Pentagon seeks nearly $30B for AI

- The Pentagon requested $29.5 billion for fiscal 2027 on May 22 to build AI supercomputers and hardened data centers under an “AI Arsenal” plan. - Emil Michael said on May 20 the department’s AI user base reached about 1.5 million, up from roughly 80,000 a year earlier. - Congress must approve the fiscal 2027 request, while Cyber Command and the NSA study classified deployment of OpenAI and Google models.

The Pentagon is asking Congress for $29.5 billion in fiscal 2027 to buy next-generation AI supercomputers and build the secure infrastructure needed to run them, according to budget documents reported by DefenseScoop on May 22. The proposal would fund hardened, Sensitive Compartmented Information Facility-accredited data centers across multiple sites and an initial fleet of GPUs and AI systems under a new “AI Arsenal initiative.” A Pentagon official told DefenseScoop the effort is meant to create “foundational, government-owned AI infrastructure” and “maximize federal buying power.” The spending request lands as the Defense Department expands AI use across the force. Emil Michael, the under secretary of defense for research and engineering, said on May 20 that AI use had climbed by roughly 1,420,000 users over the past year to about 1.5 million people out of a workforce of more than 3 million. Michael said the department is using AI at the enterprise, intelligence and warfighting levels. (defensescoop.com) ### Why is the Pentagon asking for such a large AI infrastructure budget? The fiscal 2027 request is tied to a shift away from what budget documents described as “scattered clusters” of graphics processing units toward a centralized compute portfolio covering both strategic and tactical requirements. DefenseScoop reported that the plan would scale supercomputing assets across the joint force and expand a portfolio of highly secure data centers. (war.gov) A Pentagon official told DefenseScoop that the initiative follows the White House AI Action Plan’s emphasis on “robust AI infrastructure.” The department’s January 2026 AI strategy also called for rapid buildout of foundational enablers including infrastructure, data, models, policies and talent. (defensescoop.com) ### Where does the military say AI is already being used? Michael said in Tampa, Florida, during Special Operations Forces Week 2026 that the department is embedding AI into systems used by warfighters to improve precision, speed and decision-making. He said the most important layer was warfighting use, and linked AI adoption to battlefield operations as well as intelligence and back-office functions. (defensescoop.com) The same Pentagon appearance included an example from unmanned systems. Michael said the department’s Drone Dominance Program has allocated $1.1 billion for 200,000 small lethal drones by 2027, and that officials are trying to widen competition among vendors. ### Which AI models is the Pentagon testing now? (war.gov) Bloomberg reported on May 21 that the Pentagon is testing rival AI models as it looks for alternatives to Anthropic, with 25 departmental “power users” evaluating systems. A separate report from CXO Digitalpulse said the Pentagon had begun testing models from OpenAI and Google after restricting Anthropic systems, though that account cited reporting rather than direct Pentagon comment. (war.gov) The Defense Department’s January 2026 AI strategy said the military would pursue “experimentation with America’s leading AI models” across the department. That document framed commercial model access as part of a broader push to become an “AI-first” warfighting force. ### What is the Pentagon-NSA task force supposed to do? (bloomberg.com) Politico reported on May 20 that U.S. Cyber Command is launching a joint task force with the National Security Agency to study how leading AI models can be deployed safely across their missions, including on “high-side” classified systems. The task force was announced to staff two weeks earlier by Gen. Joshua Rudd, the dual-hat leader of the NSA and Cyber Command, according to Politico. (media.defense.gov) Politico said the effort is focused in part on frontier models with advanced cyber capabilities from companies including OpenAI and Google. The report said the agencies are examining how those systems could be used on some of the government’s most sensitive networks. ### What happens next in Washington? Congress will decide whether to approve the fiscal 2027 defense request, which DefenseScoop said sits inside a broader Pentagon budget proposal totaling $1.45 trillion, including $1.1 trillion in discretionary funds and $350 billion in anticipated mandatory reconciliation funding. (politico.com) At the same time, Cyber Command and the NSA are continuing their classified AI review, while Pentagon officials expand testing of commercial models for military and intelligence use. (defensescoop.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.