Two interesting articles came out over the weekend looking at Huawei and the US Government’s attempts to hobble it.
First, some context. The US Government has issued a set of rules that seem to make it impossible for China’s telecom equipment giant to make its own chips or buy chips from any company using any American technology (which is effectively everyone). The rules are complicated, and there is a sense among many that there are many loopholes. So it is important to keep in mind there are two fronts in this fight – suppliers and Huawei’s own chips.
The first article comes from the Financial Times ($$$) which claims to outline plans for Huawei to develop chips without tripping over US sanctions. Put simply, it sounds like Huawei is helping a foundry in Shanghai to accelerate its capabilities. The key to this is that even with a lot of funding and help form Huawei, this foundry will only be able to reach 28nm processes by the end of 2021, and 20nm by 2022. Today it sounds like it is only operating at 45nm. So two years from now, Huawei’s fab will only be about 10 years behind state of the art processes.
This seems to be possible because the Shanghai foundry is largely buying used equipment, or other gear, that is beyond licensing terms and thus free from US sanction. This is a mixed blessing for Huawei. On the one hand, losing access to leading edge fabs likely means they cannot compete in handsets. The flagship smartphones, like Huawei’s Honor line, need the latest chips to be competitive. On the other hand, this foundry is probably enough to save Huawei’s core telecom equipment business. These are powered devices, with lots of analog circuitry, so leading edge is much less important. Set aside the likely quality issues and the larger power bills each base station will require, these processes will probably be good enough for most of Huawei’s products.
There are still going to be considerable vulnerabilities. Huawei is still going to need things like TCAM (specialized memory for networking) and FPGAs (programmable chips used heavily in telecom gear), adding to the concern China likely has over AMD buying Xilinx. But Huawei can possibly still access these products on the open market. Non-custom, catalog parts are both difficult for the US government to monitor and possibly (likely) covered by waivers the US Government has already granted to US vendors.
The biggest vulnerability is software tools used to design chips – commonly referred to as EDA tools. There is a very small number of these suppliers and they are heavily dependent on US Technology (as defined by the US Government). And they are absolutely required for chip design.
Which brings us the second article we want to draw to your attention. This paper from John Hopkins by Doug Fuller a professor at City University of Hong Kong. Fuller has tremendous insight into China’s semis complex, highly recommend a Twitter follow.
He similarly argues that is will take many years for Huawei to build a non-US supply chain, but it is possible. He rightly highlights the challenges around EDA tools. By his analysis, the end result of this will only be to strengthen non-US EDA tool makers. This is not a huge group today, but when faced with survival Huawei is likely to find a way to make this path work.
His argument is much more in-depth and covers all this territory and more. We recommend reading the piece. But we want to close with the comments he closes his introduction because we think they pretty accurately sum up our perspective on this whole situation.
The conclusion recommends an alternative American approach to technological competition with China that is focused on reinforcing our semiconductor capabilities instead of trying to tear down China’s.