The U.S. military's latest technology strategy emphasizes foundation models and large language models as the path to AI dominance, but this narrow focus may inadvertently benefit China by overlooking faster, more efficient open-weight alternatives that researchers increasingly favor. The White House strategy, released in August 2026, champions energy-intensive approaches favored by a handful of major tech companies while omitting open-source and open-weight models entirely, according to defense analysts.

What Are Foundation Models and Why Does the Military Care?

Foundation models, including large language models (LLMs) like those from OpenAI and Anthropic, represent one approach to building AI systems. These models work by processing enormous amounts of data through massive computing infrastructure to achieve something resembling reasoning, though it is fundamentally probability calculation at scale. The strategy explicitly lists "foundation models, including large language, multimodal, and world systems" as critical technologies for military support.

The problem, according to defense strategists, is that this approach creates a winner-take-all dynamic. If raw computational power and data volume are the primary drivers of AI capability, then only organizations with access to the most resources can compete. This concentration of power in a few well-connected U.S. tech companies may actually weaken American military readiness rather than strengthen it.

How Does China's Open-Weight Strategy Challenge U.S. Dominance?

China has been quietly advancing in open-weight AI models, which use freely available parameters rather than proprietary, protected weights. These alternative architectures achieve comparable performance to closed foundation models while consuming far less energy. Michael Schiffer, a partner at Scalare Advisors and former Deputy Assistant Secretary of Defense for East Asia, warned that the U.S. strategy overlooks this critical vulnerability.

"China's advances in open-weight AI have exposed the limits of a U.S. strategy built too heavily on the idea of denying them access to American technology and American markets," Schiffer stated.

Michael Schiffer, Partner at Scalare Advisors and former Deputy Assistant Secretary of Defense for East Asia

The omission of open-weight models from the White House strategy is particularly striking because the military faces real operational constraints that foundation models cannot easily solve. Troops deployed in the field cannot rely on the constant cloud connectivity that large language models require. Additionally, Ukraine's experience integrating AI into targeting and coordination has already revealed how Western cloud-dependent architectures strain under real combat conditions.

What Practical Problems Does This Create for Military Operations?

The military's reliance on foundation models creates several tactical vulnerabilities that open-weight alternatives could address more effectively:

• Connectivity Requirements: Foundation models depend on constant connection to cloud infrastructure, which is unreliable in forward-deployed military units and combat zones where connectivity is intermittent or nonexistent.
• Energy Consumption: Large language models require enormous computational resources, making them impractical for edge devices and field equipment where power is limited and must be rationed carefully.
• Operational Latency: Open-weight models can run locally on military hardware without round-trip delays to distant data centers, enabling faster decision-making in time-critical situations.
• Strategic Vulnerability: Dependence on a handful of private companies for AI capabilities creates a single point of failure if those companies face financial pressure, regulatory action, or supply chain disruption.

The strategy does emphasize other military priorities that align with broader deterrence goals in the Indo-Pacific region. It calls for accelerated development of drone swarms, autonomous systems, and multi-agent coordination capabilities. It also pushes the Pentagon to adopt faster, non-traditional acquisition approaches and to relax export controls that currently hobble defense companies' ability to sell technology to allies.

Why Is Public Sentiment About AI Becoming a Strategic Problem?

The concentration of AI development in a few frontier labs backed by major cloud providers has created a public trust crisis. Just 18 percent of Americans believe AI will be a positive force for the United States, according to recent polling cited in the defense analysis. If public and congressional support for AI research erodes due to concerns about energy consumption, concentration of power, or perceived corporate capture, the U.S. could lose its technological edge to China, which has openly declared its intention to pull ahead in AI.

This dynamic creates a paradox: the very strategy designed to ensure American AI dominance may undermine it by ignoring alternative approaches that are more efficient, more distributed, and potentially more resilient to the geopolitical pressures that will define the next decade of technological competition.

How Can the Military Adapt Its AI Strategy?

Defense experts suggest several adjustments to ensure the military can leverage the full spectrum of AI approaches rather than betting everything on foundation models:

• Invest in Open-Weight Research: Fund development of open-weight models and alternative architectures that can operate efficiently on edge devices and in disconnected environments where troops actually operate.
• Prioritize Operational Resilience: Design AI systems that function reliably without constant cloud connectivity, enabling autonomous decision-making in contested environments where communications are degraded or jammed.
• Diversify Technology Partnerships: Support a broader ecosystem of AI developers and researchers rather than concentrating resources in a handful of frontier labs, reducing strategic vulnerability to any single company's decisions or failures.
• Align Public and Military Interests: Address public concerns about AI's environmental and social impact by investing in more efficient alternatives, which could rebuild congressional and public support for continued AI research funding.

The stakes of this strategic choice extend beyond military capability. As China continues advancing in open-weight AI and alternative architectures, the U.S. military's narrow focus on foundation models risks ceding technological ground precisely where it matters most: in systems that work reliably in the real world, not just in data centers.