Well, the South Korean regulators are letting Samsung build sub-20-nm chips in China. U.S. regulators won't let Intel build anything below 65-nm. So it seems like some regulators understand the situation a little better than others.
it's not clear that any Chinese company has the equipment and the process capability to steal it. In fact, it's pretty clear that they don't.
I concur, but the people imposing the export restrictions won't get that. They're politicians and bureaucrats making and issuing rules for political reasons. There are any number of examples of government regulation where it's obvious those involved simply don't understand what they are regulating, and are blissfully unaware that things don't work that way and the regulations won't do what they intend. Electronics and semi-conductor manufacturing are only one place this is true.
Access to the latest technology is also an issue for the Chinese foundries.
Yep. That was part of what I was thinking of when I made my comment about whether a foundry has the process technologies in place to be able to make what I need.
The reason for this is the increasing cost and complexity with each process generation.
Do you know if anyone has tracked the increases in costs at progressively smaller geometries?I'd guess the increases become progressively larger as the geometries get smaller, but haven't seen stats on the rate of increase.
When Intel announced a fab in China, they were only allowed to transfer technology that was several generations old to the facility.
Intel was complying with US export restrictions, but I suspect other countries will have a different perception of the risks. (Like, say, Samsung.)
China cannot become a manufacturing powerhouse unless it can either access the latest technology or develop it completely within the country.
I don't see the first happening for some time. I'm not sure the second will ever happen. It's the old saw about tools and tools to make the tools. Even if they somehow did, there's the question of whether the market is big enough to support them along with everyone else.
PS- Bill McClean also made another point that I thought was pretty interesting. He said that even if the IP around Samsung's sub 20-nm process technoloy is compromised, it's not clear that any Chinese company has the equipment and the process capability to steal it. In fact, it's pretty clear that they don't. So McClean's point is, they are so far behind that maybe it doesn't really matter that much. Still, I am sure Samsung and the South Korean government wouldn't want to see that happen.
If I'm not mistaken, Uncle Sam said Intel couldn't build chips in China at geometries lower than 65-nm for the reason that you state, Jim, IP security. According to Bill McClean, the Samsung fab will be closely watched because it really is the first fab in China that will be using cutting edge process technology.
Access to the latest technology is also an issue for the Chinese foundries. Almost all of the semiconductor manufacturers around the wold are working together in one form or another to develop new process technology. The largest and broadest such efforts are the IBM development alliances. The reason for this is the increasing cost and complexity with each process generation. In fact, Intel is the only semiconductor manufacturer that continues to develop its process completely independently. But, even then, many of the technial developments are shared at the equipment vendor level.
However, there continues to be a huge distrust with Chinese firms to honor intellectual property rights, which is also backed by restrictions by many governments limiting the transfer of technology to China. When Intel announced a fab in China, they were only allowed to transfer technology that was several generations old to the facility. China cannot become a manufacturing powerhouse unless it can either access the latest technology or develop it completely within the country.
I think a lot of people thought a few years ago that SMIC and other Chinese foundries would rise to prominence because of the lower labor costs, sheer number of engineers, and the Chinese government's commitment to build a chip industry.
And I can see why, but such predictions failed to understand where the costs were. Lower labor costs and engineering talent pool weren't the critical factors. Chinese government support was, but it would still be an uphill battle.
China was trying to establish itself as a chip maker in a crowded market. If I'm looking for a foundry to make chips, my first question will be "Can this candidate successfully make the chips I need, in sufficient volume and to sufficient standards of quality?" Depending upon what sort of chips I need, candidates may be ruled out because they don't have the required process technologies in place.
Being competitive in the foundry business reminds me of the Red Queen's Race from Alice in Wonderland: you must run as fast as you can to simply stay in the same place. Getting anywhere requires you to run faster. If you have a fab, you are looking at continuing major investmentto be able to produce new generations of chips at ever smaller process geometries. The ongoing investment, over time, will be as great as the initial investment to build the fab.
If you don't get a steady stream of volume manufacturing commitments to support and justify that investment, you aren't long for this world.
There's a likely local Chinese market for chips than can support someone ike SMIC (though I doubt SMIC is anywhere near profitable yet - it's there because the Chinese government thinks there should be a Chinese foundry.) SMIC's fortunes mirror China itself: China is no longer the low cost producer growing its economy on exports, and is looking at further growth from growing and serving its own internal markets. SMIC's longer term future probably depends on the success of those efforts.
Making any significant gains outside of China seems unlikely.
Agreed. I don't think there are many Chinese chip companies that can afford or even want or need their own fab. The fabless model works well for them, just like it does for so many companies in the U.S. and elsewhere. It's a good model.
But I think a lot of people thought Chinese companies would be bigger players in the foundry business. I think a lot of people thought a few years ago that SMIC and other Chinese foundries would rise to prominence because of the lower labor costs, sheer number of engineers, and the Chinese government's commitment to build a chip industry. SMIC has been largely unsuccessful to date, but it is gaining traction and increasingly building chips for the fabless firms in China. It has a solid niche. But it's not going to catch TSMC in terms of leading edge process technology, and thus will never dominate the market like many thought it could.
The problem for China becoming a chip manufacturing power is cost,
The low labor costs aren't a factor. Those low costs helped China gain prominence in the sort of manufacturing that required product assembly, but aren't applicable to fabs. Turning out many trained engineers isn't a major factor, either.
The problem is that semi-conductor fabs are so enormously expensive to build, and as process geometries shrink, costs rise even more. And the majority of the equipment in fabs that actually makes the chips must be sourced from outside China. The infrastructure to make it locally doesn't exist.
We are seeing increasing numbers of companies go fabless, and some that still have fabs enter into joint ventures because the costs of building them are so great. How many Chinese companies can afford to build one?
Even if they can, how many Chinese companies need their own fab? As fab costs rise, the stakes get higher. The fab really needs to be running 24/7 turning out chips to cover its overhead, which requires a steady stream of volume design wins.
China's growing pool of engineering talent may make it a power in chip design, but manufacture is another matter.
SMIC's recent success with Chinese fabless customers owes a lot to the rise of smart card chips and image sensors in China -- designed by Chinese fabless, as I understand it.
As I wrote earlier this year:
Where SMIC continues to hammer, though, is what the company describes as “differentiation strategy.” At a time when SMIC’s industry rivals -- TSMC and GlobalFoundries -- are fiercely competing for the advanced digital logic sector, SMIC is trying to carve out market share in selected differentiated technologies such as embedded non-volatile memory (eNVM), CMOS image sensors (CIS) and Power Management IC (PMIC).
A Book For All Reasons Bernard Cole1 Comment Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...