Cisco and Palantir Build a Sovereign AI Security Stack
3 min readCisco used Palantir’s AIPCon 11 conference on September 10 to announce that it will deliver Palantir’s Ontology for Cybersecurity through the Cisco Secure AI Factory with NVIDIA. The same announcement names that Cisco and NVIDIA stack as the preferred full-stack foundation for Palantir’s newly published Sovereign AI Operating System.
What sovereign AI actually means here
Sovereign AI is the idea that a government or a large enterprise runs its models on infrastructure it owns and controls, with sensitive data never crossing the perimeter. It has become one of the fastest growing categories in enterprise infrastructure, driven by regulators, defense customers and companies that cannot legally send their records to a public cloud.
Cisco has been assembling the pieces since early 2026. The Secure AI Factory with NVIDIA bundles GPUs, networking, storage, security and observability into a validated design, and Cisco expanded it to rack scale in August. Palantir brings the other half: Foundry and its Artificial Intelligence Platform, built around an Ontology that maps an organization’s data into objects a model can reason over.
What Cisco and Palantir announced
The Cisco offering combines post-trained and fine-tuned NVIDIA Nemotron open models, security guardrails, and Cisco infrastructure spanning compute, networking, security and observability. The specific Palantir product involved is the Ontology for Cybersecurity, which applies that data model to security operations rather than to supply chains or logistics.
It landed alongside a separate but related announcement. NVIDIA and Palantir said the same day that they had built a sovereign AI stack for critical supply chains, and that the first deployment is running on premises inside NVIDIA’s own supply chain operations. Cisco security and collaboration lead Jeetu Patel laid out the company’s case in a blog post published the same morning, arguing that controlled AI deployment should not require a custom integration project.
Why it matters
Cisco has told investors that hyperscaler AI orders reached $9.3 billion in its last fiscal year, but the harder prize is the enterprise and government buyer who wants AI without handing data to someone else. Packaging Palantir’s software on top of a validated Cisco and NVIDIA design is an attempt to sell that buyer a finished product instead of a parts list.
The open question is whether customers will accept a three vendor bundle as simpler than building their own. Watch for named reference customers, and for whether rival full-stack designs from Dell, HPE and Supermicro pick up their own Palantir arrangements.