Pentagon Signs AI Deals With Eight Big Tech Vendors, Cuts Out Anthropic

The Pentagon has finalized a sweeping set of artificial intelligence agreements with eight of the biggest names in tech — and pointedly left Anthropic off the list. The Defense Department signed deals with OpenAI, Google, Microsoft, Amazon, Oracle, Nvidia, SpaceX, and Reflection AI to extend AI use across classified environments and core military workflows including analysis, logistics, and large-scale data processing.

Each company is contributing different layers of the stack. Microsoft, Amazon, and Oracle bring both AI models and the secure cloud infrastructure cleared for classified workloads. Nvidia is supplying its Nemotron family of open-source models for autonomous agents, while Reflection AI — a comparatively young entrant — is building bespoke models for specific defense applications. Emil Michael, undersecretary of defense for research and engineering, said the goal is to give military personnel AI tools to “maintain an advantage and achieve decision-making superiority.”

The most striking detail is who is not in the room. Anthropic, whose Claude models had been accessible to the Defense Department through Palantir’s Maven platform and reportedly used in operations including the Iran conflict and actions targeting Venezuelan leader Nicolas Maduro, was designated a “supply-chain risk” by DoD following a contract dispute. The company is contesting the designation in court, arguing that the label could blacklist it from a fast-growing slice of federal AI spending.

The fight is also personal. Defense Secretary Pete Hegseth, in congressional testimony, called Anthropic CEO Dario Amodei an “ideological lunatic,” a remark that landed against the backdrop of Anthropic’s tighter usage policies on military and surveillance applications relative to its peers. Anthropic has historically been more restrictive about how its models can be deployed in weapons-adjacent or kinetic contexts, a stance that wins it ESG credibility but appears to have cost it a seat at this particular table.

The exclusion is notable because Anthropic has been the standout commercial performer of the past quarter, with annual recurring revenue reportedly eclipsing OpenAI’s. Losing access to classified defense work narrows one of the highest-margin enterprise channels in the country and concentrates that flow toward rivals. For OpenAI, Google, and Microsoft, already deeply embedded in federal cloud, the Pentagon awards reinforce a moat that is rapidly hardening into a two-tier industry: AI vendors cleared for classified work, and everyone else.