers access to specialist GPU infrastructure, large hyperscale cloud estates and locally controlled systems through a single control plane. Customers also do not need separate toolchains, deployment scripts or identity and access management models for each provider.
Security is a central part of the update. Each cloud deployment is provisioned through a two-stage automated process that creates a full cluster without manual setup, while maintaining a single identity and access management model, one set of role-based access control policies and one approach to secrets management across environments.
The platform uses cloud-native identity mechanisms rather than static credentials. These include Workload Identity on Google Cloud, Managed Identities on Azure, Dynamic Groups on Oracle Cloud Infrastructure and IAM Roles on AWS.
Gregory Kurtzer, chief executive officer and founder of CIQ, outlined the company's view of the issue.
"AI teams today are asked to ship faster, control costs and maintain sovereignty over their data, simultaneously, across infrastructure that was never designed to work together," said Gregory Kurtzer, chief executive officer and founder of CIQ. "We built Fuzzball to solve that problem at the architectural level. When your workflow definition abstracts its requirements properly, you get portable access to every GPU environment the market offers and the freedom to route to wherever the best price, performance and data policy lives. Controlling your infrastructure and workloads is what enterprise AI infrastructure requires for production, and no other platform delivers it."
Cloud complexity
Multi-cloud use has become more common as companies seek access to scarce GPU resources, compare pricing between providers and keep some workloads in specific jurisdictions or internal data centres. That has also increased operational complexity, particularly for AI projects that combine large data sets, containerised software and varied hardware requirements.
CIQ's pitch is that a single workflow layer can reduce that burden by separating job definitions from the underlying infrastructure. In practice, that means customers can move or split work between cloud providers and on-premises systems without rewriting the workflow itself.
Bjorn Hovland, president of CIQ, said the operational burden has been one of the biggest obstacles to wider multi-cloud adoption.
"Fuzzball turns multi-cloud from a liability into a competitive advantage," said Bjorn Hovland, president of CIQ. "Five clouds used to mean five IAM models, five deployment pipelines and five sets of operational overhead, with complexity and risk multiplied."
CIQ is known as a founding support and services partner for Rocky Linux and also develops products for Linux infrastructure, cluster provisioning and container orchestration. Fuzzball deployments across CoreWeave, AWS, Google Cloud, Oracle Cloud Infrastructure and Azure are available now, while on-premises deployment remains supported on Warewulf, VMware and bare metal systems.