US’ China policy is at a crossroads

Anyone who assumed that blocking the export of EUV scanners, and later advanced DUV immersion tools as well, would stop the Chinese from manufacturing 7nm chips should have known better. TSMC started shipping N7 chips in 2018, patterned exclusively using DUV immersion equipment (the EUV-enabled N7+ node was introduced a year later). Five years ago, Chinese foundry SMIC still had unrestricted access to the West’s semiconductor manufacturing equipment. It was therefore only a matter of time for the Chinese to create their own 7nm node.

Surprise at the performance of the so-called Kirin 9000S chipset, recently discovered in a Huawei smartphone, is more appropriate. The chip based on Arm IP is reportedly on par with 1-to-2-year-old Qualcomm products, even though the US has access to more advanced process technology. Shrinking only benefits the digital parts of a chip, however, and a SoC such as the Kirin 9000S has lots of analog/RF functionality on board as well.

“The RF side of the chip is amazing, using an integrated modem that’s on par with Qualcomm’s current best,” writes chief analyst at Semianalysis Dylan Patel, noting that the RF frontend is domestically produced, a capability many thought to be out of reach of the Chinese. It’s interesting to note that the Chinese managed to produce a great 5G modem chipset, while Apple has been working on one for five years and has nothing to show for it. The recently unveiled new Iphone models still rely on Qualcomm technology.

What to make of this? If the goal of the US export restrictions is to deny China sub-14nm capabilities, as put forward in US legal documents, the sanctions are woefully inadequate. The very ASML scanner used by TSMC to ramp 7nm production – the NXT:1980i - isn’t even covered by current restrictions. In fact, the equipment maker’s DUV order book is jam-packed with these tools. Applied Materials, Lam, KLA and other companies also continue to ship equipment that can be used for near-leading-edge semiconductor manufacturing.

If, however, the goal was to stop China’s advancement in the IC industry, judgment should be postponed. Experts agree that SMIC, with extremely tight process control, could probably squeeze a 5nm-equivalent chip from the tools it currently has at its disposal. Beyond that, the company will most likely be stuck until China develops a domestic equipment industry that can keep up with the West. That said, Huawei’s smartphone proved that the Chinese can come a long way with process technology that’s not state-of-the-art. There’s no reason to suppose that they won’t be able to create powerful AI hardware, for example.

It seems that Washington doesn’t even know what it wants; the restriction package reeks of political compromise. Hawks have been pushing for tighter sanctions, arguing that the more the economies of the US and China become intertwined, the harder it becomes to act. The industry’s army of lobbyists scream bloody murder about lost revenue, thereby – one could argue – proving the hawks’ point. The result is a porous set of rules and regulations that doesn’t accomplish a clear goal.

The launch of the Huawei smartphone could tilt the balance in favor of the hawks – new export controls may be in the making. Banning a wider range of equipment, EDA tools, materials and mask technology as well as advanced chips will hit US companies significantly harder than it has until now, however. In that respect, the US’ China policy is at a crossroads. Either it bites the bullet to meaningfully decouple from China, causing major economic damage, or it doesn’t and accepts that it won’t be able to meaningfully slow down the advance of a global power that will eventually threaten its hegemony.