1 week ago
20
Ondas Inc. and Palantir Technologies have launched Regna AI. The new company will build custom AI models for governments, defense organizations and enterprises that want to keep their data, and their algorithms, under their own roof.
The announcement came on October 12, 2026. The pitch is simple, even if the engineering is not: AI that answers to its owner, runs where its owner says, and keeps working when the internet doesn’t.
Who owns what, and who does what
Regna AI is a private company. At launch, Ondas (Nasdaq: ONDS) owns all of it.
Palantir (Nasdaq: PLTR) is not taking an ownership stake at the outset. Instead, it is supplying the model-training infrastructure and engineering support.
The technical plan rests on open-weight architectures. An open-weight model is one whose internal parameters, the numbers that determine how it thinks, are available to the user rather than locked behind a vendor’s servers. That matters for sovereignty, because a government can inspect, modify and run the model itself.
Regna AI plans to train these models on each customer’s own data. The finished models will then be deployed inside environments the customer controls, using Palantir’s AIP, Foundry and Apollo platforms.
The target is what the companies call mission-specialized, autonomous AI. These are models built for a specific operational job, rather than general-purpose chatbots asked to moonlight as defense planners.
Ondas is customer number one
Ondas will also be Regna AI’s first customer.
Ondas plans to use the models to orchestrate missions on edge hardware. In this context, edge means the physical devices out in the field: drones, ground vehicles and command-and-control nodes. The AI runs on the equipment itself, not in a distant data center.
The focus is on settings with denied, degraded, intermittent or limited connectivity. A model that needs a constant link to a remote server is of little use when that link is jammed or simply absent.
The launch also arrives as Ondas reports strong commercial momentum. The company said it had booked more than $165 million in new orders as of October 7, 2026. Total orders since June 30, 2026 have climbed above $270 million.
Built on an existing partnership
This is not a first date. Ondas and Palantir announced a previous collaboration in March 2026.
The two companies have also worked together on the SkyWeaver C4I program. C4I is military shorthand for command, control, communications, computers and intelligence. SkyWeaver focused on improving multi-domain Intelligence, Surveillance and Reconnaissance, or ISR, which covers the gathering and sharing of battlefield information across land, air and other domains.
Regna AI effectively takes that working relationship and gives it a dedicated corporate home. The earlier work centered on coordinating systems. The new company centers on building the AI brains those systems run on.
The sovereign AI argument
Both chief executives framed the launch around independence. Ondas CEO Eric Brock emphasized that governments should be able to control their AI strategies without relying on outside dependencies.
Palantir CEO Alex Karp went further. He argued that sovereign AI models will outperform traditional frontier models while reducing dependencies.
Executives at both companies also pointed to the risks of depending on foreign or uncontrolled models. For a defense ministry, outsourcing the intelligence layer of its weapons and surveillance systems to a third party is a hard sell. Regna AI is designed to remove that third party from the picture entirely.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
English (US) ·