A Primer On U.S. Semiconductor Export Controls and Entity Bans
Date of Report: January 15, 2024
For decades, U.S. semiconductor policy operated under a largely market-driven, ad-hoc framework, reflecting the ideals of globalism, free-trade, and cheap shit. Those days are over. Rising technological competition, aging populations, inflation, and a global pandemic, among other factors outside the scope of this primer exposed the increasingly hot cold war between the U.S. and China. Semiconductors, and the AI supply chain as a whole, emerged as the key factor in the race for global hegemony and AI Sovereignty.
You can’t understand the US-China Cold War or AI without a primer on the backdrop: semiconductor export controls and entity bans.
For principles, executives, and investors it is important to outline this history because the recent restrictions and compliance measures are not temporary, nor are they the last. This is the new reality for AI and global commerce.
The Reactive Beginnings: Targeting Entities (Pre-2022)
By the mid-2010s, China’s ambitions in artificial intelligence had grown increasingly mature. While the U.S. had long maintained export controls for sensitive technologies, these efforts initially targeted specific entities rather than broader technological categories.
(You can read my About Me section to learn my first exposure to the shadow war on semiconductors)
In August 2018, the Export Control Reform Act (ECRA) marked a turning point. This legislation granted the president sweeping powers to regulate the export of dual-use technologies, focusing on those deemed critical to national security. Unlike earlier measures, ECRA was designed to be enduring, requiring the establishment of processes to identify and control emerging technologies.
By 2020, these controls began to sharpen their focus. The addition of Semiconductor Manufacturing International Corporation (SMIC) to the Entity List exemplified this shift. SMIC is a key player in China’s semiconductor ambitions. It was targeted for its connections to China’s military-industrial complex. This move demonstrated the U.S.'s growing recognition of the strategic role of semiconductors in national security.
SMIC Overview
- Largest and most advanced semiconductor foundry in China.
- Offers IC manufacturing at nodes from 0.35 microns to 14nm; working on 7nm and 5nm. (TSMC can now produce sub 7nm density in its new fab in Arizona).
- Mass producing 14nm FinFET, progressing toward 7nm despite estimated yield challenges.
- U.S. export restrictions limit access to advanced EUV lithography tools (See EUV Lithography Primer and ASML Audit).
- Operates fabs in Shanghai, Beijing, Tianjin, and Shenzhen.
- Clients include Huawei (on entity list).
- Identified by the U.S. as linked to China's military-civil fusion strategy.
Yet these early efforts were largely reactive, responding to perceived threats as they emerged. While effective in curbing specific entities' access to U.S. technology, this piecemeal approach proved insufficient to address the broader challenge of China’s technological rise, particularly with its diaspora of companies and shell companies smuggling restricted chips and chip manufacturing equipment.
The Shift to a Proactive Strategy: 2022–2023
In 2022, a fundamental shift occurred. Recognizing that reactive measures alone could not contain the strategic risks posed by China’s technological advancements, the U.S. adopted a more proactive and expansive approach. This phase was characterized by a focus on the broader semiconductor technology stack. The AI supply chain became the battleground.
A pivotal moment came on October 7, 2022, when the Bureau of Industry and Security (BIS) introduced sweeping new export controls. These measures aimed to restrict China's access to advanced AI chips and tools essential for semiconductor manufacturing. For the first time, restrictions extended to U.S. persons, prohibiting them from assisting in Chinese semiconductor manufacturing efforts—even for technologies not directly covered by export controls.
The impact of these measures was far-reaching. Companies around the world faced new compliance challenges, and China's ability to access advanced chips was significantly curtailed.
A New Policy Of AI Containment: 2024–Present
By 2024, semiconductor export controls had entered a new phase of escalation. These policies also introduced the concept of Foreign Direct Product Rules (FDPR), which extended U.S. jurisdiction over foreign-produced items that incorporated U.S. technology. This marked a significant expansion of the U.S.'s ability to enforce its controls beyond its borders, effectively kicking off a new containment policy for AI.
In May 2024, the federal government proposed a rule to ban federal purchases of semiconductor products from adversarial nations. Later that year, the BIS issued additional rules targeting quantum computing and cryocooling systems. These controls were accompanied by updates to the Entity List, which added more than 140 entities linked to efforts to circumvent U.S. export controls or support China’s military objectives.
Perhaps most strikingly, these measures began to explicitly address emerging concerns about AI as the future platform driving global hegemony. In January 2025, the U.S. unveiled restrictions not only on AI-capable chips but also on AI model weights—an acknowledgment of the growing strategic importance of algorithms and data in addition to hardware.
The U.S. now has a multifaceted strategyof AI containment. The most recent framework announced on January 13, 2025 draws the battleground for this policy.
Escalation of Export Controls on Advanced AI
On January 13, 2025, the BIS implemented an unprecendted, but predictable, set of controls on “advanced computing integrated circuits (ICs) and artificial intelligence (AI) model weights.” These regulations represent the most comprehensive restrictions to date on the “global diffusion of cutting-edge AI and computing technologies.” As outlined in the BIS’s newly published interim final rule, these measures aim to tighten control over the export of the most advanced AI technologies, reflecting the U.S.'s AI containment strategy.
A New Framework for AI Control
The new controls introduced by BIS expand upon previous measures by incorporating advanced AI model weights under the newly designated Export Control Classification Number (ECCN) 4E091. These weights are the numerical parameters optimized during the training of AI models, enabling their functionality. Unlike model architectures, which can often be reverse-engineered, model weights are highly proprietary (and the result of millions to now billions of dollars in training cost for the largest LLMs). The regulation requires a license to export or transfer the weights of any AI model trained on more than 10261026 computational operations. This threshold captures only the most advanced, dual-use AI models capable of sophisticated applications such as military simulations or large-scale surveillance.
Additionally, BIS has revised its licensing requirements for advanced computing ICs (ECCNs 3A090 and 4A090), imposing a global licensing requirement and expanding the scope of the Foreign Direct Product Rule (FDPR) to regulate IC exports to nearly all global destinations. These controls are complemented by new exceptions to facilitate secure transfers to “validated end-users” in allied countries. The policy also establishes a tiered allocation system for advanced computing clusters to prevent the unregulated proliferation of computational power.
The tiered system is important to understand the policy of AI containment. It essentially divides all countries into three camps: allies (Tier 1), enemies (Tier 3), and and pawns (Tier 2). Much like the role that countries in Latin America, Africa, Middle East, and Southeast Asia played in the first Cold War, these countries will now have a series of sticks and carrots to align to the U.S. AI ecosystem.
The evolutionn U.S. export controls from an ad-hoc framework to the foundational strategy for the US-China Cold War reflects a that start of a new global re-ordering. While the controls aim to secure critical technologies, they also pose challenges for U.S. companies dependent on global markets. Firms navigating these restrictions face increased compliance costs and potential limitations on their ability to compete internationally.
Investors who largely ignored this evolution and setup extensive investment vehicles and capital allocation in China have recently found themselevs in rough waters. Executives who rely extensively on access to China’s markets or supply chains now face an uncertain future.
Additionally, the export controls have prompted countermeasures from China beyond the scope of this primer. China’s response is escalting and forming the same cohesion outlined above. Where the post-Cold War era was marked by technological interdependence, the second Cold War reflects a return to strategic competition.
Prepare accordingly.