← Back to stories A modern data center featuring a computer setup with monitor and keyboard, emphasizing technology infrastructure.
Photo by Brett Sayles on Pexels
凤凰科技 2026-03-08

U.S. to expand AI chip export controls globally; exports by NVIDIA, AMD and others will require licenses

Washington moves to close loopholes

It has been reported that the U.S. Commerce Department plans to widen export controls on advanced AI chips worldwide, making shipments by NVIDIA, AMD, Intel and others subject to licenses regardless of destination. The goal is to choke off indirect flows of cutting‑edge accelerators to China by tightening rules first introduced in October 2022 and expanded in 2023, which targeted performance thresholds and high‑speed interconnects. The Bureau of Industry and Security (BIS) is reportedly preparing guidance to apply a uniform global regime as circumvention via third countries persists.

What will change

The reported policy would extend licensing to chips that exceed specified compute and networking criteria and could lean on extraterritorial tools such as the Foreign Direct Product Rule—used previously against Huawei (华为) and Russia—to reach non‑U.S. firms that rely on American technology. It may also heighten scrutiny of cloud providers that deliver AI compute to restricted end users. For vendors, that means longer lead times, case‑by‑case reviews, and heavier compliance costs. For buyers, expect more uncertainty over delivery, configuration downgrades, and shifting demand toward locally sourced alternatives.

Implications for China and the supply chain

The immediate impact would be felt by Chinese customers such as Baidu (百度), Alibaba (阿里巴巴), Tencent (腾讯), ByteDance (字节跳动) and top research institutes, which have depended on NVIDIA’s H100/H200‑class GPUs and earlier China‑specific variants (A800/H800) that were already curtailed. Beijing is accelerating domestic options—most notably Huawei’s Ascend series—yet Semiconductor Manufacturing International Corporation (中芯国际) still trails the cutting edge, limiting performance and yields. Globally, memory suppliers and contract manufacturers like Taiwan Semiconductor Manufacturing Company (台积电) could see order fragmentation, while hubs in Southeast Asia and the Middle East—popular for trans‑shipment or new data centers—face tighter screening and potential delays.

The geopolitical backdrop

The move underscores Washington’s bid to slow China’s military‑AI progress and protect what it views as force‑multiplier technologies. Allies have aligned: the Netherlands has limited ASML’s most advanced lithography tools, and Japan has tightened its own export regime. China has responded with export permits for gallium, germanium and graphite—key inputs for chips and batteries. Will a global license requirement cool the world’s AI build‑out, or simply reroute demand to less capable parts and domestic substitutes? For now, companies are reportedly drafting contingency plans while awaiting the formal BIS rule text.

AISemiconductorsSpacePolicy
View original source →