CONNECT WITH US

US directive halts TSMC's sub-7nm supply to China; Samsung, SMIC may fill gap amid AI funding and capacity strain

Staff reporter, Taipei; Levi Li, DIGITIMES Asia 0

Credit: AFP

On November 11, 2024, TSMC reportedly suspended its sub-7nm process services to China's AI and GPU chipmakers following the Huawei "chip controversy," triggering widespread industry impacts and potential order transfers. The company has already issued notifications to affected clients regarding the cessation.

US halts AI chip exports to China

The US Commerce Department issued an "is informed" directive to TSMC, as reported by Reuters on November 10, halting advanced sub-7nm chip exports to China starting November 11. This measure aligns with the US strategy to limit China's access to high-end semiconductor technology for national security reasons.

The restriction, which targets AI applications, may impact TSMC's operations in China, where demand for high-performance AI chips continues to grow. TSMC has affirmed its commitment to regulatory compliance but has not yet publicly commented on the order.

This latest directive marks another step in Washington's effort to restrict technologies with dual-use potential, building on similar measures previously imposed on other leading chipmakers like Nvidia and AMD.

Industry ripples

Industry analysts are closely monitoring the effects of TSMC's decision on global supply chains and its China strategy. Chinese firms are likely to shift orders to SMIC and Samsung, with Samsung positioned to benefit in the short term.

TSMC's decision has strained its relationship with Chinese clients, with the Huawei controversy amplifying supply chain concerns. The company's halt of 7nm GPU and AI chip supply has led to many of its clients being "caught in the crossfire," creating uncertainty for AI-driven chip companies focused on autonomous driving and cloud computing.

Following this development, Chinese clients with advanced process needs are expected to turn to SMIC and Samsung for these services.

Samsung: a viable alternative?

With TSMC unable to accept orders, Samsung also faces export restrictions, leaving Chinese foundries as the main option, according to OFweek. However, questions remain about their technology and capacity to meet demand.

US export controls impact not only American firms but also foreign companies like Samsung that use US-origin technology, potentially limiting Samsung's ability to supply advanced chips to Chinese customers, as per Financial Times.

The Korean chipmaker's cautious approach reflects broader compliance concerns, possibly limiting its ability to expand production for Chinese clients, Reuters notes.

An SMIC takeover?

SMIC's advanced process platform, overseen by Co-CEO Mong-song Liang, remains somewhat opaque. SMIC no longer reports sub-7nm processes separately in revenue, and Liang's absence from earnings calls adds to the enigmatic nature surrounding the company.

The launch of Huawei's Mate 60 and the upcoming Mate 70, both using HiSilicon Kirin chips, has put SMIC's 7nm process technology back under scrutiny.

Observers note that limitations in SMIC's EUV lithography force reliance on complex multiple-patterning processes, restricting output. Meeting demand for Kirin chips alone strains capacity, and with TSMC stepping back from China's 7nm market, SMIC's ability to absorb this demand is limited.

At SMIC's November 8 earnings call, Co-CEO Haijun Zhao addressed questions on AI innovation, its impact on China's semiconductor supply chain, and SMIC's advanced packaging plans. Zhao's cautious response emphasized that SMIC would meet customer needs within its platform's capabilities.

Zhao described SMIC as an "indirect beneficiary" of the AI wave, noting that as other firms allocate capacity to AI, some customers face shortages and may turn to SMIC as a second source.

Chinese IC clients report that new US restrictions on AI investments are constraining funding for Chinese AI chip firms, affecting R&D and expansion. With TSMC halting sub-7nm supply, China's AI chip sector faces a dual squeeze on funding and production capacity.