White House Defense Tech Strategy Prioritizes Foundation Models, Raises China-Deterrence Concerns

The White House released a technology strategy aimed at accelerating emerging military capabilities. Within AI, it highlights “multi-agent systems and swarm intelligence” and lists foundation models—including large-language, multimodal, and world-system models—as critical for deterrence in the Indo-Pacific.

Defense One

The strategy builds on earlier executive orders that sought to modernize defense acquisition and aligns with calls from defense-tech leaders and some members of Congress for a more agile system.

Defense One

• The Replicator program, a pathfinder for autonomous aircraft, receives $1 billion in funding.

Defense One

• Agencies are directed to create “streamlined pathways” for small firms, including CRADAs and ACTs.

Defense One

• The plan seeks to relax foreign-military-sales rules and export-control constraints to help early-stage companies secure investment.

Defense One

The document calls out foundation models as the primary AI avenue, omitting open-weight models and open-source software. Critics note that open-weight models—publicly available, parameter-efficient alternatives—are absent.

Defense One

Michael Schiffer, a partner at Scalare Advisors and former Deputy Assistant Secretary of Defense for East Asia, warned that emphasizing large, compute-intensive models could advantage China, which is advancing open-weight AI.

Defense One

Dario Amodei, co-founder of Anthropic, has argued that raw scale—more compute and data—drives AI capabilities more powerfully than algorithmic innovations. The strategy’s focus on scaling aligns with this view, favoring a handful of U.S. cloud providers and frontier AI labs.

Defense One

Pentagon officials say the new procurement tools are intended to “bypass traditional acquisitions” and enable rapid fielding of technologies such as drone swarms, giving younger defense-tech startups “more room to experiment.”

Defense One

• Strategic Concentration: Privileging foundation models may concentrate AI development in a few large firms, limiting diversity.

Defense One

• Operational Constraints: Large-model tools require substantial connectivity and compute, which may be unavailable to troops in contested environments. Jake Steckler of the Carnegie Endowment warned that Ukraine’s computational architecture is already strained as it integrates more AI into targeting.

Defense One

• Public Sentiment: A poll cited in the strategy found only 18 percent of Americans view AI as a positive force for the United States, suggesting challenges in securing congressional and financial support.

Defense One

The strategy’s priority areas—undersea, space, and AI/autonomy—are framed as essential to “deterrence in the Indo-Pacific.” Analysts caution that limited compute, data, and public goodwill could become a strategic vulnerability if support erodes. If China continues to advance open-weight AI, the United States may face a gap between its high-cost approach and more distributed, energy-efficient alternatives.

Defense One

• Replicator program funding: $1 billion.

Defense One

• Public perception of AI: 18 percent positive (poll).

Defense One

• Ukraine’s AI-enabled targeting architecture is reported to be straining under increased computational demands (Carnegie Endowment report).

Defense One

The sources do not provide alternative funding estimates or detailed timelines for the new procurement mechanisms, leaving the pace of adoption uncertain. No quantitative comparison of U.S. versus Chinese AI capabilities is offered, creating a gap in assessing the strategy’s effectiveness against its deterrence goals.

Defense One