Quarterly Year-over-Year Growth Slows Substantially for IC Market

[Daniel Nenni]

Analysis shows decline from 23% year-over-year growth in 1Q18 to 6% forecast for 4Q18.

IC Insights November Update to The 2018 McClean Report will present an in-depth analysis and detailed five-year forecast for the IC Industry, which is expected to enter a period of cyclical cooling after an extended period of very strong growth.

Figure 1 illustrates the worldwide quarterly year-over-year IC market increases from 1Q through 3Q and IC Insights forecast for 4Q of this year. As shown, the first half of 2018 started out with strong quarterly year-over-year growth for the IC market. However, year-over-year IC market growth dropped to 14% in 3Q. Moreover, with the softening of the memory market, IC Insights projects that year-over-year IC market growth in 4Q will be only 6%.

Figure 1
​

Third quarter sequential growth confirms the slowing year-over-year trend. In 2017, 3Q/2Q IC market growth was 11%. This year, 3Q/2Q growth slowed to a 6% increase (the same rate as the long term average). As mentioned, the softening memory market has started to become a headwind on total IC market growth. It is interesting that in 2017, the 3Q/2Q memory market growth rate was a very strong 18%. In contrast, the 3Q/2Q memory market increase in 2018 was 8%, less than half of last years rate.

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

 

[Eric Esteve]

I think we should analyze at least two type of results:
- Total semi market (like the data reported in the McLean report)
- Total semi less memory (DRAM, NAND, etc.)

I did this exercise to prepare my presentation at IP-SoC next December. The benefit is to remove an hectic and cyclical part of the semi market, so you can compare with EDA results or Design IP results (that I did in the presentation) and derive more accurate conclusion.

I have nothing against the memory market (I have even sold DRAM to HP when working for TI!), but it's so hectic that it prevent to see the real trends of the SoC market. The SoC market is growing continuously (even when the memory market collapses) and is, to my opinion, the main contributor to the IP business, and a very important to the EDA business.

Unfortunately, this memory market will stop growing in 2019, and probably decrease in 2020. This create a real issue because the people working in Wall St don't make the difference, and all the semi stocks maybe impacted by the memory market collapse!

 

[Chipper]

Two points. One is simply comparable; after an outsized year the next year's comps will just look bad, especially in any GDP+ type of business. Perhaps sequential comparisons are in order too.
Secondly what drove the outsized 2017 numbers besides memory that is gone now? We had a big crypto inventory build. PCs are actually stronger. Handsets are not much weaker. US GDP is up. Cloud infrastructure is surely up. What besides crypto and perhaps a memory glut would explain the decline?
Finally why would memory be down in 2020? Handsets? AI is memory intensive and is growing like a weed.

 

[Eric Esteve]

Finally why would memory be down in 2020? Handsets?...

Remember that we don't measure the memory market growth in term of Gb, or production units, but in $.
Here we have to remember that the memory market has ALWAYS been a cyclical market, easy to catch why:
- when the memory business is very high (price/Gb growing), it's because the demand is higher than the offer
- memory manufacturer have to build new fabs to satisfy the market demand, and possibly increase their revenues
- but, unlike with oil and OPEP, nobody regulates the price or capacity
- 18 months to 2 years after the peak, the ASP/device is going down, sometimes with negative margin, and the market collapse... Even if you sell MORE memory devices than during the peak.

Take a look at the DRAM market in $ during a long period of time, like on this pic, the cyclical nature is obvious:

View attachment 22682

2006: $35B --> 2008: $28B
2010: $41B --> 2011: $31B

 

[Chipper]

So you don't buy the "we are an oligopoly now and therefore more rational" argument or "we are not growing wafers so are getting our capacity gains from shrinks, which are not yielding as much as before" thesis?

 

[Eric Esteve]

No, I don't! DRAM manufacturers are and have been too greedy (and they probably must be because of the nature of their market)...

 

[Chipper]

Just being a devil's advocate here, but what about the end of shrinks? It seems to me that without being able to ride meaningful price-performance gains from a big new shrink in order to gain wicked share, it makes more sense for the players to just get along. Perhaps the competitive vector will be more "around the edges" in other features like speed, storage class memory or custom interfaces but the big product differentiations and therefore big market shifts are behind us now. Weak shrinks remove the potential for big differentiation and therefore the motivation to spend big to win big. Meanwhile with processors not getting much better while workloads continue expand, the best way for cloud vendors to keep up their SLAs without turning every building into a datacenter will be to feed their machines best with tons of memory.