At the SEMI Strategic Materials Conference, SMC, Lung Chu, who heads up SEMI China (and worked at Cadence at one point in his career) talked about The Rise of China IC Industry: Challenges and Opportunities . There are a number of amazing facts about the China semiconductor industry, ranging from semiconductors being a bigger import than oil into China, to 24 fabs being under construction in China, to the semiconductor market growth being negative if you take China out. Facts like these mean that what happens in China is probably the most significant thing in the semiconductor industry. For some recent updates on China, see my post SEMICON: China, China, China or my post from DAC China's IC Industry: Today and Tomorrow , or the SEMICON panel session SEMICON China Panel: Faster Than You Think . China Growth Rates The above chart shows the growth rates between 2010 and 2014. Worldwide, the semiconductor industry grew 4% CAGR over the period. But the world outside China didn't grow at all. However, the growth rate for semiconductors consumed in China by multinational corporations was 9% and by local companies higher still at 13%. As a result, over the period, China grew from 32% of the worldwide semiconductor market to 41%. If you look into the details a little, it turns out that most of the growth is strengthening prices in memory, especially DRAM. The chart above shows the growing share of the worldwide market taken by China, forecast out to 2021 (the numbers are from a different source from the first graph, hence the couple of percent difference in numbers). China Strategic Plan I won't go over the details of China's plan to become more self-sufficient in semiconductors. They are investing heavily, with the goal of having 20% of consumption being manufactured in China by 2020, and getting to 16nm by then. The big picture is in the chart above. If you look at history, electronics manufacturing was in US, Europe, and Japan. It then moved to South East Asia and subsequently to Taiwan and China. Semiconductor manufacturing was in US and Europe, then transitioned to Japan, and then to Korea and Taiwan (where it is, mostly, today). The big question is whether China is next. I am somewhat skeptical, especially on a 2020 timeframe, since most of their planned technology is not leading edge. The biggest bet is probably XMC (see my post, Memory in China: XMC ), which is a plan to create a new veritcal NAND flash memory company from scratch, using the discontinuity of the switch from planar to vertical NAND as an opportunity. I also think getting to 16nm, which means FinFET presumably, on that timeframe is simply not going to happen. It is only three years away. The big foundries can ramp leading-edge processes in a few months after years of experience. I just don't see that happening in China from a standing start. The other strategic part of China's investment was intended to be in M&A. China has much higher P/E ratios (higher stock prices) so their equity is cheap as a currency, plus the Chinese government(s) are making billions available. But the US, in particular, is pushing back and the current feeling is that CFIUS will reject any significant acquisition of US technology, just as it already did with Lattice (and if you don't know what CFIUS is, then read my post The Four Ts: Trade, Tax, Talent, and Technology Funding ). So the M&A aspect is largely on hold for now, at least. The long-term perspective for China is that they want to transition from low-cost manufacturing, to being a value-added solution provider, to innovation. For about 10 years now, there has been a big investment in education, so there are lots of qualified new graduate engineers, and there are also lots of people who have been in the industry over that same period, and so now have ten years of experience and can take on management roles. Like in the US 20 years ago, the smartest students are going into engineering. I don't think everything will happen by 2020 as planned, but you take this longer term view, then I don't see any reason to doubt that they will get there eventually (barring some major geopolitical event or huge economic event like a debt crisis). Fabs The above chart shows the plans worldwide for new fabs (high-volume IC fabs). There has been a huge ramp in the last three years in China (and, to a lesser extent, Taiwan). That is forecast to moderate going forward. But there are actually 24 fabs under construction in China. Some are 300mm but a lot are 200mm. But the situation in China with fabs is like states in the US with Hollywood movies—every state wants one and they all throw money at it. Lung said that, despite being the head of SEMI China, he has started to tell states and municipalities that call on him, "Please don't build a fab." Even though every fab is great for the equipment manufacturing industry, he is worried that they can't all be successful, and could provoke a price war. There needs to be more investment in design, to create the product to fill all the planned fabs. Fabless Most design in China is fabless. There are no restrictions on being able to use leading-edge processes at the big foundries, especially in Taiwan. The above chart shows the top 10 fabless semiconductor companies in China. The top two, HiSilicon and Spreadtrum, are #7 and #10 globally. The next tier are a lot smaller. Most key technologies that go into electronics that are assembled in China are still not available from domestic suppliers (some are available from China-based fabs of multi-nationals like Samsung or Intel), in particular, DRAM, NAND flash, server, and PC microprocessors. The one area where there is some penetration is the application and communication processors for mobile, at around 20% available from domestic sources (that is the heart of HiSilicon's and Spreadtrum's businesses). SEMICON China If you want to know more, then join 70,000 other semiconductor professionals at SEMICON China in Shanghai next March. SEMICON West seems pretty big to me, but SEMICON China is nearly three times the size. Sign up for Sunday Brunch, the weekly Breakfast Bytes email.![]()
↧