Pentagon’s Big AI Awards
The office running Pentagon AI work has contracted major AI labs for agent-style workflows — roughly $600 million went to Anthropic, Google and xAI on top of an earlier $200 million deal with OpenAI — with procurement framed around ethical safeguards for operational use. (x.com)
The Pentagon is no longer buying just one artificial intelligence chatbot at a time. In June 2025 it gave OpenAI a contract with a ceiling of $200 million, and on July 14, 2025 it gave Anthropic, Google, and xAI their own contracts with the same $200 million ceiling. (openai.com) (ai.mil) All four deals came from the Pentagon’s Chief Digital and Artificial Intelligence Office, the unit that acts like the Defense Department’s central buyer for data and artificial intelligence systems. The office said these awards are meant to speed up adoption of “advanced AI capabilities” across national security work. (ai.mil) (cnbc.com) The new phrase in the paperwork is “agentic artificial intelligence.” That means software that can carry out a chain of tasks on its own, more like a junior staffer following a checklist than a chatbot answering one question at a time. (ai.mil) (breakingdefense.com) The Pentagon says those task chains could touch both office work and operational work. OpenAI described its June 2025 pilot as covering health-care access for service members and families, acquisition data, and proactive cyber defense, while Defense Department reporting also said frontier systems could be used in frontline warfighting operations. (openai.com) (breakingdefense.com) This is also a procurement strategy, not just a technology strategy. By giving separate contracts to OpenAI, Anthropic, Google, and xAI, the Pentagon can compare models from rival labs instead of locking itself into one company’s system too early. (ai.mil) (defensescoop.com) The money sounds huge, but “ceiling” is the key word. A $200 million ceiling is the maximum the government could spend under each award, not proof that all $800 million has already been paid out. (ai.mil) (cnbc.com) The Pentagon wrapped the July 2025 announcement in language about “responsible use” and safety guardrails. Google said its award would support secure deployment across the defense ecosystem, and the Defense Department said the goal was operational use with appropriate safeguards rather than a free-for-all rollout. (ai.mil) (cloud.google.com) The timing tells you where Washington thinks the market is going. OpenAI won first in June 2025, Pentagon officials signaled more frontier artificial intelligence partnerships were coming the next day, and by mid-July the department had lined up the rest of the top U.S. model makers. (defensescoop.com) (ai.mil) So the real story is not one contract. It is the Defense Department building a bench of commercial artificial intelligence labs for jobs that range from paperwork triage to cyber defense to mission support, with each company getting a shot at becoming part of the Pentagon’s everyday operating system. (openai.com) (breakingdefense.com)