Get email delivery of the Cadence blog featured here
Yesterday I covered the first half of the CASPA meeting last Saturday about memory in China. That was the big picture.
Simon Yang, XMC
The meeting actually opened with Simon Yang, the CEO of XMC, giving a little bit of history. But the bulk of the information about XMC came at the end of the afternoon, so I've put it all together here in one place.
XMC has had a bit of a checkered history, originally under the name XInxin and a partial subsidiary of SMIC. SInce 2013 it has been fully independent. Their big bet is on 3D NAND. They have developed this jointly with Cypress (was Spansion before Cypress acquired them) and on the back of a lot of licenses and IBM patents. They have invested $200M in this and have demonstrated their first 3D test chip. But the big money is coming from the Chinese government (mainly). XMC asked for $15B, but they were told to take $24B on the basis that if they were going to be serious about being a world leader then they needed to match the world leaders' investment. They are building a fab in Wuhan (central China) with eventual capacity for 300M wafers per month (wpm).
Michael Chen, XMC
The final speaker was Michael Chen, who is the CBO of XMC, talking about Memory, the Turning Point of China's Semiconductor Industry. China is coming up fast. For example, in 2009 only HiSilicon (a subsidiary of Huawei) was in the top 50 in semiconductors and now there are 9 companies. If you live in the SoC world, it is easy to think that the whole world is fabless companies and foundries. But once you include memory, foundries are 18% and 72% is IDM. But in China there is less than 1% IDM. So China is a latecomer.
But sometimes latecomers have an advantage, especially with tool usage since they haven't invested in assets that get stranded. But China needs a whole food chain, which is largely outside: manufacturing equipment, EDA, key IPs such as processors and DDR. The reality today is that in the bill of materials (BOM) cost for smartphones, nearly 60% is IC-related but none is from China. The same in servers, with no domestic providers of the ICs.
China needs to build up a whole ecosystem, not just ICs. Michael talked about how the top 20 of Silicon Valley consist of computer/system (Apple, HP), networking (Cisco, Juniper, NetApp), semi (Intel, AMD, SanDisk), software (Oracle, Symantec, VMWare, Adobe), Internet (Google, eBay, Yahoo, Facebook), equipment (Applied, Agilent). It is the inter-relationship of all these companies that forms a mutually reinforcing business ecosystem.
Some of the shift in ecosystem can be seen in the forecasts for the smartphone market (in units, not dollars). The above chart shows the transition. In blue is the total phones sold in China. In red, the sales of local Chinese brands (so the difference is primarily Apple and Samsung). In green are the local Chinese brands sold in the rest of the world, going from next to nothing and forecast to be almost as big as the number of phones sold domestically in China.
But memory is central to much of this, as it is to so many markets. Another big one is automobile, where China is already the #1 manufacturer and the #1 consumer. With the coming switch to ADAS and autonomous vehicles, the semiconductor content and the memory content are only going to increase.
Today memory is 22% of the total semiconductor market at nearly $80B, but it is expected to reach $114B in 2019 when it will be 30% of the total market. That's a good reason for XMC to place a big bet in 3D flash.
But it won't happen overnight. Simon, in the Q&A, said that they will lose a large amount of money for 3-5 years before they reach the break-even point. But the move from 2D to 3D makes this a good transition point and the right moment to enter the market. He was also asked whether they were vulnerable to fabless competitors, but he pointed out that in memory, manufacturing and design are too inter-related, plus the value chain doesn't really support two companies' margins.
When he was asked how he was going to spend $24B, his answer was that a lot would be for equipment so companies like Applied Materials would be happy. But headcount needs to grow 10X in the next three to four years, so they are almost building a city to accommodate everyone.
One final slide from Tammy. It's all about money. The slide above shows the percentage of capex by region (Asia Pacific is non-Japan Asia). The orange goes from next to nothing in1985 to over half the investment last year. Furthermore, to put China's $25B investment in XMC into perspective, last year the total capex investment worldwide was $62B. Of course, the $25B investment is not in one year, but even so it is very significant. If China fails to establish a 3D NAND flash memory supplier in XMC, it won't be for lack of investment.
Previous: Memory, the Turning Point of China Semiconductor
Next: "Interoperability is the Only Way to Prove Standards Compliance"