The announcement landed on a Friday afternoon, and by the following Monday, the contours of the largest government AI contract in history were still coming into focus. On May 1, 2026, the Pentagon confirmed it had reached agreements with seven artificial intelligence companies to deploy their technology across classified military networks, a move that will fundamentally reshape how the US armed forces operationalise artificial intelligence in combat environments.
The companies involved read like a who's who of the AI industry: SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, and Amazon Web Services. Each has agreed to terms that allow the deployment of their systems on the Defense Department's most sensitive network environments, known as Impact Level 6 and 7, which handle classified intelligence and combat operations. The agreements accelerate what Defense Secretary Pete Hegseth described in January as the push to make the US military an "AI-first fighting force." [1]
The deal is unprecedented in its scope. For the first time, frontier AI models from commercial labs will be integrated into classified government infrastructure at this scale. OpenAI, whose ChatGPT has defined the consumer AI era, will now power decision-support tools inside networks that are inaccessible to the public. Google will do the same with Gemini. Nvidia will supply the underlying compute infrastructure. [1][3]
The Coalition That Wasn't: Anthropic's Exclusion
Notably absent from the agreements is Anthropic, the AI developer behind the Claude chatbot. The company refused to include language permitting "any lawful use" of its technology by the military, a stipulation the Pentagon requires of all its AI contractors. The dispute escalated sharply last month when the Defense Department designated Anthropic a supply-chain risk, the first time an American company has received that classification. [1]
The designation effectively bars Anthropic from Pentagon contracts. Defense officials have made clear the move is designed to pressure the company toward compliance, betting that signing Anthropic's competitors will bring the holdout startup back to the negotiating table. Anthropic has sued in response, arguing the designation is unconstitutional and that the lawful use clause could enable autonomous weapons and domestic mass surveillance. [1][2]
The Pentagon's strategy appears to be working. With OpenAI, Google, and others now embedded in classified networks, Anthropic faces the prospect of permanent exclusion from a market it helped define. Whether that brings the company back to the table or hardens its position remains to be seen.
Reflection AI: The Unknown Variable
Among the seven signatories, Reflection AI is the least known and perhaps the most politically charged. The two-year-old startup has never released a publicly available model. Its stated goal is to build open-source AI systems as a counter to Chinese firms such as DeepSeek. According to the Wall Street Journal, Reflection is seeking a valuation of $25 billion and has received investment from Nvidia and 1789 Capital, the venture fund where Donald Trump Jr is a partner. [1]
The inclusion of a company with no public product in the Pentagon's most classified AI environments is striking. It suggests either exceptional confidence in Reflection's technology or exceptional confidence in its political alignment. The company did not respond to requests for comment.
What $54 Billion Buys
The Pentagon has requested $54 billion for autonomous weapons development alone in 2026. The AI agreements announced May 1 represent the operationalisation of that spending. The contracts cover integration across decision-support, intelligence synthesis, and operational planning systems. The goal, according to the Pentagon's statement, is to "maintain decision superiority across all domains of warfare." [1]
The broader context is a dramatic acceleration in government AI spending. In the first half of fiscal year 2026, the Department of Defense committed over $32 billion in contract ceiling to AI, cloud computing, cybersecurity, and data analytics programs. That figure does not include the infrastructure deals being signed with commercial AI firms, which are being funded through separate appropriations. [4]
The commercial AI sector is spending even more aggressively. Amazon, Meta, and Oracle have collectively committed between $660 billion and $690 billion in capital expenditure for 2026, nearly doubling 2025 levels, according to the Futurum Group. Much of that spending is being funnelled directly into the government-adjacent cloud and infrastructure contracts that these Pentagon deals validate. [5]
The Bigger Picture: AI's Military Turn
The May 1 agreements mark the moment when the AI industry's commercial ambitions and the US government's military strategy became structurally inseparable. Every major AI company now has a direct financial stake in the expansion of classified government AI infrastructure. That alignment creates powerful incentives to support further government spending in this area.
It also raises questions the contracts do not answer. The lawful use clause covers use by the US military, but what happens when AI systems are deployed in coalition operations with other nations? What constraints apply when AI tools built for decision support are repurposed for autonomous targeting? The Pentagon's statement references "any lawful use," but lawful and ethical are not synonymous.
Anthropic's refusal to sign was framed by the company as an ethical stance. The Pentagon's response was to label it a supply-chain risk. That framing matters: it positions non-compliance as a security threat rather than a moral objection, which has implications for how other companies might evaluate their own red lines. [1]
The agreements are not static. They are framework arrangements that will be refined as the integration work progresses. Impact Level 6 and 7 networks are among the most sensitive environments in the US government, and the technical challenges of deploying large language models in those contexts are significant. The coming months will test whether the promises of May 1 translate into working systems.
For now, the line has been drawn. Seven AI companies are inside the perimeter. One is outside. The deals they have signed will shape the nature of military AI for years to come.