Palantir Technologies announced the delivery of new sovereign AI outcomes for a slate of flagship customers, expanding on its work with NVIDIA to help organizations build and control AI systems using their own data, models, and infrastructure. The latest deployments span defense, construction, semiconductor supply chains, mission-critical autonomy, and health care.
L3Harris has unified more than 30 enterprise resource planning systems, connected hundreds of software applications to Palantir’s Ontology, and created roughly 3.5 million data connections alongside more than 5,000 scheduled data feeds. Building on open-source models and NVIDIA’s Nemotron family, L3Harris is moving to own its models, compute, and operational data. The company is exploring bespoke AI models to inform mission systems pairing in the field, with ambitions for rapid super intelligence and enhanced capabilities on next-generation hardware such as VAMPIRE. “We are thrilled to explore and develop capabilities required for critical defense use cases. By building our own models trained on proprietary data, we believe we can surface unique capabilities for critical use cases in our industry,” said Heidi Wood, Chief Transformation Officer at L3Harris. “We have tested the capabilities of sovereign models already on internal work where initial results have been promising. In a few days we were able to outperform the frontier labs in our test environment and cut costs by 95%. Frontier labs are not training their models with relevant data for our industry, where we also have industry-specific needs for security and zero-retention. Given our unique experience and our ability to leverage our data, we can build AI that is purpose-fit for the mission and retain critical rights to terms of use using open-source models. We view sovereign AI as a mission critical to our work supporting Western defense.”
McCarthy Building Companies is using Palantir’s Foundry, AIP, and Ontology as an AI operating system to infuse more than 160 years of construction expertise into project workflows from design through field execution. The company fine-tuned and benchmarked NVIDIA Nemotron against other models on more than 100,000 production RFIs to anticipate cost and schedule impacts. “Our clients get a higher level of confidence and greater cost and schedule certainty at every phase of the project. We built this model on 160 years of McCarthy experience, and it outperforms the frontier at a lower cost,” said Rico Bertucci, Vice President of Strategic Initiatives. “No one else can pick it up, because it is tuned to how we build.” McCarthy has prioritized RFI classification—fine-tuning a Nemotron 3 Nano 4B model in Foundry with labeled RFI data—to flag emerging risks and triage issues earlier, and an AI-assisted subcontractor drafting workflow that uses post-trained Nemotron on NVIDIA compute to translate scopes of work into contract language. The latter is delivering better-than-frontier performance at significantly lower cost and saving 30,000 hours annually.
NVIDIA and Palantir have built an operating system that shortens the path from wafer to deployment across one of the world’s most complex supply chains, combining Palantir’s AIP platform, NVIDIA’s Agent Knowledge Foundation, and post-trained Nemotron models. The system helps teams detect potential supply constraints sooner, evaluate alternatives faster, and optimize supply flows while maintaining control over proprietary data, customized models, and workflows. “As demand for intelligence scales rapidly, speeding up our supply chain accelerates how quickly our customers reach the first token. With Palantir and post-trained Nemotron models, we spot potential problems earlier, act on them sooner, and put our best expertise behind every call,” said Jeff Whitmer VP of System Supply Chain at NVIDIA. “This is what sovereign intelligence looks like in practice – built with our own data, under our control, and delivering for our customers at the speed business demands.”
Ondas is expanding its collaboration with Palantir across manufacturing, mission planning, and edge intelligence, and is launching Regna AI to deliver sovereign AI for mission-critical operations. Built on Palantir’s AIP and leveraging open-weight models including NVIDIA Nemotron, Regna AI enables customers to adapt, evaluate, and deploy AI using their own data and domain expertise while retaining control of proprietary data, models, and infrastructure. “Through Regna AI, we are expanding our partnership with Palantir and leveraging open models from NVIDIA to enable governments and enterprises to develop and deploy mission-specialized AI on their own terms,” said Eric Brock, Chairman and CEO of Ondas. “This expands on our existing projects with Palantir, which span from building the manufacturing & supply chain backbone for Ondas Sentinel through Warp Speed all the way to significant progress on a true mission autonomy solution for SkyWeaver. At the enterprise level, Palantir has enabled rapid expansion in corporate development and post-merger integrations through deployment of Palantir’s CRM and M&A Workflows, which enabled the integration of DZYNE into Ondas Sentinel on the order of weeks. Our focus is on delivering sovereign AI capabilities that ensure customers retain full ownership and control of their data, models, and infrastructure while transforming proprietary knowledge into differentiated intelligence and mission-critical applications.”
Tampa General Hospital, a customer since 2021, is using Foundry, Ontology, and AIP to reduce patient hold times, improve nurse staffing ratios, respond to natural disasters, flag early signs of sepsis, reduce mortality rates, and improve patient outcomes. As it expands AI, the health system is implementing sovereign controls that keep sensitive data and model processing local. Tampa General is classifying large language model traffic against its own risk and privacy standards, fine-tuning and post-training Nemotron to improve performance by 7% and reduce latency by a factor of 22. “AI Innovation in health care only works if our patients and team members can trust it,” said Dr. Jaimie Weber, Associate Chief Medical Informatics and AI Strategy Officer at Tampa General Hospital. “By keeping their data local and setting our own controls, we can expand AI that improves outcomes for our patients and makes the work easier for the team members caring for them.”
The developments follow a recent joint update from Palantir and NVIDIA on how Palantir’s Foundry, Ontology, and AIP are accelerating NVIDIA’s supply chain and speeding the path from wafer to first token, with multiple enterprises now pursuing sovereignty efforts that emphasize ownership of data, models, and compute.







