REGISTER | LOGIN
Breaking News
News & Analysis

Samsung’s Capex Seen Crushing Memory Startups

11/15/2017 12:01 PM EST
14 comments
NO RATINGS
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
ubm112211
User Rank
Author
Re: ...
ubm112211   11/16/2017 9:17:27 PM
NO RATINGS
for backend there are some room for AIs. Fab uh.. no. it's already highly automated. the more complicate it goes the more labor need to maintain it.   Ironically , for this time the general public is not criticizing the chinese govt for whatever reasons. the world has suffered enough the monopoly power of the memory giants.  yeah, who else can lower the memory price to a more reasonable level beside chinese govt at this moment?

Alan Patterson
User Rank
Author
Re: ...
Alan Patterson   11/16/2017 9:04:26 PM
NO RATINGS
Time for the lights-out, totally automated AI fab?

ubm112211
User Rank
Author
Re: ...
ubm112211   11/16/2017 6:46:39 PM
NO RATINGS
samsung should have trouble in finding so many labors for their new factory. this act will backfire.

resistion
User Rank
Author
Toshiba sale
resistion   11/16/2017 4:31:31 PM
NO RATINGS
The Toshiba sale is about 18 billion, has not even started to consider the capex to catch up. Toshiba only invested 1.76 billion so far. Too late for buyers beware. https://www.japantimes.co.jp/news/2017/10/24/business/corporate-business/bain-says-western-digital-legal-protest-threatens-toshiba-joint-venture/#.Wg4CSkqWY2w https://www.theregister.co.uk/2017/08/03/western_digitals_latest_toshery/

realjjj
User Rank
CEO
Re: ...Chinese Startups
realjjj   11/16/2017 2:00:12 PM
NO RATINGS
China's growth is also about management and workforce. SMIC is claiming volume prod on 14nm in 2019 and that is well behind but it is somewhat remarkable if they can do it. Ofc the business model is a bit different, technology is much more important for a foundry while for memory, cost per bit is king.Anther big difference is scale and utilization, with a foundry scale matters and you can't go from 0 to 500k WSPM overnight but it's a bit different with memory.

On patents, at least Sandisk and Spansion got some revenue from licensing, not sure about others. Your point about some markets preferring leading edge is valid but it's a very big market and China will have limited share at first. Look at phones and bit crossover to DDR4, it took a while.

Anyway, the main difference of opinion here seems to be China's ability to close the gap. You are pessimistic while I believe that high volume production, the scaling slowdown and new NVMs shifting the incumbent's focus , will enable China to close the gap.

bmccleanicinsights
User Rank
Author
Re: ...Chinese Startups
bmccleanicinsights   11/16/2017 1:13:57 PM
NO RATINGS
In general, my comparison to the 2000 timeperiod for Chinese foundries was targeted at the comparison of technology and not China's economy or financial means.  Everyone agrees, China has plenty of money and their economy is huge.  The issue still comes down to technology.  SMIC and the other Chinese foundries have had 15 years or more, including the last five years of a booming Chinese economy, to catch up or even close the gap with TSMC and have just not been able to do it.  I don't think it will be any different in the memory market.   What China's memory producers will be able to make is lagging generation parts that would only be price competitive on a per bit basis if they sell below cost.  This would be similar to what the Japanese DRAM producers did in the 1980s to gain a foothold in the market. This was followed by numerous trade hearings and litigation between the U.S. and Japan (which I was involved in).  Moreover, in our opinion, it may be near impossible for new Chinese memory startups to make even lagging-edge DRAM or 3D NAND without stepping on numerous Samsung, SK Hynix, or Micron patents.  The current memory leaders leave each other alone with regard to patents because each company has their own arsenal.  This will not be the case with the Chinese memory startups.  The typical plan here, let the Chinese startups make and sell their memory devices for about a year and then drop the lawsuit bomb.  Should be interesting!   Making lagging-edge parts addresses only so much of the market.  With the drive toward smaller, thinner, more portable systems, many electronic system producers will not want to meet their memory needs with 3 Chinese DRAM devices instead of 1 Samsung device.  Not saying there wouldn't be a market for lagging-edge Chinese memory, its just that it isn't that large.   Yes, we are looking at the long term bit growth for memory in the ranges you mentioned.

realjjj
User Rank
CEO
Re: ...Chinese Startups
realjjj   11/16/2017 12:42:12 PM
NO RATINGS
Hi, thanks for the clarifications.

The comparison with year 2000 is less than ideal, where was China's economy back then, their GDP was under 1 trillion and it is maybe around 12 now. Their chances of success are much higher today as their economy and society have evolved.

I do have a question , how do you see blended bit cost per year evolving in the next few years for incumbents - at 20%, 30%, a bit more , a bit less. I am asking per year not per new node and blended so overall output not new node vs previous node. On the NAND side there is also QLC that will be a substantial part of the cost reduction and if China manages to have 64L QLC in 2020, blended cost differences might not be that large.

And then ofc there is price, supply vs demand and where the industry is in the next 5 years. Excluding China, are you assuming bit growth at 35-45% for NAND and 20-25% for DRAM or higher?  

bmccleanicinsights
User Rank
Author
Re: ...Chinese Startups
bmccleanicinsights   11/16/2017 11:54:30 AM
NO RATINGS
I would like to clarify some of IC Insights' opinions expressed in the article on Samsung's capex. When mentioning that Samsung's enormous level of capex will "kill" the hopes of the startups it was meant to mean kill the hopes of the startups of becoming major memory players.  It does not mean it will kill the "attempt" of the Chinese companies at becoming major suppliers.  When "business" becomes driven by nationalistic motives, any economic reasoning goes out the window, and this will be especially true for the Chinese startups.  New Chinese memory companies can lose billions for years and not blink.  However, this does not mean that they will ever be major suppliers. Money is not really the issue here for the Chinese, they have plenty of it.  The stumbling block that we believe they will fail to overcome is the technology.  Without a joint venture with an existing major memory supplier, the Chinese memory companies will lag woefully behind the current leaders for at least the next 10 years, or maybe until a completely new memory technology that replaces the current NAND and DRAM architectures levels the playing field. A good example of where IC Insights envisions the future of the Chinese memory producers can be seen in history of the Chinese producers in the IC foundry industry.  In the early 2000's China's government had a plan to establish major Chinese players in the IC foundry business.  The Chinese government did all it could to help these companies, like SMIC, get established.  They were going to take on TSMC, UMC, etc. head on. Flash forward to 2017.  SMIC is an excellent foundry, has been very profitable recently, and is China's largest player in the foundry market.  SMIC has been in business for 17 years and is still about 4-5 generations behind TSMC in technology.  SMIC also has only a 6% share of the pure-play foundry market.  In fact, in total, the Chinese companies have a 9% share of the pure-play foundry market. So yes, the Chinese startups are likely to spend billions of dollars attempting to gain a foothold in the memory market and have a long term outlook that does not include being profitable.  However, similar to the foundry market, the Chinese memory producers are also likely to continue to lag woefully behind the current leading memory producers in technology and have less than a 10% share of the memory market even 10-15 years from now. Bill McClean President, IC Insights

HangLai
User Rank
Author
Re: ...
HangLai   11/16/2017 11:09:53 AM
NO RATINGS
I agree.  Those people focus on present or near term profit will be in for a surprise.  Chinese firms will be competitive in a few years; Korean is no match in this fight.  They do not have the resources in finance nor human factor for a long term fight.  Samsung invested more factory facilities now only to drive down the ASP, to hurt themselves and others in a short term.

realjjj
User Rank
CEO
Re: ...
realjjj   11/16/2017 10:09:24 AM
NO RATINGS
Hey, IC Insights seems to be looking at China the same way you look at a publicly traded company but from China's perspective, the fundamental goal is to capture more of the BOM and memory is a large part of that. They'll be behind at first and maybe lose some money for a few years but that's a given. The assumption that China needs to compete "on the same level" anytime soon is false and this move does not "kill any hope". China needs to get the ball rolling and then they'll work hard to close the gap. This is not about making a profit from day one and it's not a failure if they lose 2 billion while Samsung makes 20,  this is a long term effort. it matters where they are in 2025 and 2040 not 2020.

Another notable aspect is that CAPEX is not a big problem for China, it's more of a problem for others, especially if prices fall and they are in the red for long periods of time. At the same time, scaling has slowed down, a new node doesn't give you 50% lower costs so a 2 year gap is not what it used to be , cost wise.

Samsung is less afraid to engage than others, we've seen that in displays in the last decade. China came from behind there too and Samsung did well to compete but that did not stop China.

How this plays out, the incumbents target higher margin segments, then transition more and more bits to new NVMs, maybe some can pivot to monolithic 3D neuromorphic and become the next Intel or Qualcomm or maybe foundries take a huge chunk of their business with embedded 3D NVMs. China will do its best to keep up and close the gap, I would keep a close eye on what China does in the new NVMs area.

BTW no article on SMIC's results this time around?

Page 1 / 2   >   >>
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed