Palantir’s Military AI Is Getting Real — And Demand Is ‘Unprecedented’

Palantir shares jumped as much as 21% early Tuesday after the company previewed its new AI platform, AIP, and dropped some pretty wild claims about demand.

CEO Alex Karp, during Monday’s analyst call, said the demand for AIP is “nothing I’ve ever seen in 20 years of being involved in Palantir.” He described the boom in large language models as a revolution “that will raise ships and sink ships.” Palantir is “reorganizing our efforts aggressively to capitalize” and “running hard” at the opportunity.

AIP stands for Artificial Intelligence Platform. The first version rolls out to some customers this month. The company’s website and demo videos show it being used by militaries to tap the kinds of AI models that power ChatGPT, but for battlefield intelligence and decision-making. One demo displays enemy targets, identifies hostile situations, proposes battle plans, and sends those plans to commanding officers for execution.

The video emphasizes safety and security — clients decide what data the models can see and what they “can and can’t do on behalf of humans.” Karp said: “If you wheel these technologies correctly, safely, and securely, you have a weapon that will allow you to win, that will scare your competitors and adversaries.”

Civilian applications are also on the table. Another demo shows a manufacturing company using AIP to prepare for a hurricane strike — analyzing distribution center operations and deciding whether to accelerate, delay, or cancel orders. It forecasts the impact on customer orders and revenue. An insurance client with early access described it as “years ahead” of other solutions. Within days, that client built a “collaborative AI agent” to automate claims processing.

Details like pricing and terms are still being decided, though Palantir has had conversations with “hundreds” of potential partners. The company recently refocused engineering teams and other resources around AI to meet demand.

Palantir will share more on June 1 at an event in Palo Alto.

I’ve seen a lot of AI hype cycles, but this one feels different. Palantir has always been in the defense and intelligence space, but AIP is the first time I’ve seen them package LLMs into something that feels actionable rather than just a demo. The fact that an insurance client built a working agent in days is telling — but I’m skeptical about how well these models will perform under real battlefield conditions, where latency, reliability, and adversarial manipulation are critical. The safety controls are a good start, but they’re only as good as the people configuring them. We’ll see how this plays out.