Chips & PolicyLOT 261122 · 3 min read
TSMC halts China-bound shipments as U.S. rules flag decade-old 16nm and 14nm nodes
TSMC has paused shipments to Chinese and non-Chinese designers after a new BIS rule classifies 16nm and 14nm chips as advanced computing parts until proven otherwise.
Key takeaways
- TSMC has paused shipments to Chinese and non-Chinese designers pending compliance reviews under a new BIS interim final rule.
- The rule presumes chips at 16nm or 14nm — nodes about 10 years old — are export-controlled advanced computing chips.
- Chips with fewer than 30 billion transistors and no HBM cannot be assumed exempt from high-performance computing classification.
- Export licenses are now required for advanced AI chips shipped to Indonesia, Malaysia, Singapore, Brazil and Thailand.
- Sources cite smuggling of 1.5 million Nvidia AI GPUs, worth roughly $7.5 billion, as a driver of the crackdown.

TSMC has put multiple chip shipments on hold — including orders from Chinese designers — after a new U.S. export rule classified chips made on decade-old 16nm and 14nm process nodes as presumptively export-controlled advanced computing products. Nikkei Asia first reported the pause, which reaches far beyond the data-center AI chips Washington originally targeted.
The U.S. Bureau of Industry and Security (BIS) published the new controls as an interim final rule (IFR). Its stated aim is to stop Chinese AI companies from leasing compute capacity outside China to reach advanced U.S. chips such as Nvidia's Blackwell AI GPUs. In practice, the rule also drags consumer and automotive silicon into the licensing net.
What does the rule actually change?
Previous U.S. export controls concentrated on high-performance AI processors, including Nvidia's Blackwell and the upcoming Vera Rubin generation. The new IFR expands the scope to all chips and shifts the compliance burden onto the supply chain itself.
Under the rule, front-end foundries and OSAT companies — the outsourced assembly, packaging and test firms — must treat any semiconductor at the 16nm or 14nm node, or anything more advanced, as an export-controlled advanced computing chip by default. The rule closes an obvious loophole: even a chip with fewer than 30 billion transistors and no HBM memory cannot be assumed to be free of high-performance computing use.
The practical consequence is documentation. A designer building on 16nm or newer — even for a smartphone, television or car rather than a data center — must now obtain certification from an approved designer covering the chip's Total Processing Performance or performance density before the fab will release the order without extra scrutiny.
How wide is the blast radius?
The 16nm and 14nm nodes are roughly ten years old and ship in enormous volumes inside consumer devices and vehicles. That makes the new presumption a compliance event for a broad slice of the industry, not a niche restriction on AI accelerators.
A source told Nikkei Asia: "The new rule is very strict this time, and the affecting scope is far beyond only chips for data centers but covers smartphone processors, autonomous driving and other types of chips for consumer electronics as well as others if they are using nodes that are 14nm or 16nm or more advanced tech." The source added: "The move would delay shipments to many chip developers worldwide before they can clear compliance and hit many emerging startups in many Asian nations."
In short, the pause is not limited to TSMC's Chinese customers; designers elsewhere also face holds until their paperwork clears.
Why the smuggling crackdown intensified
The IFR adds another measure: exporters now need licenses to ship advanced AI chips to Indonesia, Malaysia, Singapore, Brazil and Thailand. The rule identifies these countries as frequent transshipment points for chips diverted to China, citing the Supermicro and Megaspeed investigations. In those cases, Nvidia servers were routed through one of these destinations before moving on to China.
Enforcement pressure had already been building. Sources say the real reason behind Jensen Huang's departure from Nvidia was the smuggling of roughly 1.5 million Nvidia AI GPUs that occurred under his watch, beyond the reported relationship with Megaspeed executive Alice Huang. She is suspected of helping divert $7.5 billion worth of Nvidia chips and of passing information to Chinese intelligence.
The takeaway for the supply chain is simple: until TSMC and other foundries clear compliance reviews, lead times for mainstream-node chips will stretch, and the cost of proving a chip is ordinary now falls on the designer.
Tom's Hardware (Source)
More from Tom Whitfield
Show full bio
Staff writer covering companies, deals and strategy in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
1,200 articles