Cisco, Palantir, NVIDIA Partner on Sovereign AI Infrastructure

Cisco announced a collaboration with Palantir and NVIDIA to deliver enterprise AI infrastructure designed for organizations that require strict control over proprietary data and critical systems. The partnership centers on Palantir's Ontology for Cybersecurity, delivered through Cisco's Secure AI Factory with NVIDIA serving as the foundation for Palantir's Sovereign AI OS.

The initiative addresses a recurring challenge as companies scale AI deployments: maintaining operational control while meeting compliance requirements in regulated industries and government settings. According to Cisco, the collaboration will provide validated reference architecture combining compute, networking, storage, security, and observability as a unified system.

The intelligence-cost-control framework

Cisco framed the partnership around three vectors that define practical AI strategy: intelligence, cost, and control.

On intelligence, the company noted that while frontier models achieve impressive benchmark scores, post-training open models like NVIDIA Nemotron with business-specific data often delivers better accuracy at lower cost for particular use cases. The relevant measure is how well a model performs on tasks specific to a business need, not abstract capability.

Cost extends beyond price per token to useful intelligence per dollar, per watt, and per unit of infrastructure—metrics that ultimately affect long-term business health.

Control means the ability to customize models and deploy them in chosen environments—cloud, edge, or on-premises—while maintaining authority over data, security posture, and operating conditions. This becomes critical for protecting intellectual property and meeting compliance mandates.

Cisco argued that the optimal model for any organization sits at the intersection of these three factors for specific workloads. A fraud-detection pipeline requires different architecture than a customer service agent, and classified workloads that cannot leave specific facilities demand entirely different approaches.

Architecture and deployment

The partnership combines distinct capabilities from each vendor. Cisco provides secure, observable infrastructure with unified management through Cisco Cloud Control. NVIDIA Nemotron offers a platform for post-training with proprietary data. Palantir contributes the operational AI layer through Palantir Foundry and its Artificial Intelligence Platform, grounded in the Palantir Ontology.

The focus on cybersecurity and IT operations reflects demand from organizations managing complex security environments. Forward-deployed engineers will assist companies and governments in building custom AI implementations.

Cisco noted that NVIDIA and Palantir already use this architecture as the blueprint for managing one of the world's most complex supply chains, with plans to expand to additional industries.

Why it matters

The collaboration reflects a strategic shift as post-trained open-weight models increasingly match or exceed frontier models on real-world applications. This enables enterprises to diversify AI strategies—using cloud-based frontier models for general tasks while running custom-trained models on-premises where control is paramount. For regulated industries and government agencies, the ability to maintain data sovereignty while deploying advanced AI represents a significant operational requirement that general-purpose cloud AI services cannot address.

Details were first reported by Cisco.