Is AI Containment for National Security Possible?

KEY POINTS

• The Interim Final Rule on Artificial Intelligence Diffusion (“The Rule”) suffers from semantic disconnects that demonstrate a fundamental lack of grounding of policy assumptions in AI technology and related industry/ecosystem.
• Policymakers and business leaders should be cognizant of unintended consequences associated with the application of The Rule especially in regard to AI technological leadership and global market leadership.
• The Rule is part of a continuum of protectionist tech trade and national security policies originating from the first Trump Administration and will likely be preserved in some substantial form with the incoming Trump Administration.

On January 13th of 2025, the Biden-Harris Administration released the Interim Final Rule on Artificial Intelligence Diffusion (“The Rule”) to “streamline licensing hurdles for both large and small chip orders, bolster U.S. AI leadership, and provide clarity to allied and partner nations about how they can benefit from AI.” The rule, which is slated to go into effect in 120 days, well into the second Trump Administration in just about a week.

The rule proposes to foster “responsible diffusion of U.S. technology” with six mechanisms (link):?

• No restrictions apply to chip sales to 18 key allies and partners.
• Chip orders with collective computation power up to roughly 1,700 advanced GPUs do not require a license and do not count against national chip caps.
• Entities that meet high security and trust standards and are headquartered in close allies and partners can obtain highly trusted “Universal Verified End User” (UVEU) status.
• Entities that meet the same security requirements and are headquartered in any destination that is not a country of concern can apply for “National Verified End User” status, enabling them to purchase computational power equivalent to up to 320,000 advanced GPUs over the next two years.
• Non-VEU entities located outside of close allies can still purchase large amounts of computational power, up to the equivalent of 50,000 advanced GPUs per country.
• Government-to-government arrangements cultivate an international ecosystem of shared values regarding the development, deployment, and use of AI.

Allies and partners of the U.S. included in the no restriction list include Taiwan,?Australia, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, the Republic of Korea, Spain, Sweden, and the United Kingdom.

Primary countries of concern include China, Russia, and Iran.

ANALYSIS?

The way that The Rule reads raises some concerns about the grounding of its assumptions and definitions of AI and supercomputing concepts as it relates to AI applications, the role and position of foundation/frontier models, and the bearing of “chips” as a unit of control and constraint for the stated purpose of the rule – national security.?

SEMANTIC DISCONNECTS

Firstly, the “chips” (GPU) that go into these AI supercomputing systems (or superclusters) are having the least influence on the highly fluid definition and standard of “scaling” of these systems and are bound by Moore’s Law or the perceived death of it.

High-bandwidth memory (HBM), Interconnect and networking including inter-datacenter interconnect are playing a much bigger role especially for superclusters larger than 100,000 GPUs for which power, I/O and memory walls tend to be the bottlenecks for the next level of system performance or any variety of stacked “scaling laws.”

According to 英伟达 , a rack-scale GB200 NVL72 has tens of thousands if not hundreds of thousands of parts including GPU, CPU, DPUs, switches, and represents the basic unit of GenAI supercomputing. The GPU, or “chip” in Washington’s parlance, is but one of many important components of an AI training supercomputing system.

To confuse matters even more, Jensen Huang often refers to an entire Blackwell supercluster that might be comprised of tens of thousands of physical GPUs as a “GPU” capable of running as a massive virtual GPU.?

Yes, throughly confusing.

MISCONCEPTION OF AI DEVELOPMENT VS AI CHIPS

(See full neXt Curve report)

A MISGUIDED UNIT OF AI CURRENCY – THE CHIP

(See full neXt Curve report)

OBVIOUS GPUAAS CLOUD OMISSION

(See full neXt Curve report)

THE DILEMMA OF CHINESE INGENUITY

(See full neXt Curve report)

UNINTENTIONAL ABETTING OF CHINESE AI LEADERSHIP

(See full neXt Curve report)

THE INEVITABILITY OF MUTUAL RETRIBUTION

(See full neXt Curve report)

RECOMMENDATIONS?

The securing the future of AI and securing technological and market leadership in AI are important for the U.S. and its allies. This point is not up to debate especially in an increasingly contentious geopolitical environment that is the world today.

In many ways, the AI chip industry is becoming a victim of the enormous GenAI hype. The Rule has many statements that put into question its grounding in the realities and nature of the semiconductor industry and “AI”.?

In principle, The Rule seems to struggle articulating an effective set of measures to meet its stated objectives but with a confidence you would expect from an LLM (link).

• Continue to ensure that advanced semiconductors sold abroad are not used by countries of concern to train advanced AI systems, while still permitting access for general-purpose applications from telecommunications to banking.
• Restrict the transfer to non-trusted actors of the model weights for advanced closed-weight models.
• Setting security standards to protect the weights of advanced closed-weight AI models.

PREPARE TO LIVE WITH THE RULE

(See full neXt Curve report)

PREPARE FOR ACCELERATED CHINA + BRICS COMPETITION

(See full neXt Curve report)

REFRENCES:

• “FACT SHEET: Ensuring U.S. Security and Economic Strength in the Age of Artificial Intelligence”(January 13, 2025), The White House
• “DeepSeek-V3 Technical Report” (December 27, 2025) by DeepSeek AI , Aixin Liu, Bei Feng, et al., arXiv
• “Intel issues revenue warning after US revokes Huawei export licenses — further efforts to restrict China’s access to AI chips” (May 8, 2024) by Anton Shilov , Tom's Hardware
• “NVIDIA Statement on the Biden Administration’s Misguided ‘AI Diffusion’ Rule” (January 13, 2025) by Ned Finkle, NVIDIA
• “China tells firms to avoid Nvidia chips” (September 30, 2024) by 彭博资讯 , Taipei Times

This material may not be copied, reproduced, or modified in whole or in part for any purpose except with express written permission or license from an authorized representative of?neXt Curve. In addition to such written permission or license to copy, reproduce, or modify this document in whole or part, an acknowledgement of the authors of the document and all applicable portions of the copyright notice must be clearly referenced.

? 2025 neXt Curve. All Rights Reserved