TSMC Admits Chip Supply Chain Lacks Downstream Tracking

TSMC´s latest report reveals it cannot trace how its chips are ultimately used, spotlighting challenges in enforcing semiconductor export restrictions for Artificial Intelligence hardware.

TSMC has publicly acknowledged significant limitations in its ability to track the end use of its semiconductor products, as detailed in its recent annual report. The company stated it ´inherently lacks visibility regarding the downstream use or user of final products,´ highlighting a gap in oversight that has come under scrutiny following incidents involving sensitive chip shipments to restricted entities. Specifically, the report addresses how 7 nanometer chips originally manufactured for Sophgo were eventually found in Huawei´s Ascend 910B/C Artificial Intelligence accelerators, despite those devices being subject to stringent US export controls.

According to TSMC, the standard chip fabrication workflow involves taking design files (GDS files) from clients, mostly via intermediaries, checking for technical compliance, crafting photomasks, and producing the actual wafers, all without context on where or how the chips are ultimately deployed. This process, while typical for foundry operations, means that once chips leave TSMC, there is little visibility until—if ever—they are analyzed in finished products. Subsequent investigations revealed these chips comprised the heart of more than one million dual-chiplet Artificial Intelligence accelerators, translating to two million dies delivered to Huawei, an entity facing US sanctions.

The report cautions that should supply-chain partners neglect to secure proper import, export, or re-export licenses, TSMC itself could face significant regulatory action, penalties, or investigations—even if it follows all current protocols. The US has already floated a possible billion-dollar fine in connection with these compliance issues. As international sanctions, particularly from the US targeting Chinese technology companies, become more rigorous, foundries like TSMC are under escalating pressure to improve end-user tracking. Yet, the complex, layered nature of the global semiconductor supply chain renders comprehensive oversight challenging and may ultimately lead firms to reconsider their dealings with Chinese clients to sidestep regulatory risks.

81

Impact Score

Hybrid Web3 strategies for the artificial intelligence era

Enterprises are starting to blend Web2 infrastructure with decentralized Web3 technologies to cut costs, improve resilience, and support artificial intelligence workloads, while navigating persistent interoperability, regulatory, and user experience challenges.

Artificial Intelligence, chips, and robots set the tone at CES 2026

CES 2026 in Las Vegas put Artificial Intelligence at the center of nearly every major announcement, with chipmakers and robotics firms using the show to preview their next wave of platforms and humanoid systems. Nvidia, AMD, Intel, Qualcomm, Google, Samsung, Hyundai, and Boston Dynamics all leaned on Artificial Intelligence to anchor their product strategies.

Inside the UK’s artificial intelligence security institute

The UK’s artificial intelligence security institute has found that popular frontier models can be jailbroken at scale, exposing reliability gaps and security risks for governments and regulated industries that rely on trusted vendors.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.