Will Samsung's 2017 Semi Capex Deliver Knockout Blow to Competition?

[Daniel Nenni]

Will Samsung’s 2017 Semi Capex Deliver Knockout Blow to Competition? The company’s $26 billion in 2017 outlays expected to be more than Intel and TSMC combined.

IC Insights has revised its outlook for semiconductor industry capital spending and will present its new findings in the November Update to The McClean Report 2017, which will be released at the end of this month. IC Insights’ latest forecast now shows semiconductor industry capital spending climbing 35% this year to $90.8 billion.

After spending $11.3 billion in semiconductor capex last year, Samsung announced that its 2017 outlays for the semiconductor group are expected to more than double to $26 billion. Bill McClean, president of IC Insights stated, “In my 37 years of tracking the semiconductor industry, I have never seen such an aggressive ramp of semiconductor capital expenditures. The sheer magnitude of Samsung’s spending this year is unprecedented in the history of the semiconductor industry!”

Figure 1 shows Samsung’s capital spending outlays for its semiconductor group since 2010, the first year the company spent more than $10 billion in capex for the semiconductor segment. After spending $11.3 billion in 2016, the jump in capex expected for this year is simply amazing.

To illustrate how forceful its spending plans are, IC Insights anticipates that Samsung’s semiconductor capex of $8.6 billion in 4Q17 will represent 33% of the $26.2 billion in total semiconductor industry capital spending for this quarter. Meanwhile, the company is expected to account for about 16% of worldwide semiconductor sales in 4Q17.

IC Insights estimates that Samsung’s $26 billion in semiconductor outlays this year will be segmented as follows:

3D NAND flash: $14 billion (including an enormous ramp in capacity at its Pyeongtaek fab)

DRAM: $7 billion (for process migration and additional capacity to make up for capacity loss due to migration)

Foundry/Other: $5 billion (for ramping up 10nm process capacity)

View attachment 20710
Figure 1
​

IC Insights believes that Samsung’s massive spending outlays this year will have repercussions far into the future. One of the effects likely to occur is a period of overcapacity in the 3D NAND flash market. This overcapacity situation will not only be due to Samsung’s huge spending for 3D NAND flash, but also to its competitors in this market segment (e.g., SK Hynix, Micron, Toshiba, Intel, etc.) responding to the company’s spending surge. At some point, Samsung’s competitors will need to ramp up their capacity or loose market share.

Samsung’s current spending spree is also expected to just about kill any hopes that Chinese companies may have of becoming significant players in the 3D NAND flash or DRAM markets. As our clients have been aware of for some time, IC Insights has been extremely skeptical about the ability of new Chinese startups to compete with Samsung, SK Hynix, and Micron with regards to 3D NAND and DRAM technology. This year’s level of spending by Samsung just about guarantees that without some type of joint venture with a large existing memory suppler, new Chinese memory startups stand little chance of competing on the same level as today’s leading suppliers.

Report Details: The 2017 McClean Report
Additional details on capital spending and other trends within the IC industry are provided in The McClean Report—A Complete Analysis and Forecast of the Integrated Circuit Industry (released in January 2017). A subscription to The McClean Report includes free monthly updates from March through November (including the Mid-Year Update), and free access to subscriber-only webinars throughout the year. An individual-user license to the 2017 edition of The McClean Report is priced at $4,090 and includes an Internet access password. A multi-user worldwide corporate license is available for $7,090.

To review additional information about IC Insights’ new and existing market research reports and services please visit our website: IC Insights | Semiconductor Market Research.

A PDF version of this Research Bulletin can be downloaded from our website at http://www.icinsights.com/news/bulletins/

 

[hist78]

Who do you think will benefit most from Samsung's spending in terms of equipment sales?

 

[Arthur Hanson]

Excessive capital spending can also be a trap, if a new technology renders it obsolete before the pay back period, or it drives prices down sharply enough that it proves uneconomic to operate. Also to fully utilize equipment will take more development in a tighter time frame to make the equipment pay for itself. This raises the question does Samsung have the products developed to match the life span and capabilities of the equipment before newer technologies lower its economic value. Equipment now has a shorter and shorter economic life, unless it can be applied to totally new products that change the economic life span. I also see a vast expansion of the market in a whole array of semis, mems, sensors and power sources, so it depends on the specific equipment ordered and its range of applications. There are many, many vast untapped or underserved markets that can benefit from the latest leading edge technologies. The companies and their partners that create new markets will win. Just look at Samsung's new vacuum cleaner, a very minor product, but a step in home automation. Now look at medical where inexpensive devices of many types could sell literally in the billions. Nanotech that requires these tools is still in its infancy in technology and markets. Refer to my forum on semis and nanotech.

 

[Daniel Nenni]

Here is a note from Robert Maire on Samsung:

Samsung- The hundred year flood of capex or the new baseline?
Capex spending by Intel has been soft, spending by TSMC has been very good but Samsung has been so far off the charts that people are dumbfounded. If anybody in the industry tells you they predicted or expected Samsung's spend level they are clearly lying.

Cities don't build flood protection based on once every hundred year floods but warming and other trends may have increased the frequency of what used to be 100 year flood levels to a more common time period.

Is what we see at Samsung a spike within a "new normal" range? or is it simply a once in a lifetime, black swan event never to be repeated again in our lifetimes.

Take a look at the numbers...
Over the past five years, from 2012 to 2016 Samsung has spent an average of about $12.35B per year. It has also spent within a fairly tight range with a high of $13.5B and a low of $11.3B or a deviation of plus or minus about a billion dollars or roughly 10%.

That is until 2017 when all hell broke loose.......
Industry number crunchers are expecting about $26B in total capex spend with about $21B of that in memory and $14B on NAND alone. That $14B is more than Samsung has ever spent in total going back over the years and even blows away Intel at its peak. We are a a spend level roughly twice that of Samsung's highest year. This is the equivalent of a 100 year flood.

We are very big proponents of NAND storage and the data revolution that we are in the middle of. We think that much of the NAND capacity is for SSD's that will wipe out all rotating media. We continue to make the case that NAND is the poster child for elastic, price driven demand with a virtually unlimited thirst given all the new data hungry applications such as AR, VR, AI and IOT etc;, however, we are talking about a river of NAND.

We can convince ourselves , for the time being, that NAND spending could be within the range of a new normal given the new circumstances, and that the new normal for NAND is much higher. DRAM, however, is a different story. We continue to say that DRAM is relatively inelastic demand as compared to NAND as most applications require a finite amount. The growth is DRAM is more linked to the growth of the number of applications rather than the amount used per application which is the main NAND driver.

With these two points being made we could say that $14B in NAND could be sustainable or a peak within a new higher range but $7B for DRAM is likely outside of even a new normal range for DRAM spend.

If Samsung's old, 5 year average capex spend was $11B+- $1B could the new capex spend be $20B +- $3B of which averages NAND at $12B, DRAM $4B and foundry $4B with all three seeing above the range growth in 2017 due to a confluence of events? We think DRAM spend has the most risk but NAND while high is not crazy given demand. Foundry seems roughly in line with what we expect and could see more spend as EUV cranks up with Samsung stepping on the gas.

This of course is just our "SWAG" on the numbers which will obviously be way off in reality but its as good a start as any.

Its important to remember that this is just Samsung we are talking about. While $26B may not be long term sustainable, $20B or more is a very good consolation prize, especially coming off an $11B long term average. Samsung , by itself probably lifts the entire industry by close to a $10B a year in a spending increase step function.

Kind of like 10 year floods happening every year.........

 

[name99]

Does the money targeted at "NAND" mean NAND or persistent storage? Does Samsung believe they have (or need to find, very rapidly) something more-or-less equivalent to Optane, to sell into more-or-less the same markets. (Where these could be broadly understood, eg a ramp-up of some sort of spin-based storage that may not be as dense as Optane, but is persistent and is lower power, or lower latency.)

I don't think the bulk storage market is especially interesting for a new memory type, not yet, but the market Optane should have been in from the start (and supposedly will be in next year...) of persistent storage addressed as RAM *is* interesting; and it would make sense for Samsung to target that for the whole world that Intel will likely omit in its Optane rollout (so high-end phones, ARM-based laptops, non-x86 based servers and HPC, things like that).

 

[Arthur Hanson]

Any opinions on Automata and the ramifications for memory and processing if it comes to be. Mixing memory and processing in one package could be a game change or a fool's folly. Could Samsung, TSM, Micron and Intel all be chasing versions and variations of mixing memory and processing?
Any thoughts, views or even speculation on this would be appreciated. I think we are in the infancy of tech reaching into many areas where it hasn't been in its current form and feel demand will surprise most, but apparently from the numbers not Samsung. I follow the money for a living and the trail is getting very interesting.

 

[Fred Chen]

Does the money targeted at "NAND" mean NAND or persistent storage? Does Samsung believe they have (or need to find, very rapidly) something more-or-less equivalent to Optane, to sell into more-or-less the same markets. (Where these could be broadly understood, eg a ramp-up of some sort of spin-based storage that may not be as dense as Optane, but is persistent and is lower power, or lower latency.)

I don't think the bulk storage market is especially interesting for a new memory type, not yet, but the market Optane should have been in from the start (and supposedly will be in next year...) of persistent storage addressed as RAM *is* interesting; and it would make sense for Samsung to target that for the whole world that Intel will likely omit in its Optane rollout (so high-end phones, ARM-based laptops, non-x86 based servers and HPC, things like that).

Samsung is in fact developing Z-NAND to directly counter Optane. It is a faster, less dense form of its 3D NAND.

 

[wgt46]

​

"IC Insights believes that Samsung’s massive spending outlays this year will have repercussions far into the future. One of the effects likely to occur is a period of overcapacity in the 3D NAND flash market. This overcapacity situation will not only be due to Samsung’s huge spending for 3D NAND flash, but also to its competitors in this market segment (e.g., SK Hynix, Micron, Toshiba, Intel, etc.) responding to the company’s spending surge. At some point, Samsung’s competitors will need to ramp up their capacity or loose market share."

Wrong. Nobody, including Samsung, is focused on bit share. Here is Samsung from the recent earnings call - "[FONT=&quot]While closely monitoring market conditions,[/FONT] we will continuously improve profitability by expanding sales of differentiated products and managing a profit-focused product mix." The move to 64L will build supply (Micron is on the high end of the NAnD projections for 2018 at 50%, but they have been consistently high (overestimating the 2017 capacity increase for example). Here's Samsung (also from the Q3 earnings call) "[FONT=&quot]As for supply,[/FONT] although suppliers’ production of 64 layer may expand, supply growth is expected to be limited compared to demand growth[FONT=&quot], due to decreasing planar capacity and increasing technical difficulties."
[/FONT]
"Samsung’s current spending spree is also expected to just about kill any hopes that Chinese companies may have of becoming significant players in the 3D NAND flash or DRAM markets." Wrong again. Not because the Chinese have little to no chance of making an impact in the DRAM/NAND markets - they don't, but that is for technical and IP ownership reasons. But the argument that CapEx is an influencing factor for the Chinese is a total misreading of the Chinese initiative. The Chinese effort is a strategic projection of State power, and is not in any way an "economic" matter. Because this is so they will not be deterred by any concerns that might be expected to motivate a profit oriented company.

 

[Arthur Hanson]

I do see lower prices, but I also see lower costs as the technology progresses while seeing an very large expansion of the market, especially with IOT and advanced mobile.

 

[wgt46]

Agreed!