Customer stories > Scottdata
Scott Data Built the World's First Multi-Tenant AI Cloud on NVIDIA Spectrum-X Networking Platform with Netris
In March 2025, Scott Data became the first GPU cloud operator in the world to deploy NVIDIA Spectrum-X networking with full multi-tenancy, and only the second Spectrum-X deployment globally.
Scott Data is a data center and AI cloud operator headquartered in Omaha, Nebraska. An NVIDIA Cloud Partner (NCP), the company began as a colocation facility, expanded into managed services, and evolved into an operator of GPU cloud infrastructure.
Scott Data’s facilities meet Department of Defense and National Intelligence Community security criteria, serving more than 100 customers ranging from small businesses to Fortune 100 enterprises.
With the Netris NAAM (Network Automation, Abstraction, and Multi-Tenancy) platform as the foundation for AI networking, Scott Data transformed from a traditional data center operator into a fully automated and multi-tenant AI cloud provider.
Today, Scott Data can provision new customers in seconds to minutes, dynamically resize GPU pools, and securely reassign resources across tenants in real time — adding GPU capacity for customers ramping up projects or reallocating when workloads wind down — serving customers across a wide range of industries, including highly regulated sectors like finance, government, defense, and education.
"Our whole goal with this Spectrum-X-based AI cluster was to make it dynamic and multi-tenant. Netris made that possible."
Dominic Romeo
Head of Technology Services at Scott Data
The Challenge: No Fabric Manager, No Automation, No Multi-Tenancy
When Scott Data decided to launch a GPU-based AI cloud on NVIDIA Spectrum-X, they faced a fundamental problem: Spectrum-X didn't include a fabric manager, automation platform, or multi-tenancy layer. Scott Data's team — with decades of networking expertise across enterprise and service provider environments — recognized these gaps immediately. Without these capabilities, Scott Data couldn't launch their AI cloud quickly, onboard tenants in minutes, deliver hard isolation, or dynamically reallocate GPU resources.
Scott Data's Requirements
1. Launch the AI Cloud Quickly
Scott Data needed to get their AI cloud business operational fast so their GPU investment could start generating ROI immediately.
2. Hard Multi-Tenancy
Scott Data needed hard isolation (enforced on networking hardware) to securely share GPU resources across multiple customers while meeting strict requirements for data sovereignty and compliance.
3. Onboard New Customers Quickly
Scott Data needed to provision new tenants in minutes — not days or weeks — with full network isolation configured automatically. Manual provisioning through tickets, scripts, and ad-hoc configurations would create bottlenecks and delay customer activation.
4. Respond to Changing Customer Needs in Real Time
Scott Data needed the ability to resize GPU pools on demand — expanding capacity when customers ramp up projects or reallocating resources when workloads wind down — without downtime or disruption to active jobs.
5. Uptime and SLA Requirements
Scott Data needed to deliver consistent uptime and avoid SLA violations. Manual network configuration carries an error rate of approximately 20% — misconfigurations that could disrupt customer workloads, create security vulnerabilities, and damage customer trust.
Why Building In-House Wasn't Viable
Scott Data faced a choice: build network automation and multi-tenancy in-house, or find a purpose-built platform.
After evaluating what it would take to build in-house, they concluded it could burn years and waste millions of dollars. The result would be fragile and expensive to maintain. It would also pull focus away from the AI cloud business — essentially turning Scott Data into a network automation company instead of a GPU cloud operator.
Without automation, any tenant lifecycle changes (e.g. adding, removing, or resizing tenants) would take days or weeks and carry a high risk of disrupting other operational workloads. Dynamic hard isolation (enforced on networking hardware), mandatory for securely sharing GPU resources across multiple customers, would not be operationally achievable through manual processes.
"If we had to manually spin up or resize tenant isolation without Netris, we'd be looking at days or weeks for the provisioning time."
Dominic Romeo
Head of Technology Services at Scott Data
The Evaluation: Why Scott Data Selected Netris
Scott Data went through a thorough evaluation of multiple solutions and ultimately selected Netris.
Scott Data selected Netris based on rigorous hands-on testing and NVIDIA validation.
Using NVIDIA Air, their engineers ran proof-of-concept deployments and implemented all of their specific use cases. Netris delivered everything Scott Data needed and performed exceptionally well across the entire evaluation. NVIDIA engineers confirmed that Netris was aligned with NVIDIA's deployment guides and validated for Spectrum-X environments.
"We were not able to find anything else that was even remotely competitive."
Dominic Romeo
Head of Technology Services at Scott Data
A dedicated Slack channel with same-hour response times reinforced confidence. Netris became the essential platform for AI networking, automating both east-west and north-south traffic across the Spectrum-X environment.
Netris NAAM Platform Capabilities
AI Networking Foundation
Netris is the foundational platform for AI infrastructure operators, delivering cloud-provider-grade network automation, abstraction, and multi-tenancy across the entire AI networking stack: Ethernet, InfiniBand, NVLink, DPUs, Virtual, and Edge Networking. Netris absorbs the complexity and networking nuances of various reference architectures, accelerating the launch and operation of AI clouds.
Network Automation
Netris delivers the missing automation layer for Ethernet and DPU fabrics, with best practices built into its auto-configuration algorithms. Engineers define the end state, and Netris configures VRFs, VXLANs, BGP, and everything else — all in under two minutes with closed-loop assurance. This network automation speeds day-0/1 deployment and gives operators confidence and agility for day-2 maintenance.
Abstraction
Netris provides a cloud-grade network abstraction layer via web console and REST API, which is essential for AI cloud services. It enables AWS-style constructs like VPCs, peering, elastic IPs, load balancers, and more, while Netris algorithms translate them into precise configurations across Ethernet, InfiniBand, NVLink, DPUs, Virtual and Edge Networking. This delivers simplicity over complex, multi-fabric infrastructure.
Hard Multi-Tenancy
Netris enables true multi-tenancy by providing hard isolation (enforced on networking hardware) across bare metal, VM, and container workloads. This hard isolation — now the industry standard for data sovereignty and compliance — is fully enforced on switches, DPUs, and fabrics. Operators simply define the endpoints, and Netris algorithms automatically generate and apply the correct configurations across all network elements.
"When we set up the cluster template and click 'Deploy' with Netris, it's done in seconds or minutes. It's amazingly fast, and it's very consistent."
Dominic Romeo
Head of Technology Services at Scott Data
"Being able to abstract all that away, not having to have my engineers spend a year learning a new network operating system, and just being able to administer the network through the point-and-click GUI or the API of Netris was really powerful for us."
Dominic Romeo
Head of Technology Services at Scott Data
Key Outcomes
Netris delivered immediate and ongoing operational impact for Scott Data — enabling them to launch their AI cloud quickly, onboard tenants in minutes, and instantly resize GPU pools as customer needs change.
Faster time to revenue — Scott Data got their GPU cloud business operational quickly, without wasting years and millions of dollars building in-house automation.
Tenant provisioning in seconds to minutes — Scott Data onboards new customers in minutes with full network isolation configured automatically — no tickets, no scripts, no bottlenecks.
Real-time GPU pool resizing — Scott Data resizes GPU pools on demand — expanding capacity when customers ramp up projects or reallocating resources when workloads wind down — without downtime or disruption to active jobs.
Maximized GPU utilization — Scott Data eliminates idle capacity by dynamically reallocating GPUs across tenants as customer needs change.
Reduced operational risk — Scott Data delivers consistent uptime and avoids SLA violations by replacing manual processes with automated, validated orchestration — eliminating the misconfigurations that disrupt customer workloads.
Future-proof — Netris evolves with each new NVIDIA GPU generation, Reference Architecture (RA), new networking platforms, and implementation guidelines.
Support and partnership — A dedicated Slack channel provided same-hour technical assistance response times. Not only does Netris provide exceptional implementation and support, but it is also a great source for learning about best practices based on their vast experience of working with other customers.
"The Netris team was really eager to help us, even when I probably would say those are not things that they needed to solve. They were invested in us being successful."
Dominic Romeo
Head of Technology Services at Scott Data
Looking Ahead
With Netris as the core of its AI networking infrastructure, Scott Data has grown its customer base, expanded GPU capacity, and continues to deliver the reliability and flexibility their customers depend on — while maximizing GPU utilization and ROI.
"It's very consistent, much more consistent than humans can be, because we know that we all have typos or places where we get distracted."
Dominic Romeo
Head of Technology Services at Scott Data
Technical Environment
Scott Data operates a production, multi-tenant AI Cloud supporting both training and inference workloads. The platform is built around hard multi-tenancy, treating tenants as mutually untrusted by default and enforcing strict, hardware-level network isolation with controlled connectivity across the entire environment.
The east–west (backend) network is rail-optimized and is designed for non-blocking GPU-to-GPU communication based on NVIDIA Spectrum-X reference architecture. The multi-tenant north–south (frontend) network provides secure, hardware-isolated tenant ingress, egress, and external integration without compromising performance. A highly-available 3-node Netris controller interprets the customer’s intent (e.g., tenant-acme: {server51, server12, server37}) and, in a closed loop, automatically configures each switch in both fabrics to enforce the desired tenant isolation.
Scott Data operates NVIDIA H100 GPU-based HGX servers, alongside additional accelerators such as NVIDIA L40S GPUs, supporting a mix of training and inference workloads within the same AI Cloud platform.
Both the east–west and north–south fabrics are built using NVIDIA SN5600-series switches for both leaf and spine network layers, as well as SN2201 switches are deployed as leafs for management (IPMI, BF3 IPMI, switch management). East–west network leverages BlueField SuperNICs, with the Netris host agent enforcing consistent SuperNIC configuration aligned with NVIDIA-validated designs.
Additionally, to enable controlled, tenant-aware, cloud-like ingress and egress at the north-south fabric edge, Netris SoftGate provides address translation (NAT) and Layer-4 Load Balancing services.
Netris enables Scott Data to operate the entire AI Cloud with a lean infrastructure operations team, with only a small subset specializing in network engineering. Scott Data engineers use Netris’ UI for everyday tasks, Netris Terraform provider for Infrastructure-as-Code (IaC) operations, and Netris REST API for integrating with other management systems and platforms.
About Scott Data
TBA
About Netris
Netris is the leading Network Automation, Abstraction, and Multi-Tenancy (NAAM) Platform purpose-built for GPU Clouds and Enterprise AI Factories. It provides the essential foundation for transforming GPUs from idle capital expense into sustainable revenue. By automating complex multi-fabric network environments across Ethernet, InfiniBand, NVLink, DPUs, Virtual, and Edge networking, Netris enables operators to launch AI clouds in weeks instead of years, maximize GPU utilization, and ensure secure, compliant multi-tenancy without the cost and risk of in-house automation. To learn more, visit www.netris.io