The secret supply chain: How TSMC chips ended up in Huawei’s latest AI tech

The semiconductor industry, driven by the race for technological dominance, has become an arena for geopolitical competition.

Huawei, one of China’s leading technology giants, has been accused of using advanced chips produced by the Taiwan Semiconductor Manufacturing Company (TSMC) in its latest AI technology, despite US sanctions intended to prevent such dealings.

The revelation came to light last week after reports suggested that Huawei sidestepped US trade restrictions by working with third-party companies, such as China-based chip designer Sophgo and cryptocurrency mining equipment company Bitmain.

While both companies deny the allegations, corporate registration records show that Micree Zhan—who co-founded Sophgo—was also involved in the founding of Bitmain.

Following the US sanctions, China was promoting self-reliance in its semiconductor industry.

As part of these efforts, last month Huawei began testing the Ascend 910C chip as a potential competitor to Nvidia’s H100.

However, Huawei's recent use of TSMC chips, despite the trade bans, has reignited debate over China's narrative of technological independence and raised concerns about the enforcement of global trade restrictions.

As reported by Bloomberg, Huawei’s reliance on TSMC chips output may reflect China's ongoing struggle to manufacture advanced chips in sufficient quantities and further dent the Asian giant's ability to produce high-end chips domestically.

‘Backdoor tactics’

According to Tony Loughran, a risk analyst and director at Zero Risk International, the use of third-party suppliers to bypass sanctions—commonly known as "backdooring"—is not a new tactic.

"Backdooring, or third-party sourcing, is a method that has been employed across nations for years," Loughran tells TRT World.

He illustrates the concept with a historical example: "During both the Iraq wars, despite an export ban on oil, countries managed to buy and sell oil on the black market. Where there is high demand, there’s often a lucrative pipeline and someone willing to create a workaround.”

A similar pattern appears to be unfolding with Huawei.

Its Ascend 910B processors are likely to be manufactured using TSMC’s advanced 7-nanometer technology, and Bloomberg reports that at least one of these chips was discovered in a teardown of Huawei’s latest AI offerings.

Given Huawei's blacklisting in August 2020, it has been largely relying on domestic manufacturers like Semiconductor Manufacturing International Co. (SMIC) for chip production.

However, the poor yields reported for AI chips produced by SMIC—only 20 percent of their output operates as intended—highlight the limitations of China's semiconductor industry and may have possibly forced Huawei’s hand to source chips elsewhere.

How it works

The process of acquiring restricted technology through indirect channels often involves third-party businesses that officially claim to serve legitimate customers, while discreetly redirecting resources to embargoed entities.

“This typically involves a third-party business claiming to serve legitimate clients, while gradually redirecting batches to embargoed nations.” risk specialist Tony Loughran explains.

In Huawei's case, the chips could have been routed through multiple suppliers before ending up in Huawei's hands. This convoluted supply chain allows for the circumventing of restrictions without directly implicating the involved manufacturers.

Loughran warns of the risks associated with such practices: “The major concern is that if semiconductor chips can bypass restrictions this easily, there should be even greater vigilance over sensitive products in high-risk areas, such as advanced weapons and defence systems.”

Drawing a parallel, Loughran likens it to a case involving Israel’s pager attack.

“This scenario resembles the recent case of Taiwan's pagers being sold to a third-party contractor, who then modified them to target Hezbollah.”

His comparison displays the potential consequences of weakly regulated tech supply chains.

Struggle for self-reliance?

China’s self-reliance ambitions have been a cornerstone of its tech policy, with President Xi Jinping emphasising the need for technological independence in light of rising tensions with the West.

However, Huawei's reliance on TSMC for its latest AI processors indicates that these ambitions are still far from reality.

The reports paint a clear picture of Huawei's challenges in producing its own chips at a commercially viable scale.

After the Biden administration introduced strict export controls, TSMC and other suppliers were barred from selling products or services to Huawei without a US Department of Commerce export licence, as their operations rely heavily on American technologies.

In response, Huawei was forced to localise production for nearly all its components—a costly process, potentially running into hundreds of millions or even billions of dollars, and requiring several years to complete.

The limited yields of SMIC's 7nm chips illustrate the difficulty in achieving the level of quality control required to compete with TSMC.

The path forward

As the debate over Huawei's access to advanced semiconductors continues, Loughran suggests proactive measures that governments and tech companies could take to prevent future violations of this nature.

"Technology companies should engage independent auditors to verify the integrity of their supply chains, especially in areas subject to sanctions or requiring higher security compliance," he notes.

This measure could help maintain transparency and prevent the type of backdoor transactions that appear to have taken place in the Huawei-TSMC case.

Additionally, Loughran mentions the role of the United Nations in regulating such activities, emphasising the importance of international cooperation.

"The role of the United Nations and potential sanctions should also be examined to validate the current US stance," he says.

Such an approach could help create a global framework for monitoring and enforcing sanctions, reducing the likelihood of future breaches.