Palantir Technologies is betting big on a simple idea: The U.S. government should never have to depend on someone else’s AI model to protect the country.
That idea is now a formal partnership.
Palantir (PLTR) recently announced it is teaming with NVIDIA to give federal agencies and companies that run critical U.S. infrastructure their own version of powerful AI models, ones they can control, retrain, and keep locked inside secure walls.
The announcement comes at a moment when Palantir is already riding one of the strongest growth stretches in its history, and it signals the company wants to plant its flag even deeper in the national security world that built its reputation.
Why sovereign AI matters in 2026
For years, the biggest worry about using commercial AI models inside government has been simple.
Sensitive data goes in, and nobody outside the vendor really knows what happens to it once it’s incorporated into a model’s training.
Palantir and NVIDIA (NVDA) say their new offering solves that problem.
It combines NVIDIA’s open-source Nemotron models with Palantir’s existing products, including AIP, Foundry, Ontology, and Apollo, so agencies can train and run AI without handing proprietary information to an outside company, according to a company statement.
More Palantir:
- Palantir’s 2 deals answer a question investors keep asking
- Michael Burry pulls back on massive Palantir short bet
- Palantir flashes a warning signal Wall Street can’t ignore
The system includes several built-in protections.
These cover explicit data authorization, secure isolation between customers, and the ability for an agency to fully erase its data whenever it wants.
Perhaps the most notable feature is what the companies call a self-improving model.
The system collects data on how users interact with it and uses that information to retrain the model over time, so it gets better at the specific job it was built for, whether that is battlefield logistics or fraud detection.
“Combining Palantir infrastructure with NVIDIA’s AI and Nemotron models will allow the U.S. government to unleash the full power of LLMs while removing the underlying security risks,” said Palantir CEO Alex Karp.
Nvidia CEO Jensen Huang framed the deal in similar terms, calling open-source AI foundational to national security and U.S. technology leadership, according to the statement.
Fabrice COFFRINI / Getty Image
Palantir’s growth streak sets the stage
This partnership comes right after Palantir posted one of the strongest quarters in the company’s history.
- In the first quarter of 2026, Palantir grew revenue 85% year over year, its fastest pace ever as a public company.
- Its U.S. business surpassed 100% year-over-year growth for the first time, climbing 104%.
- U.S. government revenue grew 84% year over year during the quarter.
- The company’s Maven Smart System, used to support military operations, saw usage roughly quadruple over the past year.
Chief Technology Officer Shyam Sankar described the moment as a turning point for cybersecurity as well, noting that AI systems paired with Palantir’s platform are now finding vulnerabilities in operating systems and browsers faster than defenders can patch them.
Related: Cathie Wood makes striking Palantir move as shares tumble
That backdrop helps explain why the Nvidia deal focuses so heavily on control and security, rather than just raw AI capability.
Palantir has built its brand on being the company government agencies trust with their most sensitive data, and the new offering doubles down on that positioning at a time when rivals are racing to sell general-purpose AI tools into the same market.
What comes next for agencies and investors
The joint offering will roll out through deployment engineering for classified and air-gapped environments, along with tools that let agencies fine-tune how the models behave in production.
Nvidia AI Enterprise software, including Nvidia NIM microservices, will support the rollout.
For Palantir shareholders, the announcement reinforces a theme that has driven the stock’s run over the past two years.
The company keeps expanding its footprint inside the institutions that matter most for national security, while positioning itself as the layer that keeps AI models grounded in real, auditable operations rather than experimental demos.
Whether this translates into new contract wins will likely become clearer in the coming quarters, as agencies decide how much control they are willing to build in-house versus buying it directly from Palantir and Nvidia.
Related: Palantir CEO gets painfully honest about AI’s biggest problem
