Last week, the US government announced major new semiconductor export restrictions against China.
This was a long-predicted response to what the US sees as an onslaught of Chinese economic warfare over recent decades, and a way to prevent China’s military from becoming too advanced. China, unsurprisingly, views it as an unprovoked attack.
So, what are the main restrictions, and how will it affect western semiconductor companies trying to do business here in China?
The US government released two key documents:
The first adds 31 Chinese companies to the ‘unverified list’ and removes nine as their credentials have been approved. The second sets new restrictions on the export of advanced computing products, semiconductor manufacturing equipment, and other parts of the semiconductor supply chain.
The unverified list (UVL)
The ‘UVL’ is essentially a directory of Chinese companies for whose products the US government can’t determine the end use – typically because it can’t contact the companies, or they won’t co-operate.
The US Bureau of Industry & Security (BIS) can now effectively exclude UVL companies from global semiconductor technologies when it wants – by putting them on the now-famous ‘entity list’. Huawei is on this list, and it has hurt the corporation significantly.
The ‘Savings Clause’ in the document seems to indicate that UVL companies cannot receive US exports after 7 November 2022, suggesting they’ll be completely cut off unless they comply with US government audits to confirm end use and so get removed from the list.
The UVL contains large semiconductor companies, government institutes and universities. Noteworthy additions include fab equipment maker Naura and NAND memory manufacturer YMTC.
Some smaller fab equipment companies have also been added, suggesting this could be a new target area for US restrictions, though there are still many such Chinese companies not included.
The problem facing China now is that the US can add any company to this list if auditors aren’t able to confirm end use, which the Chinese government normally won’t allow. It must cooperate with BIS or see its companies get cut off.
That said, as mentioned above, nine companies have just been removed from the UVL - presumably because they complied, and end use was confirmed and accepted.
This part of the new restrictions covers the equipment used to manufacture chips, as well as AI and supercomputing chips themselves. It says its objective is to restrict China’s military development.
The controls cover any AI chip with both “an aggregate bidirectional transfer rate over all inputs and outputs of 600 Gbyte/s” and “a bit length per operation multiplied by processing performance measured in TOPS, aggregated over all processor units, of 4800 or more”.
This means major products from companies like America’s Nvidia, the UK’s Graphcore and China’s Biren would be covered.
The document also lists 28 Chinese companies on a supercomputing entity list. The rules suggest Nvidia chips going back five years may be deemed too powerful to be sold to these companies.
Definition of ‘direct product’
The real kicker, however, is the definition of ‘direct product’ – the term applied to semiconductor items made outside the US which include American input. While restrictions used to apply to non-US products that included a certain amount of American technology – 10% or 25% depending on the product – it appears that the definition of the term has now been expanded to apply to items that have any American input, including software.
So, a Taiwanese fab like TSMC or Korean group like Samsung won’t be able to work with a Chinese fabless company designing high-end chips, because they use American tools and equipment. Overseas design services companies are also unlikely to be able to work with Chinese firms, creating a problem for China’s AI and GPU chip design industry.
The rules also expand on previous photolithography equipment bans to include any equipment that can be employed in 16nm or less processes which use non-planar technology. This means no FinFET or GAAT for China. There are also bans on NAND equipment which can do 128 layers or more, and DRAM equipment that can do 18nm half pitch or better.
These pose real difficulties for YMTC (NAND) and Changxin Memory (DRAM) here in China. It could halt their expansion plans and leave them with products far behind global competitors. Perhaps YMTC’s CEO, who stepped down recently, saw this trouble looming?
It is hard to understand how these rules will be enforced, though. Older logic equipment that isn’t designed for 16nm or less could be used for it, just as we saw SMIC create 7nm chips using DUV equipment. Does that mean all DUV is now banned?
NAND equipment for 64 layer can also do 128 layer, so does that mean YMTC won’t even be allowed to do 64 layer in case it uses the same equipment for 128?
And will export licences be handed out for exemptions left, right and centre? Or will the US be stricter this time?
Some exceptions include global companies in China with HQs in the US or allied countries, which will be reviewed on a case-by-case rather than be denied by default. This is likely to help with the CHIPS 4 Alliance which South Korea was unsure about, as both SKHynix and Samsung have fabs in China.
The new regulations are long and complex, with a great number of restrictions and exceptions. From feedback I’m hearing on the ground here, Chinese companies are worried about what this means for their expansion plans - or whether they can even get a product out if they lose access to overseas fabs like TSMC, UMC, GF, and Samsung.
Some are even worrying that they’ll have to choose between their US passport and their company as the rules dictate that US citizens can’t support Chinese semiconductor firms, many of which were founded by US returnees who still hold US passports.
On the flip side, this is a plus for our non-US clients that don’t come under the direct product rule: we’re seeing renewed interest in them from Chinese companies.
And that’s not to say there’s no interest in US technologies anymore. Quite the opposite: Chinese companies want them if they can get them.
Depending on the company, if they’re cut off, they’ll either have to shut up shop, halt expansion plans, find other overseas options or use local alternatives. But most local alternatives – if they exist at all – lag far behind.
It will take a while to see how this all plays out and where the new boundaries lie.
It’s doubtless that the tighter rules will have an impact on China’s semiconductor industry and prospects in the country for western firms. But China has surprised us many times before and could well do so again.
One thing’s for sure: if new opportunities emerge, they’ll need careful navigation. What’s more, the world will still need semiconductors; and if these rules mean capacity expansion in China becomes difficult, this expansion will have to go elsewhere – presenting a potential win to some other countries in the region.
To discuss the situation for your semiconductor business in China, you can contact Stewart at email@example.com