Given that the semiconductor industry is clearly in the midst of a down cycle (even though there are cycle deniers, also members of the flat earth society…), most investors and industry participants want to know the timing of the down cycle and the shape of the recovery as we want to know when its safe to buy the stocks again. We also want to try to predict the ramp rate of the recovery so as to value the growth of the industry and the stocks.
There can be a lot of speculation as to the shape of the bottom of the cycle. Is it a “V” shape, which offers a quick, bounce back recovery? Is it a “U” shape which coasts into a trough with a slower return to growth? The CEO of a conservative company in the semi space has talked about always preparing for a “canoe” shaped down cycle which has an extended bottom. Then there are always the pessimists who talk about an “L” shaped bottom from which there is a very tepid recovery.
In order to try to predict “the shape of things to come” ( also a book by HG Wells….) we need to look at what drove the last cycles and what those drivers look like now and what new drivers or headwinds have emerged since then.
No two cycles have ever been the same as the industry drivers continue to evolve.
It seems at this point that we are likely not seeing a “V” shaped bottom with a sharp snap back and a quick return to growth after minimal disruption. Underlying strength in technology suggests that growth will return but the real question is at what rate?
At this point we assume no “black swan” event such as a true trade war which would cause a “nuclear winter” and the dreaded “L” shape bottom.
The thirst for infinite memory
Memory used to play a much smaller role in the semi ecosystem as it was largely confined to DRAM for use in PC’s to run the OS and applications. NAND has now replaced much of the rotating media as well as replacing optical disks, magnetic tape and photographic film as a storage medium. New faster, non volatile memory is on the horizon which could power another memory spike.
Memory hit an unsustainable peak of 84% of Lam’s business before falling. The historical norm of memory and logic/foundry being in closer balance was obviously upended.
Much of that was due to the conversion of 2D NAND to 3D NAND which is obviously a one time event. We are now past that spike in demand and back to a more normal production process flow.
Simply put, when we look back on the past few years of the strong up cycle we had a perfect storm of unusually high demand as most memory applications changed to silicon based memory coupled with a significant change in process technology requiring huge investments in the new process.
We are now at a new, more stable plateau of memory demand that while growing is not experiencing the “step function” growth seen in end use change. Additionally we are at a more stable production environment where any major NAND player is producing using 3D technology as 2D is essentially dead from a financial model.
While AI, VR and new applications will drive the TAM growth for memory we will not see the step function growth experienced in the past when other storage media was replaced in a very short period by NAND.
All this suggests is that memory growth while very good, will not be as “rocket-like” that we experienced over the previous years.
A “mix shift” away from dep/etch towards litho/metrology
We will also see a mix shift in demand for equipment companies products due to changes in process technology to keep up with Moore’s law.
A second factor underlying the unusually strong growth in the prior cycle for some equipment companies, and not others, was the fact that the long delay in EUV forced the industry into a “make do”, substitute, process technology of multi patterning.
Rather than the industry progressing on its normal path of better litho techniques to drive Moore’s law, it got side tracked into doing multiple deposition and etch steps as a substitute technology due to the unavailability of EUV.
Now with EUV starting to enter high volume manufacturing we can expect less multi patterning and thus less deposition and etch equipment and more lithography and lithography supporting equipment such as metrology and mask inspection.
Currently the industry leader (TSMC) is running a “hand full” of layers using EUV technology at 7NM. This is expected to enter more or less “full” EUV implementation at 5NM with a “low teens” number of layers processed with EUV. To be perfectly clear, multiple patterning doesn’t go away entirely as it will still be needed for EUV but its higher than market rate growth is essentially stopped.
This means that in addition to a slowing memory spend, AMAT, LRCX, TEL and other dep/etch companies will get hit by a double whammy of slowing multi-patterning spend in the cyclical recovery.
ASML will obviously be a big beneficiary of the shift to EUV but also metrology/yield management companies such as KLAC, NANO etc;, to deal with the more difficult litho techniques.
Overall CAPEX spend for EUV will be lower as compared to multipatterning as EUV is, after all, a cost savings technology as well as a Moore’s law driver. While not a huge headwind for the semi equipment industry it is none the less an impediment to getting back to a higher growth rate.
More importantly for investors, it likely changes the companies to invest in and the valuation of those companies as the growth rates will be different in the coming up cycle as compared to the prior up cycle. While all companies will benefit, the relative winners will be different in the coming up cycle.
The “black swan” cloud over the recovery
Another complicating factor that is near impossible to predict but must be factored into investors minds is the potential of a trade war or other global macro event. In prior administrations, the likelihood of a macro disruption to the industry was so low as to be ignored. Now, the probability of a trade disruption to the very international semiconductor industry cannot be ignored.
While the odds are still low, the potential continues to rile the markets and has to be taken into account by companies in the industry for their planning.
This indirectly can slow growth the semi market as companies may be more conservative and cautious as to their growth plans. Companies may be slower to spend or start new projects.
So while all bets are off on a trade war, we can assume a bit of a wet blanket on enthusiastic growth.
From Sunny & Warm to Cooler and Cloudy with a tornado watch
Overall we see the great growth factors which drove the industry over the past years cooling a bit in the next up cycle so we are looking for a slower recovery at least initially.
The long term for semiconductors is up and to the right as it has always been but with a little less gusto. While we obviously can’t predict the next product/application that causes a dislocation in demand but those new things that are out there, AI, VR, AR, autonomous vehicles etc; are longer term in nature and part of a strong foundation of growth rather than an upward spike.
On the supply side we have China coming on as a major new source of supply that will be incremental to existing supplies in Taiwan, Korea and the US. Predicted new supply from China will likely easily offset new demand as discussed above and likely at competitive pricing keeping revenue growth lower.
China supply will be increasing over the next few years as we come out of the current down cycle which could act as another headwind or impediment to a return to faster growth. However they are still not a factor at the leading edge so impact will be somewhat limited.
We still see no reason to jump back into the stocks as the momentum continues to be negative. Quarterly earnings reports will not be all that positive.
We don’t think semi equipment companies will be calling for a recovery in one quarter again after being burned. Analysts are still capitulating, after the fact, and dropping like flies.
Not all companies have capitulated. That capitulation may be the indication of a bottom. Stock buy backs and large cash hoards will soften the impact but not zero it out. Everyone is still making lots of money, just a little less of it.
Right now, our best guess is a “U” shaped down cycle with a bottom in H1 2019. We see a slower return to growth out of the bottom as the high growth factors have cooled a bit. We would also remind investors that Q1 is the seasonally weakest time for the chip industry so we don’t expect it to be the catalyst of a recovery.
The stocks could be off up to 50% from their highs. AMAT seems on its way to $30, LRCX feels like its on its way to $125, ASML to $160 and KLAC on its way to $90. This feels like a fairly normal cyclical correction in stock prices. As usual, smaller cap names and sub-suppliers are harder hit as they are on the end of the whip.