Summary: On January 9, 2026, the US Department of War published an AI Acceleration Strategy and three memoranda for an AI-first force spanning more than 3 million personnel. The plan names seven Pace-Setting Projects and includes GenAI.mil access to Gemini and Grok at Impact Level 5 and above. A model-parity provision sets a 30-day deployment goal after public release.
On January 9, 2026, the US Department of War (DOW, the renamed Department of Defense) put out its AI Acceleration Strategy, paired the same day with three memoranda shoving the whole department toward an AI-first fighting force. Secretary Pete Hegseth’s framing skips the nuance: build “an AI-first warfighting force across all domains.” Drafted at the President’s direction, the strategy spans three tenets — warfighting, intelligence, and enterprise operations — and touches more than 3 million military and civilian personnel.
This is not one more vision binder. The three memoranda tackle strategy execution, a data-platform restructuring, and an overhaul of the innovation bureaucracy, each carrying concrete deadlines from 30 to 270 days out. Inside the Pentagon’s document culture, those dates are where accountability anchors.
Seven Pace-Setting Projects, From Warfighting to Enterprise
The strategy’s heart is a set of seven “Pace-Setting Projects,” distributed across the three tenets:
- Warfighting: Swarm Forge, Agent Network, Ender’s Foundry
- Intelligence: Open Arsenal, Project Grant
- Enterprise operations: GenAI.mil, Enterprise Agents
Details on most projects remain classified, but the direction reads cleanly. The warfighting side wagers on drone swarms, agent networks, and simulation. The intelligence side concentrates on open deployment of models and analytic tooling. The enterprise side treats generative AI as infrastructure for the entire department. Layered on top, every component must nominate at least three “fast-follow” projects inside 30 days — so the list keeps lengthening.
GenAI.mil: Frontier Models for 3 Million People
The most concrete entry is GenAI.mil: department-wide access to frontier generative AI, with the official release naming Google’s Gemini and xAI’s Grok as available at Impact Level 5 (IL-5) and above. Emil Michael, the Under Secretary of War for Research and Engineering, said it without ornament: “Speed defines victory in the AI era, and this strategy ensures our people can move fast, adapt faster, and win. We cannot let bureaucracy serve as an excuse for mediocrity.”
An attached “AI model parity” provision obliges the department to field the latest models within 30 days of public release — in effect forcing the military to track the commercial release cadence rather than letting procurement cycles stretch a version gap into a generational one.
Rewiring the Institution: Data, Procurement, and Bureaucracy
The remaining two memoranda handle the business of making the organization move. One rebuilds the Advana data platform into a “War Data Platform” with milestones at 30, 60, 120, and 270 days, aimed at finally clearing financial audits that have dragged for years. The other crowns the Under Secretary for Research and Engineering as the department’s sole Chief Technology Officer, disbands layers of steering and working groups, and elevates organizations such as the Defense Innovation Unit into direct-reporting field activities.
Procurement and talent run on their own clocks: model objectivity benchmarks inside 90 days, “any lawful use” contract language inside 180 days, AI talent hiring plans inside 60 days, and a monthly Barrier Removal Board to knock obstacles flat. The same memorandum demands Service Innovation Plans from each branch within 90 days and carves a dedicated innovation increment out of the FY2028 budget — this agenda arrives with money stapled to it. Holland & Knight’s analysis hands defense contractors the plain signal: decision authority is consolidating around the CTO, and vendors who hold the timelines take the seats.
What It Means for Contractors and Builders
Three practical effects. First, contract language tilting toward “any lawful use” moves model and agent vendors onto a playing field of capability and delivery speed rather than bespoke procurement packages. Second, data interfaces are converging on a single platform — integrating with the War Data Platform will cost less engineering than yesterday’s multi-point hookups, and components will expect vendors to plug in on the milestone schedule. Third, with CES 2026 the same week throwing its spotlight on physical AI and autonomous systems, civilian and military AI release cycles are lining up — the 30-day model parity clause is precisely that alignment cast as policy. The seven project names plus the 30-day fast-follow pipeline sketch the demand list for the next two years; anyone building agents, simulation, or data infrastructure can read that list as a roadmap.