HGJ, Hanxin, SMIC, HiSilicon, Hongxin, the Big Fund: these few names are more or less a summary of China's semiconductor industry over the past twenty years, each one an expression of a different idea about how to develop it. Looking back at those turning points, it seems we never got out of one particular mental trap, and instead keep making the same mistake over and over.
The Bind
HGJ and Hanxin
The strength of a whole-of-nation system is that it can concentrate resources on big things. Following the logic of "independent and controllable", HGJ (核高基) set out from day one to close the loop on fully domestic chip design and manufacturing. The Hanxin (汉芯) scandal then showed in full just how badly that Great Leap Forward approach failed. The result is that after twenty-odd years, China's semiconductor industry is neither "independent" nor "controllable", and it took US sanctions before everyone realized there was not a single card left in hand. Amusingly, the Hanxin scandal dragged Loongson down with it and left it carrying the blame for twenty years. Even now, few people can tell the two apart.
From SMIC to Hongxin
Once it was clear the whole-of-nation approach was not working, bringing in outside talent and backing it with large domestic capital became the new direction. SMIC is a relatively successful example, going from zero to one over the past twenty years by continuously recruiting people from TSMC. But SMIC's heavy dependence on the global supply chain also leaves it struggling in the middle of the US-China contest, and it still has a hard time breaking through below 14nm. SMIC's model is also not easy to copy. As national investment in semiconductors grew, so did all kinds of chaos. Wuhan Hongxin, with over a hundred billion yuan behind it, copied SMIC by recruiting a TSMC veteran to front for it and even got hold of the mainland's first EUV lithography machine, and the whole thing turned out to be a scam.
Huawei and Globalization
Huawei, the first Chinese tech giant to move into international markets, understood the importance of chip independence very early. HiSilicon's growth was rough going too: from switch chips to wireless base stations to phone SoCs, HiSilicon became China's largest Fabless company and earned a chance to stand alongside the global giants. But Huawei was over-dependent on the global supply chain, and that is exactly where it ended up trapped. Because of US sanctions, its foundry access, its EDA tools and its IP were all cut off, so Huawei is now stuck using 4G chips from two years ago, and HiSilicon has had to fall back to SMIC's 14nm process.
The Big Fund's Corruption
In response to the Western technology blockade against China, the state set up the integrated circuit Big Fund. Looking at it now, more than a hundred billion yuan of that investment basically went down the drain. Worse, the Big Fund also left behind messes like Tsinghua Unigroup along with various bad debts, which over the long run will affect the state's willingness to keep investing in semiconductors at scale. Given that someone needs a face-saving way out, arresting a few corrupt officials is probably the easiest option. But you can never arrest them all, and no matter how many people take the blame, it does not change the efficiency problem of large government-led funds.
The Trap
Ren Zhengfei has made a series of points about how China can get out of the bind its semiconductor industry is in. In summary: 1) do not rush; 2) do the basic scientific research properly and patiently. I agree with both of these completely. Reading Ren's earlier remarks, there is a third point that is hard to state publicly now that Western sanctions are in place: 3) learn from the US, and find partners worldwide by every means possible.
More than any other industry, semiconductors depend on a global division of labor, because the field embodies the accumulated knowledge of all of humanity and no single country can do it alone. So "independence" does not really exist. For China, the more reliable direction for improving semiconductor competitiveness is to achieve some degree of "control". As for how to get there, personally I think there are a few directions:
- Invest more in basic science and attract high-end talent from abroad, especially ethnic Chinese scholars returning home. Against the backdrop of US-China confrontation, a large number of PhD students in basic science fields have already come back over the past two years, and that trend will not change over the long run.
- Encourage startups to develop and break through in specific advanced niches, in the hope that they can play a key role in the global supply chain.
- Open the Chinese market further, so that foreign semiconductor companies can get the same policy support domestically. The more international companies invest in manufacturing and R&D in China, the more domestic talent gets trained.
- Acquire and restructure key foreign companies in the global supply chain. The relatively successful examples of the past couple of years are the stake in ARM and the acquisition of Imagination, two IP companies, where setting up Chinese entities secured access to core IP licenses.
Closing
Seen as a cycle in human history, we are at the start of a wave of deglobalization and nationalism, most likely followed by local conflicts and a worldwide recession. The only way to break out of that downward cycle is a breakthrough in science and technology.
The bind China's semiconductor industry is in is one important expression of that cycle. As Moore's Law comes to an end, the West's most important advantage in the critical information industries, semiconductors, will cease to exist. That means an enormous redistribution of wealth, and with it a shift in the relative strength of different countries. Naturally that shift is not something today's advanced countries want to see. And as the party being suppressed, if China falls into the same "deglobalization" and "nationalism" trap, it will only miss out on this important chance at redistribution.