Nokia and Lenovo Sovereign AI Data Center Blueprint Solution

As AI becomes more pervasive in our lives, its data sources, data rights/permissions, where it is stored and where it can be used has become a critical determinant of where training and inference models are built and operated. Nation states and regulators are continuing to define new policies and laws that protect the rights of the data owners and ensure the security and privacy of that data. Gartner predicts that by 2028, 65% of governments worldwide will introduce technological sovereignty requirements to improve independence and protect from extraterritorial regulatory interference, highlighting the urgency for infrastructure that supports Sovereign AI.

New Sovereign AI clusters and workloads will require a new generation of high-performance servers with higher density CPU and GPU processing together with bandwidth capacity, scalability and improved performance of the date center network. These workloads can also be distributed across cloud networking infrastructures which requires network operations management visibility and automation capabilities that cut across network boundaries.

Nokia and Lenovo have architected, tested and validated a Sovereign AI data center networking blueprint solution that’s built to meet the demanding processing, network performance and security requirements for sovereign AI. It combines the high-performance Lenovo ThinkSystem line of servers and storage systems together with the Nokia Data Center Fabric solution.

Sovereign AI can be defined as AI systems that are designed, developed and deployed under the control and jurisdiction of a specific nation, organization or enterprise that removes any dependency of a foreign nation, technology, platform or outside influence. For details, see the NVIDIA blog post, What Is Sovereign AI?.

Key concepts of Sovereign AI include:

• National Control and Governance: Sovereign AI means having full control over the entire AI lifecycle—from data collection, data storage, and model training to deployment and governance—using domestic infrastructure, internal resources, and data. (Dale, 2025)
• Strategic Resilience: In order to reduce reliance on foreign AI systems. Nations need to prepare and adapt their own AI System to protect against geopolitical risks, potential security or system vulnerabilities in their critical infrastructure. (Alduhishy, Muath, 2024)
• Ethical and Regulatory Alignment: Sovereign AI systems are designed to comply with local laws, ethical standards, strategic priorities and core values, ensuring that AI technologies serve national interests. (Sideco, 2025)
• Economic and Innovation Boost: By investing in domestic AI capabilities, nations can accelerate innovation, create high paying jobs that attracts new talent in order to remain competitive in the global digital economy (Alduhishy, Muath, 2024)

Sovereign AI is a strategic response to the growing importance of AI in national security, economic competitiveness, and digital sovereignty.

Lenovo and Nokia have designed and optimized a high-performance, scalable AI pod blueprint solution that allows Government, Service Providers and Enterprises to quickly deploy and turn-up new AI services and use cases. The following table shows examples of the industry segments and use cases for sovereign AI.

Designing networks for Sovereign AI involves building infrastructure that ensures data control, security, performance, and compliance within national borders. According to IDC research commissioned by Lenovo, data sovereignty, compliance, and availability are the top success factors for AI implementation among CIOs.

Here are the key network design considerations and why they matter:

1. Data Sovereignty & Path Control. Sovereign AI must ensure that data used for training and inference does not leave authorized jurisdictions.
2. High-performance AI Fabric. AI workloads require massive parallel processing and low-latency communication between GPUs and nodes. (Ray, 2025)
3. Distributed Workload. AI training and inference involve large datasets and distributed compute nodes.
4. Security & Compliance. Sovereign AI must comply with national regulations and protect against cyber threats.
5. Modular & Scalable Architecture. Sovereign AI must support diverse use cases and scale rapidly.
6. Host Management & Orchestration. Smooth operation and debugging are essential for AI lifecycle management.
7. Federated & Hybrid Models. Full sovereignty may not be feasible for all nations.

Nokia and Lenovo have partnered to deliver high-performance, scalable, and secure blueprint solution for AI/ML data center combining Nokia’s Data Center Fabric and Event-Driven Automation with Lenovo’s ThinkSystem AI-ready portfolio to meet the demands of modern workloads.

This blueprint solution is built in modular AI pod units that support a wide range of AI applications or workload processing, footprint requirements or budget.

The basic AI pod unit consists of high-performance AI GPU servers, storage systems and management equipment together with a rail-optimized switching network. The AI pods units can be scaled out to support hundreds of servers and petabytes of storage.

The following figure shows an example of an AI pod unit that supports 64 GPUs.

Figure 1. AI-DC Blueprint Solution Equipment Rack Layout example for 64 GPUs

The Lenovo-Nokia Validated Design reference architecture is composed of three main networking segments. The backend network, the backend storage network, and the combined front-end and management network, as shown in the following figure.

Figure 2. Lenovo, Nokia Sovereign AI Validated Design Architecture

The major components in the architecture are as follows:

• The backend “scale-out” network supports GPU to GPU connectivity. It consists of Lenovo ThinkSystem SR685a V3 air-cooled AI GPU servers that host AMD Instinct GPUs and 400G (Gig) NICs that are interconnected with Nokia 7220 IXR-H4 GPU leaf switches. This design ensures that only a single hop is needed for GPU to GPU inter server communication.
• The backend storage network Nokia 7220 IXR-D5 high-performance switches that interconnects to Lenovo ThinkSystem SR630 V3 servers and DDN AIX400 storage nodes.
• The converged front end and in-band management network is represented by Lenovo ThinkSystem SR630 V3 GPU and management servers connected through Nokia 7220 IXR-D3L
• Nokia’s Event Driven Automation (EDA) platform, was used to simplify the configuration, validation and operational aspects of managing an AI cluster. This included the use of its back-end configuration application to design, configuration and deployment of the back-end fabric including multi-tiered architectures, multi-tenancy, IP addressing, and network parameter settings such as for congestion avoidance, and traffic prioritization.

Lenovo and Nokia ran an extensive suite of MLCommons.org performance tests on the referenced Sovereign AI design. These include:

• MLPerf tests for standard RoCEv2 packet sizes ranging from 256 to 4096 bytes. The testing results met or exceeded industry standards for the average bi-directional bandwidth achieved across 400Gb (gigabits) interfaces.
• MLPerf tests were also run for maximum RoCEv2 packet size of 4096 bytes. The testing results met or exceeded industry standards for the average bi-directional bandwidth achieved across all 400Gb Interfaces.
• MLCommons benchmark testing was run on the performance of a 70 billion Llama 2 training and inference models that measured both throughput and latency metrics.

The Nokia Data Center Validated Design Hub is a repository of rigorously tested, multi-vendor network designs to ensure performance and reliability standards are met as well as providing a template for rapid deployment and ease of operations.

The test results are included in the Lenovo-AMD-Nokia AI-DC Validated Design document.

The Nokia and Lenovo data center networking solution has been rigorously tested and validated to support the capacity, scale, performance, and reliability required for integrating AI workloads into enterprise environments. It combines Nokia’s expertise in data center networking and IP routing with Lenovo’s deep experience in high-performance servers and storage systems.

Key business benefits for enterprises include the following:

• Pre-validated, orderable server rack bundles
  Ready for deployment and available at discounted package pricing to accelerate time-to-value.
• Flexible network management options
  From DIY setups using open-source tools and APIs to full lifecycle automation via integrated orchestration platforms.
• Carrier-grade networking stack
  Proven reliability and robustness for mission-critical AI workloads.
• High availability and uptime
  Enabled by built-in component redundancy and active-active link configurations.
• Digital twin and rollback capabilities
  Support pre-validation, version control, and rapid restoration in complex environments.
• Open configuration and observability
  Designed for transparency, customization, and operational control.
• Comprehensive API support
  Facilitates seamless integration across compute, storage, and networking layers.
• Streaming telemetry with granular insights
  Enables real-time monitoring and proactive performance optimization.
• Infrastructure sovereignty and control
  Peace of mind with a trusted data center network built on your own assets.

Lenovo and Nokia together bring together a Sovereign AI solution that is unmatched in best-of breed technology, services and support:

• Open multi-vendor eco-system solution
  Providing flexibility of choice to build a solution tailored to your own needs.
• GPU/CPU chipset agnostic
  Multi-GPU support: AMD/NVIDIA/Other GPUS
• Built with industry standard specifications
  No proprietary protocols: based on standard RoCEv2 / future UEC protocols
• Certified full AI DC solution
  Blueprints that are rigorously tested and jointly validated in the labs
• Scalable AI pod configurations for simplified ordering
  Pre-defined full-stack AI blueprint solutions that take the guess-work out of ordering
• Network customization and operation the way you want it
  Open-APIs to interface both with the server and networking layers
• Rapid trouble-resolution and support
  Global tech support to answer your call, wherever you are located
• End-to-end data center solution across your network
  Seamless integration between the data center and the Wide Area Network (and across DCs)
• Industry leaders in AI, HPC Data Centers and IP/Optical
  Lenovo is deployed in over 25% of the top 500 Supercomputers in the world and Nokia is #1 market-share leader in IP edge routing worldwide.
• Advanced automation with AI for data center operations
  Network Automation to gain insights, simplify operations and reduce human errors
• Power-efficient platforms
  Optimized designs for lower power consumption
• Best of breed data center technologies
  Deep Networking Expertise from Nokia combined with advanced AI Server technologies from Lenovo
• Carrier-grade reliability
  Track record of reliability from Lenovo and Nokia with field-proven, carrier-grade systems that combined to deliver a powerful, reliable and resilient solution
• Global presence and expertise
  Expert teams present in the field to engage with customers and partners to identify the best solutions and support with any technical topics

Visit our website or talk to one of our experts:

• Nokia Data Center Networks
• Talk to a Nokia expert
• Lenovo AI Solutions
• Lenovo Data Center Support

For more information, see these references:

• Dale, Roberts: Cambridge University Press "Sovereign AI in 2025", August 2025.
  https://www.cambridge.org/core/journals/natural-language-processing/article/sovereign-ai-in-2025/C51560626AF518BDF280891C406E9553
• Mauth Alduhishy: World Economic Forum "Sovereign AI; What is it? And 6 strategic pillars for achieving it", April 2024. Muath Alduhishy - Agenda Contributor | World Economic Forum
  https://www.weforum.org/stories/2024/04/sovereign-ai-what-is-ways-states-building/
• Francis Sideco: Forbes "The AI Era Enters Its Sovereign Phase", June 2025.
  https://www.forbes.com/sites/tiriasresearch/2025/06/03/the-ai-era-enters-its-sovereign-phase/
• Trisha Ray, "Sovereign remedies: Between AI autonomy and control" Atlantic Council, April 2025.
  https://www.atlanticcouncil.org/in-depth-research-reports/issue-brief/sovereign-remedies-between-ai-autonomy-and-control/

Mark Wallis is a Senior AI Solutions Architect at Lenovo. He is passionate about building and testing the latest AI solutions to meet customers’ needs with a background in managing complex installations, performance testing and software automation. He is originally from the United Kingdom, but has been based in the USA for the last 25 years. When he is not working, he can be found spending time with his family or out and about, hiking and cycling.

David Nowoswiat is a Senior Product and Solutions Marketing Manager at Nokia and has over 25 years of telecom industry experience in both wireless and wireline technologies. He is currently in Nokia’s IP and Optical Business Group supporting Cloud Packet Core Marketing. He holds a Bachelor in Industrial Engineering degree from Georgia Institute of Technology and a Masters in Business Administration from Northeastern University.

Miguel Simões is the Alliance Director for the Global Nokia Partnership in the Lenovo Infrastructure Solutions Group. With more than 28 years of industry experience, Miguel is a veteran of the telecommunications world, now leading Lenovo’s global alliance with Nokia. In his current role, he drives strategic alignment, joint innovation, and business acceleration across enterprise and telco ecosystems worldwide. Miguel’s expertise spans operations, engineering, pre-sales, and business development, with previous leadership roles at Nokia where he led SD-WAN and data center transformation initiatives in Europe. Known for his customer-centric vision and technical depth, Miguel brings a collaborative spirit and global perspective to the stage. He holds a BSC in Electronics and Telecommunications Engineering and is a passionate advocate for agile, sustainable infrastructure