Leap Frogging with China Semis

We have learned in recent years that predicting anything is largely pointless. That being said, we are fairly certain that we will see an immense amount of semiconductor and hardware (aka Deep Tech) venture funding in China this year. And while we are searching for an authoritative aggregator of this, we spent some time looking at activity in this space in late December. This is largely an anecdotal exercise, but just that thin sliver of activity turned out to be highly informative.

There are really two topics to study in regards to China’s semis. The big one, which everyone seems to focus on, is China’s quest for semiconductor “self-reliance” – can China build up its own semiconductor fabs for manufacturing chips? This is such a big topic because it has important geopolitical implications. But here at D2D, we like to focus on more prosaic, base commercial topics. This focused us on the growth of China’s fabless chip design companies. With the world obsessed with supply chain questions, the focus is on the first, manufacturing question. But we would argue that the second topic is also incredibly important. China has several thousand chip design companies today (or is it 10,000?). Most of these will fade away, but we would argue strongly that a handful will emerge as major, global players. The fun part will be figuring out in the next few years who will comprise that small handful.

From even this quick glance we can discern some patterns. First is the sheer quantity. There have been two to three funding announcements for chip and hardware companies every week that we have been following this. We do not have a great data source aggregating Chinese venture capital activity but our guess is that semis investing in China is roughly 10x the number of deals as the US. Probably more.

Second is the diversity of end markets these companies are attacking. There are a lot of AI-related chip companies in China. AI is among the most loaded words in tech right now. It means different things to different people, and of course has taken on geo-political overtones. That being said, much of the work of building “AI chips” is fairly mundane. As we frequently point out AI chips are fairly simple, built for doing a lot of very specific math around matrix multiplication. The Chinese companies that are emerging around this space generally fit one of three profiles: AI accelerators (aka TPUs or NPUs) intended for use in data centers by webscale companies; AI for automotive uses of which there are many, not all related to autonomous driving; and AI for other uses – of which the majority is video recognition and manipulation. There are a huge number of AI chip design companies in China today across all these domains.

Another common thread is chips for “Industrial uses”, a large number of which are some form of analog chips. Much of the time when we speak of semiconductors, we are taking about the high volume, high profile chips for mobile phones, PCs and data centers. These tend to move in well understood and highly scrutinized ways (e.g. Apple’s annual iPhone updates and AMD vs. Intel data center CPU share). Industrial chips, especially analog, move differently. In particular, these chips tend to be much lower volumes, millions of units a year as opposed to hundreds of millions, but this is offset by generally higher prices and gross margins. Sales cycles are also very different. The makers of industrial machinery of all sorts spend years designing their products and take their time to select chip suppliers. These designs then typically get locked down for years. So a key differentiator in the space is the ability to provide engineering and design support to get those design wins. This is particularly important in China which has a growing field of industrial equipment makers. Local chip companies can support these customers in a much more localized fashion than the global players, very broadly speaking. So it is no surprise that Chinese chip companies are emerging to serve these customers.

Nowhere is this more true than in automotive. China has gotten to the point that it has so many electronic vehicle (EV) makers that the government is starting to “encourage” consolidation for fear of the industry self-swamping itself so early in its growth. A large subset of chip designers in China are emerging to serve these customers. Their products cover the range of automotive semis from low-price microcontrollers and actuators, to power management chips (crucial for EVs) and all the way up to infotainment automotive application processors and full-blown compute for autonomous vehicles.

Through all this an important theme is apparent. Each of these areas is essentially greenfield. While there are a small number of Chinese GPU and CPU makers, the vast majority of companies getting funded in China today are leapfrogging or maneuvering around markets with entrenched, consolidated competition. In the case of AI that means entirely new markets, for Industrial and Automotive that means new customers many of whom are only a few years old themselves. As much as the press and policy makers focus on high-profile areas of competition, the real story is here.

For China’s 1,000+ fabless chip design companies the real test for their viability will ultimately be their ability to compete outside of China. Most of those 1,000 will never get there, but many will, and our guess is that most of that small group will come from companies who cut their teeth in these green fields and use that to bootstrap their way to larger, more competitive product portfolios.

Leave a Reply