Get email delivery of the Cadence blog featured here
Professor Wei Shaojun of Tsinghua University presented on China's IC Industry: Today and Tomorrow. As you probably know, in 2014 China announced plans to promote the development of the IC industry in China. One big reason for this is that China imports an enormous quantity of semiconductors, 60% of the worldwide production of semiconductors. Of course, many of these are re-exported in finished goods (for example, an Apple Ax chip manufactured in Taiwan gets imported into China and then most of those are re-exported in iPhones). I don't know if it is true today, but a couple of years ago at the Gartner Semiconductor meeting before SEMICON, one of the analysts pointed out that China imports a greater value of semiconductors than it does oil. It's a little bit of an unfair comparison, because little of the oil is re-exported, but it is a dramatic statistic nonetheless. One of the goals in this area is that China should manufacture 20% of the semiconductors and only import 80% by 2020.
The Chinese market is not homogenous. In the coastal areas, GDP/person is over $10K, in the middle of the company it is between $6-10K, and in the west it is under $6K. This means that different products fit different parts of the market. In particular, a lot of the market is extremely cost sensitive.
I will use the ¥ sign for the Chinese Yuan (元) also known as the Renminbi or RNB. One US $ is approximately ¥7.
There was rapid growth in 2016 with the IC industry in China reaching ¥433B, up 20.1% from the year before (compared to 1.1% for worldwide semiconductor). The annual growth rate from 2004 to 2016 has been 18.9%, albeit starting from a very low base.
The Chinese IC industry is divided into three sectors: fabless, manufacturing (foundry and IDM), and package & test. All three sectors were above ¥100B, fabless became the biggest sector, and manufacturing became the fastest growing.
In 2016, the manufacturing sector grew 25%, not only because of global demand and full production, but also the role of government policies. New capacity and new fabs started to produce.
The main manufacturing sites are shown on the map to the right and are:
SMIC grew by almost 40% and Shanghai Huali by over 50%. They both had 28nm in production. Note that this is polysion, not HKMG. Depending on just how you count, the most advanced process technology is a couple of generations back from the leading-edge foundries and IDMs. However, production is only 150K WPM (thousand wafer starts per month), only 15% of what is required, which is around 1M WPM.
Fabless has been growing very fast. From 1999 until 2016, the CAGR was 44.9%, although starting from almost zero. A lot of the companies are either foreign or JVs. There are 1362 fabless companies today. However, 20% of the companies take 80% of the revenue (more precisely, 161 take 75% of the revenue). Total revenue in 2016 was ¥165B.
The top two fabless semiconductor companies in China, Hisilicon (¥260B) and Uni-group (¥125B) are sixth and tenth in the global ranking of fabless semis. Much of the design is at 14/16nm (manufactured outside China, largely in Taiwan and Korea). 503 fabless semiconductor comapnies (out of the 1362) were profitable. The gross margin of the top 10 is 35%, of the top 100, 30.6%.
Electronic IC imports remain high, with a value of over ¥200B for four consecutive years. As I said in the opening paragraph, 60% of all the ICs manufactured worldwide were imported into China. Approximately 27% of them were consumed in China, and the rest were re-exported in final goods.
In general, China is at the low end of the value chain. In many markets, the share of Chinese domestic production is zero. For the more advanced products, I assume that this is at least partially because only last year did 28nm production come online, and more advanced processes are not available domestically. Only in some mobile chip segments does the share get into double digits.
The value of manufacturing is only 0.6% and the rest is the cost of the components. This is one of the big motivations for China to develop their own manufacturing.
There are 23 300mm wafer fabs under construction in China, to go with the 16 already existing ones. He said 26 at one point, so there may be a couple missing from the chart. Once they are complete, total production will be 1.11M WPM, split with 400K WPM for foundry business, and 650K WPM for memory businesses.
In the chart below, the fabs in white are existing fabs, the ones in blue are being constructed. The existing fabs have a capacity of 149K WPM and the new ones under construction over four times as much at 615K WPM. Last year was also the first year that capital investment was over $10B (and I assume that doesn't include most of the equipment for the fabs under construction yet).
All this design, especially in the fabless companies, is potentially new opportunity for EDA companies. The top 10 fabless companies in China increased their share of the fabless market from 30% to 43% between 2010 and 2016, but this is a lot less concentrated than in the US where the top 10 fabless companies are over 80% of the market. That means that many of the fabless companies are small, with 90% of them having less than 100 employees, and they struggle to afford tools.
There is some EDA development in China (not including development groups in China of companies like Cadence). Empyrean, Proplus, Platform, and a few universities are the main forces in tool development and IC design methodology research. In 2016, the China Electronics Design Automation Expert Committee was established.
Although China's IC industry has experienced rapid growth, the scale is quite small and the quality is poor. A special challenge is how to meet the market requirements while keeping the cost low, given that China is a country with a huge population, vast territory, growing rapidly, but with an unbalanced economy and diverse cultures.
The rapidly growing IC industry in China will force design and EDA engineers both inside and outside China to innovate and collaborate. There is a large talent pool.
The current state of play can be summed up in a few bullets: