Aviz Networks and Spectro Cloud have announced a strategic partnership to transform NVIDIA AI infrastructure into a repeatable, governed platform-as-a-service. By combining Aviz ONES networking orchestration with Spectro Cloud PaletteAI management, the joint solution delivers unified operations for GPU-aware Ethernet fabrics, Kubernetes fleets, and NVIDIA AI Enterprise software, ensuring consistent performance, secure multi-tenancy, and simplified lifecycle management from Day-0 deployment to Day-2 operations.
Building and scaling AI factories involves far more than deploying GPUs—organizations need integrated orchestration to align topology, automate cluster lifecycles, enforce governance, and maintain predictable performance across diverse environments. The partnership addresses these challenges by creating a unified operations layer that standardizes infrastructure deployment while incorporating validated NVIDIA AI Factory reference architectures.
The joint solution combines Spectro Cloud PaletteAI’s capabilities for deploying and managing AI environments—including NVIDIA AI Enterprise components—with Aviz ONES’ deterministic fabric orchestration. This alignment ensures GPU clusters operate on optimized Ethernet networks with consistent east-west traffic performance, critical for both training and inference workloads.
Enterprises and service providers can enforce strict isolation across fabric segments, GPU and DPU resources, and Kubernetes namespaces using policy-driven guardrails. Zero-trust principles protect multi-tenant environments, enabling secure sharing of infrastructure while preventing unauthorized access or interference.
The platform automates Day-0 provisioning through Day-2 lifecycle tasks across hybrid, multi-cloud, edge, and sovereign setups. Unified observability integrates telemetry from networking fabrics, Kubernetes clusters, and AI workloads, accelerating troubleshooting, optimizing GPU utilization, and providing clear visibility into system health.
This partnership is tailored for enterprise AI platforms requiring governed multi-tenant clusters, large-scale AI factory deployments demanding deterministic performance, GPU Platform-as-a-Service providers offering isolated environments with SLAs, neo-cloud and GPU cloud operators delivering catalog-based services, and hybrid/sovereign operators unifying regulated environments.
“Cloud providers and enterprises want to consume GPU infrastructure like a product—without reinventing orchestration and operations for every cluster,” said Leon Jones, VP Alliances at Spectro Cloud. “With PaletteAI and Aviz ONES together, customers can standardize AI stack deployment, automate lifecycle management, and operate AI factories with consistent performance and governance across environments.”
“Aviz was founded to make AI networking simple, open, and AI-driven—while giving infrastructure teams a deterministic way to operate AI fabrics at scale,” said Thomas Scheibe, Chief Product Officer at Aviz. “With Spectro Cloud, we’re delivering a unified operations stack that aligns Kubernetes fleets, GPU-aware Ethernet fabrics, and multi-tenant segmentation so AI factories can be deployed faster, run predictably, and be consumed securely—like true infrastructure-as-a-service.”
The joint stack builds on NVIDIA AI Factory reference architectures, with Spectro Cloud handling full-stack AI environment management and Aviz providing GPU-aware networking alignment.
About Aviz Networks
Aviz modernizes networking for the AI era with a full-stack, software-first platform that helps enterprises accelerate AI factory investments and maximize ROI.
About Spectro Cloud
With our Palette and PaletteAI platforms, Spectro Cloud solves how enterprises and public sector organizations manage full-stack application and AI infrastructure in any environment: from edge to cloud, and from metal to model. Using the power of cloud-native technologies like Kubernetes, we give platform engineers and operations teams the flexibility to choose their preferred stack while benefiting from complete, repeatable consistency.