Supermicro builds a full-stack AI factory reference architecture with SUSE

Highlights

  • Cuts the PoC-to-production cycle for AI considerably, with a proven end-to-end solution.
  • Maximizes GPU utilization with fractional GPU sharing, boosting ROI.
  • Automates workload prioritization to protect mission-critical inference workloads, supporting business continuity.
  • Enables robust data sovereignty and compliance, securing AI-powered operations in even the most highly regulated industries.

Products

Supermicro is a leading global IT solutions provider, specializing in building and deploying high-performance, application-optimized server and storage infrastructure. The company operates in over 100 countries and has more than 7,000 employees. Able to produce $50 billion worth of racks annually, Supermicro’s robust supply chain is helping meet the accelerating global demand from customers for AI infrastructure, while cutting their time-to-deployment using its unique Data Center Building Block Solutions®.

At-a-Glance

Supermicro chose SUSE to build an end-to-end AI factory reference architecture. The combination of Supermicro and SUSE technologies helps organizations accelerate their AI rollout and maximize their ROI while maintaining sovereignty and governance. Supermicro is publishing a reference architecture, coming soon, that scales from a single server with one GPU to multi-rack systems with thousands of GPUs. This approach enables customers to use the same sizing methodology for their initial AI deployment and future expansions.

Building a complete AI factory architecture

Supermicro is committed to providing first-to-market infrastructure solutions tailored to enterprise, cloud and AI platforms. Supermicro has delivered complete rack-level systems — networking, cooling and software preinstalled and validated — for many years. The modular plug-and-play architecture, called Data Center Building Block Solutions (DCBBS), helps customers deploy new infrastructure faster.

The rise of AI-enriched solutions brings new challenges for the enterprises that are Supermicro’s target customers. Many organizations have retrofitted GPU capacity into their data centers but are struggling to achieve ROI.

To help its customers rebalance the economics of AI, Supermicro wanted to create a factory reference architecture optimized for the entire AI lifecycle. An AI factory combines compute, data, models and orchestration to create scalable, continuous AI services.

Supermicro aimed to deliver an architecture that integrates the latest-gen GPUs and innovative liquid-cooling solutions to meet its customers’ demands for AI processing at scale. However, to help customers deploy AI effectively, Supermicro needed not just powerful hardware, but the modular, scalable software that could support customer transitions from AI pilots to production. To enable end-to-end turnkey AI data center deployment, Supermicro looked for a software partner with open, sovereign and supported solutions.

 

“With SUSE AI Factory in place, our customers can easily deploy AI on just one server or create a gigafactory with thousands of GPUs. The SUSE technology helped us make our AI offering flexible and repeatable, giving every customer exactly what they need to cut time-to-value.”

Supermicro AI Factory Architecture Team

Why SUSE solutions?

Having been a long-term SUSE partner, Supermicro chose to work with SUSE once again for key elements in its planned AI reference architecture and selected SUSE solutions which deliver open source flexibility with enterprise-grade backing, supporting Supermicro’s own strategy of offering as much choice as possible to customers.

SUSE’s focus on digital sovereignty was also a key factor in the choice, helping Supermicro give its customers confidence that their proprietary information will stay secure. At the same time, SUSE matches Supermicro’s global footprint, enabling Supermicro to deliver and support the solution anywhere in the world.

SUSE Linux Enterprise Server acts as the resilient, AI-ready foundation for the software infrastructure. For AI processing at the edge, SUSE Linux Micro offers a small, immutable footprint suited to edge and disconnected sites.

The SUSE Rancher Suite combines solutions optimized for containerization on-premises, in the cloud or at the edge. SUSE Rancher Prime enables seamless Kubernetes cluster management and scalability across any CNCF-certified Kubernetes distribution, with exclusive K3s and RKE2 Long Term Support of up to five years. SUSE Observability and SUSE Security deliver comprehensive insights and zero-trust controls. Together, they streamline compliance without impacting developer productivity. Finally, SUSE Storage provides low-latency persistent block storage for Kubernetes, keeping data secure.

At the top of the stack sits SUSE AI Factory. SUSE AI Factory is the management plane for the AI platform itself: it lets a team discover AI applications and compose them into immutable, version-controlled stacks — referred to as blueprints — so that what runs in production is a declared, reproducible object rather than an assembly of parts. Customers can also choose to deploy either SUSE AI Factory or SUSE AI Factory with NVIDIA, which is embedded with NVIDIA AI Enterprise software for best-in-class AI deployments. 

 

The impact of SUSE solutions

From PoC to production faster

SUSE AI Factory is a Private Enterprise AI solution that enables production-ready AI workloads and mission-critical enterprise operations. With SUSE AI Factory, Supermicro accelerated the creation of its AI factory architecture. The SUSE technologies augment Supermicro’s hardware to create a true out-of-the-box sovereign AI capability for Supermicro's customers. SUSE AI Factory offers pre-validated blueprints for business use cases, as well as the freedom to create customized deployments. This means that customers with strong ideas but only limited AI experience can hit the ground running and quickly achieve value from the solution.

Together, Supermicro and SUSE provide services that help customers right-size architecture and select the appropriate tools for their use cases. 

Customers can manage every layer of the stack through a user-friendly, centralized UI in SUSE Rancher Prime, reducing the potential operational burden on their IT teams. At the same time, rapid cluster provisioning with SUSE Rancher Prime supports effortless scalability.

“With SUSE AI Factory in place, our customers can easily deploy AI on just one server or create a gigafactory with thousands of GPUs,” confirms the Supermicro AI Factory Architecture Team. “The SUSE technology helped us make our AI offering flexible and repeatable, giving every customer exactly what they need to cut time-to-value.”

Maximizes GPU utilization to boost ROI

SUSE AI Factory with NVIDIA embeds NVIDIA AI Enterprise into the SUSE platform, and with it NVIDIA Run:ai, which handles GPU orchestration. A single GPU can be divided into slices that several workloads use at once, and scheduling policy decides who gets what share. A GPU bought for a peak workload therefore continues to do useful work between the peaks instead of waiting for them.

With hardware costs rising continuously, optimizing utilization will help customers significantly boost their ROI for AI projects. In addition, this path to more cost-effective scalability will further help organizations move from experimentation into production.

Automates workload prioritization to support business continuity

Selecting SUSE AI Factory with NVIDIA enables Supermicro to offer intelligent, automated balancing of development, training and production workloads to its customers. This ensures that there is always enough GPU power for mission-critical applications, supporting business continuity.

In practice that means production inference holds a guaranteed share of the cluster, and development and training work runs on what is left — expanding when the cluster is quiet and yielding when it is not. The customer sets the policy once; the platform enforces it without requiring manual adjustments.

Enables robust data sovereignty and compliance

By choosing SUSE solutions, which are secure by default and give organizations complete ownership and auditability over every layer, Supermicro can meet customer demands for AI sovereignty. The solution gives customers the freedom to deploy SUSE AI Factory on any infrastructure, including on-premises. By using SUSE to enable on-premises deployments in this way, Supermicro is also helping to meet digital sovereignty standards and emerging regulations like the EU AI Act. 

To support regulated industries, the reference architecture ensures that proprietary data remains completely within the private infrastructure. Moreover, SUSE publishes Software Bills of Materials for its product media and container images, and its platform components carry Common Criteria and FIPS 140-3 validations, which can be reliably reviewed by auditors. 

“As governments move to leverage AI solutions, our use of SUSE AI Factory enables Supermicro to give them the confidence that sensitive data will remain protected,” adds the Supermicro AI Factory Architecture Team.

What’s next for Supermicro?

Supermicro is preparing to launch its AI factory reference architecture, with further refinements to follow. The Supermicro AI Factory Architecture Team concludes: “Together with SUSE, we are not just selling licenses and hardware, we are supporting our customers at every stage. With complete integration and pre-validation across the stack, we give organizations the platform to do what they do best.”

The work continues in two directions: extending the architecture across more workload sizes and storage designs, and turning the sizing method into something a customer can apply to their own model and concurrency targets without help.