#semiconductor

On July 10, 2026, China's Ministry of Commerce and General Administration of Customs announced a temporary ban on exports of helium (HS code 2804290010), citing the Foreign Trade Law of the People's Republic of China. This decision could significantly tighten global helium supply, which is already strained due to Middle East conflicts, affecting critical industries such as semiconductor manufacturing, medical imaging, and scientific research. The ban takes immediate effect, and any future adjustments will be announced separately. Helium is essential for cooling extreme ultraviolet lithography (EUV) machines and as a carrier gas in wafer processing.

TSMC's Arizona Fab 21 has begun mass production of NVIDIA Blackwell wafers using the custom 4NP process, but these wafers must be shipped to Taiwan for CoWoS-L advanced packaging. This highlights a critical gap in the US semiconductor supply chain: while advanced logic fabrication is now possible domestically, the US still lacks high-volume advanced packaging and HBM memory capabilities, creating ongoing reliance on Taiwanese facilities and delaying full supply chain independence until at least 2028-2029. The 4NP process is a customized 4nm-class node for NVIDIA, and CoWoS-L combines chip-on-wafer-on-substrate with an RDL interposer and local silicon interconnect. Amkor, TSMC, and SK Hynix are building packaging and HBM capacity in the US, but these facilities are not yet operational.

Anthropic has started developing its own AI chips and is in early negotiations with Samsung Electronics for manufacturing, aiming to gain more control over the compute infrastructure for its Claude model. This move signals Anthropic's strategic push toward vertical integration in AI hardware, reducing reliance on external chip suppliers like Google and Amazon. If successful, it could enhance performance and cost-efficiency for Claude, similar to OpenAI's chip efforts. The project is still in early stages, and Anthropic may ultimately decide to only purchase AI chips instead. Custom AI chip development is extremely costly, with R&D potentially exceeding $500 million.