Silicon Security in the New Era of AI, Export Controls, and U.S.-China Rivalry.

上午 9:40 - 上午 10:05

The October 7, 2022 export controls were treated at the time as a Biden-era departure. Three and a half years later, they have been tightened (December 2024), carried forward by the second Trump administration, joined by Japan and the Netherlands, and met by a Chinese indigenization push that has reshaped — but not broken — the regime they were designed to defend.

This presentation argues that the architecture has proven more durable than its critics predicted, but that its long-term effectiveness now turns on three questions the next two years will answer. First, can Washington and its allies close the smuggling and diversion gap that has allowed restricted chips to reach Chinese AI data centers? Second, will the allied coalition — the United States, Japan, the Netherlands, South Korea, and Taiwan — hold together through a period of trade pressure, leadership transitions, and divergent commercial interests? And third, will the Trump second administration be consistent enough in the policy design and implementation? 

The talk works through the evidence on each question and lays out the planning assumptions that semiconductor executives, equipment suppliers, OSATs, and the investors funding the AI build-out should be carrying into 2027 and beyond.

 

Key Technologies Covered

  • AI chip export controls (Nvidia H100/H200/Blackwell-class GPUs and successors) 

  • Semiconductor manufacturing equipment controls: EUV and advanced DUV lithography 

  • High-Bandwidth Memory (HBM) and advanced packaging restrictions 

  • The Foreign Direct Product Rule (FDPR) and its 2024 expansion 

  • Entity List enforcement and the new Restricted Fabrication Facility (RFF) license category 

  • Allied export-control architecture: Japan, the Netherlands, South Korea, Taiwan 

  • AI chip smuggling: scale, routes, and enforcement gaps 

  • China's indigenization push: SMIC, Huawei Ascend, and domestic SME (Naura, AMEC, SMEE) 

  • DeepSeek and AI-compute efficiency: implications for compute-threshold control logic 

Featured Speakers

Mr. Gregory Allen

Mr. Gregory Allen

Founder & CEO, Decision Tree Research

Gregory C. Allen is the founder and CEO of Decision Tree Research, an independent research firm focused on the geopolitics of strategic technologies. For more than a decade he has led work shaping U.S. and allied policy on artificial intelligence, semiconductors, and defense tech.

As founding director of the Wadhwani AI Center at the Center for Strategic and International Studies (CSIS), Mr. Allen authored the influential reports "Choking Off China's Access to the Future of AI" (2022) and "The True Impact of Allied Export Controls" (2024), described by The Economist as "very much at the center of the formulation of current U.S. policy." Earlier, he was Director of Strategy and Policy at the U.S. Department of Defense's Joint Artificial Intelligence Center, overseeing the DoD's AI strategy. He has testified multiple times before the U.S. Congress on AI and the semiconductor supply chain.

Mr. Allen hosts The AI Power Podcast and writes Decision Tree, an evidence-based newsletter on AI, semiconductors, and defense tech. He holds a joint MPP from the Harvard Kennedy School and MBA from Harvard Business School.