The Pentagon has quintupled its enterprise AI agreement with Scale AI, raising the contract ceiling from $100 million to $500 million in a move that signals the Department of War is no longer treating artificial intelligence as a science project. The expansion, disclosed in early May 2026 and detailed in defense-industry reporting on May 8, comes roughly eight months after the original Production Other Transaction Authority (OTA) was awarded.
The contract is run out of the Chief Digital and Artificial Intelligence Office (CDAO), the Pentagon's central AI organization, and uses an OTA mechanism specifically designed to bypass the fragmented, multi-year acquisition cycles that have historically slowed defense software adoption.
From pilots to production
When the original $100 million ceiling was set in September 2025, it was sized for what officials assumed would be a measured rollout. According to Scale AI and reporting from Bloomberg and GovCon Wire, demand from across the Department of War rapidly outran that envelope. Components from the Army, Navy, and Marine Corps, along with defense agencies and Office of the Secretary of War offices, all began initiating Project Agreements under the same vehicle.
The scope listed in the expansion is notably broad: computer vision tooling, generative AI decision-support, data operations and labeling pipelines, classified generative AI applications, and systems supporting autonomy and targeting. In practice, that means a single contracting vehicle now covers everything from analyst workflows to weapons-relevant capability development.
Hegseth's AI memo, in dollars
The expansion aligns with a January 2026 strategy memo from Defense Secretary Pete Hegseth calling for accelerated AI adoption across the department and a reduction in bureaucratic hurdles. The Scale deal is the most concrete dollar figure attached to that direction so far, eclipsing most prior CDAO-led commercial agreements.
It also lands in a politically charged moment for AI procurement. Anthropic remains in litigation with the Pentagon over a separate supply-chain dispute, and the White House is reportedly drafting an executive order to vet new frontier models before deployment. Against that backdrop, a quietly expanded Scale AI ceiling effectively re-anchors the department's near-term AI infrastructure strategy on a Meta-backed data and tooling provider rather than on a frontier-model lab.
Why this matters
For the AI industry, the headline number is less interesting than the structural signal. A five-fold ceiling lift on an existing OTA tells vendors and rivals that the Pentagon's AI line item has crossed an internal threshold: enough programs are using these tools day-to-day that procurement is now scaling to meet demand, not seeding it.
For Scale AI, fresh from its multibillion-dollar Meta tie-up in 2025, the deal cements a defense business that complements its commercial data-labeling and evaluation work. For competitors — Palantir, Anduril, Microsoft, and frontier-model labs courting classified workloads — it sets a benchmark every future award will be measured against.
The broader lesson is straightforward. After several years of pilots, demos, and proofs-of-concept, U.S. military AI is starting to be priced like infrastructure rather than research.
