The geography of artificial intelligence is shaped not only by research talent and electricity, but by who can buy advanced accelerators and the equipment needed to manufacture them. Export controls have turned those transactions into instruments of national security policy. Their effects extend beyond blocked shipments: they influence data-center locations, supply-chain due diligence, domestic subsidies, product design, and incentives to develop substitutes.

The central policy argument is that leading AI chips and high-bandwidth memory can support military intelligence, weapons development, surveillance, and other strategic capabilities. The countervailing concern is that controls can reduce suppliers' revenue, impose compliance costs on third countries, and accelerate indigenous alternatives. The result is a moving boundary rather than a permanent technological embargo.

The controlled object keeps expanding

Early public debate often focused on a few graphics processors. The U.S. regime has grown to cover a broader stack: advanced computing integrated circuits, servers containing them, high-bandwidth memory, semiconductor-manufacturing equipment, certain software, end uses, and named entities.

In December 2024, the U.S. Bureau of Industry and Security announced controls on 24 types of semiconductor-manufacturing equipment, three types of software tools, and certain high-bandwidth memory. It also added 140 entities to the Entity List and modified 14 entries. The BIS announcement described tools for etching, deposition, lithography, ion implantation, metrology, inspection, and cleaning—evidence that manufacturing capability, not finished chips alone, is the target.

Rules can also follow activity across borders. BIS guidance issued in May 2025 said licenses may be required when a provider knows U.S.-controlled chips will be used by an overseas infrastructure service to train models for parties headquartered in China or other D:5 destinations. It also addressed certain support by U.S. persons. The policy statement makes end user, end use, and knowledge central compliance questions even when hardware is physically located elsewhere.

Policy volatility becomes a business condition

The controls have changed repeatedly. In January 2025, the United States issued an AI Diffusion Rule that would have created country-based authorizations and limits for advanced chips and certain model weights. In May, before its principal compliance requirements took effect, the Commerce Department announced that it would rescind the framework and develop a replacement while issuing new chip guidance. That rescission notice illustrates why companies cannot treat a one-time classification as durable strategy.

In January 2026, BIS revised its policy again, allowing license applications for Nvidia H200, AMD MI325X, and comparable products to be reviewed case by case for approved Chinese customers. Applicants must show, among other conditions, that exports will not reduce capacity available to U.S. customers, that buyers use compliance procedures, and that products undergo independent U.S. performance and security testing. The agency's stated principle was: “Export controls should evolve with changes in technology, while protecting national security.” The January rule summary is a reminder that controlled performance tiers and licensing presumptions can move in both directions.

For chipmakers, cloud providers, and buyers, this means product road maps must account for regulation. A chip designed just below a threshold may serve a restricted market until the rule changes. Data-center operators need screening, recordkeeping, physical security, and controls on remote access. Customers face uncertainty over upgrades and replacement parts.

The map is allied, not simply bilateral

AI Chip Export Controls Are Redrawing the Technology Map — supporting image
AI Chip Export Controls Are Redrawing the Technology Map — supporting image

No country controls every essential step. Advanced chips combine U.S. design tools and architectures, manufacturing concentrated in East Asia, Dutch lithography, Japanese materials and equipment, and global packaging capacity. Effective restrictions therefore depend on coordination, while differences create opportunities for diversion or substitution.

The Netherlands introduced national authorization requirements for specified advanced lithography equipment in September 2023, expanded them in September 2024, and added specified measuring and inspection equipment from April 2025. The Dutch government stresses that applications are assessed case by case rather than under a blanket export ban. Its 2025 notice also links national responsibility to the country's unusually strong position in manufacturing equipment.

At the same time, industrial policy is pulling production toward favored jurisdictions. The U.S. Commerce Department finalized up to $6.6 billion in direct CHIPS Act funding for TSMC Arizona in 2024, supporting more than $65 billion in planned investment across three fabs. The award announcement ties supply security to domestic capacity. Export controls restrict one set of flows; subsidies and tax incentives create another.

Controls reshape incentives, not just access

The near-term effect is segmentation. Firms may offer different products by destination, place compute in approved regions, or use licensed cloud access instead of hardware sales. Countries seeking dependable access have stronger reasons to negotiate trusted-partner status, host compliant data centers, and fund domestic packaging or fabrication.

Longer term, outcomes are less certain. Controls can slow access to frontier equipment, but they also raise the return on alternative chips, software optimization, smuggling, stockpiling, and local manufacturing. Broad rules may burden allied firms; narrow rules may be easier to route around. Measuring effectiveness therefore requires more than counting denied licenses. Policymakers need evidence about capability delays, diversion, allied alignment, supplier revenue, and the pace of substitution.

The technology map is being redrawn through a combination of licensing borders and investment corridors. The durable question is whether policy can preserve a meaningful security advantage without fragmenting the supply network so severely that it weakens the innovation base it is intended to protect.

Sources