Description

DISCLAIMER: Nothing expressed below constitutes financial advice or offer to buy / sell any security, and the author or parties associated with the author may hold positions in securities mentioned below. You are encouraged to obtain the services of a registered financial advisor, and are further encouraged perform your own research to come up with your own conclusions on any topics discussed below.

Over the years, several semiconductor capital equipment ('semicap') companies have been written up on VIC  (including AMAT and KLAC writeups by yours truly).   A relatively niche segment of the market, semicap share prices have performed rather well in the past 1 / 2 / 3 / 5 / 10 years, but I remain convinced they are still positioned to deliver attractive LT returns, both in an absolute and relative sense.   Therefore, I offer yet another write-up about semicap. This one is not company-specific but industry-specific.   There is no ETF for semicap currently in existence so I am fudging it and submitting it here as SOXX.  SOXX is the ticker for the iShares ETF that mimics the Philadelphia Semiconductor Index; this is the conventional basket favored by investors who wish to gain a diversified exposure to semiconductors.   The Philadelphia index is the best proxy here, even though it is imperfect and has some issues as an investment (because of its large weight in Intel… but this perhaps is a conversation for another day).  Even though this write-up is titled SOXX, in reality I am advocating for semicap companies (primarily AMAT, KLAC and Lam, but I also like a few others outside of the US).  I think owning semiconductors generally is fine, but owning semicap is even better. I hope that the thoughts offered below will be acceptable as fresh analysis and will be counted as a full submission under my membership requirements.   

Semiconductors are a better and less cyclical value chain than before

Semis GDP used to grow by 1.5x the growth rate of global GDP but in more recent years this accelerated closer to 2x+ GDP.  I believe this is sustainable.  Both in our personal and our business lives, we are increasingly accustomed to better, faster devices and better device-enabled experiences.  This is coinciding with the broadening of semis demand sources from PCs and smartphones to servers and auto / industrial applications.  Digital entertainment is becoming more immersive, immediate, social and unavoidable than ‘analog’ entertainment.  And in our professional / business lives, the cost of not having access to and not being proficient with the latest devices (and software / services) is becoming higher and higher, and I cannot see what will reverse this.  All of this translates to numbers.  Philadelphia Semiconductor Index revenue CAGR during the period of 2011-2021e is 11%, and the CAGR for EBITDA over the same period is 17%.   If you strip out the semicap companies from this semi index, the index’s sales CAGR is the same 11% and the EBITDA CAGR is even slightly higher, at 18%.   This EBITDA in turn produces the operating cash flow that the semi value chain uses to finance its equipment capex.   Note also that the index (ex semicap)'s EBITDA as a % of total assets has also increased significantly in recent years, further corroborating the improved condition of the industry.   If you have a fundamentally positive LT view of technology, I believe semiconductors and semicap equipment specifically are an underappreciated, inexpensive way to express that view.   The Philadelphia Semiconductor Index is on 28x trailing earnings, with a 12% ROA and 24% ROE.  The 3 largest American semicap companies (AMAT, Lam and KLA) are all currently trading in public markets at a price of between 19x and 24x trailing earnings and between 16x-17.5x forward consensus earnings, with ROA% in the 17%-25% range and ROE% of 49%-76%.   In the long term the semicap businesses are tightly linked to the overall semis value chain, so I think the LT revenue growth of semicap businesses should be similar to the LT growth of overall semi GDP.  But (1) the starting valuations for semicap are cheaper, (2) semicap industry has a more predictable oligopolistic setup, (3) semicap has stronger and more predictable capital allocation, and (4) semicap is on average much less capital intensive than the overall value chain.  Finally, (5) semicaps also have an important feature of their business model in the constantly growing installed base of machines, which attracts a predictable long-term stream of service, parts and upgrade revenue.  As a result of all this, I think semicaps are a significant improvement over an already attractive proposition of just being long the broad semiconductor index.

Semicap equipment is not about to drop off a cliff and is a secularly growing industry

Analysts frequently use the shorthand of % WFE (annual wafer fab equipment spend) / semi GDP or TSMC’s % capex / revenue to try to gauge where we are in the semicap cycle.  In 2021, % WFE / semi GDP is likely to be at least 14% and potentially as high as 16%. This is markedly higher than the average levels seen since 2011 (c.12% WFE / semi GDP).   A direct result of this is that when you look at recent sell-side analysis of companies such as Lam or Applied Materials, the analysts think 2022 / 2023 / 2024 (if their models even go as far as 2024 – and often, they don't) will be way down compared to 2021, both in sales & profitability terms.   But this is too simplistic.  Dividing WFE (or semicap sales) by semi industry sales (or GDP) makes little logical sense, but people still do it and fixate on it because it is the most widely available / easiest-to-calculate number.  As I stated previously, the value chain has structurally improved, and one of the symptoms of that is higher profitability ie higher EBITDA margins among most of the layers of the value chain.   If you divide WFE by the Philadelphia Semiconductor Index's EBITDA (excluding semicap EBITDA), this metric has steadily declined from a peak of 70% in 2012 to 43% in 2020.  In 2021, it is expected to further decline to 32%.  This is a far better, more apples-to-apples way to think about semicap's "affordability" for the rest of the value chain.  WFE constitutes semicap industry's sales and makes up a big portion of the capex that the semis value chain (with the exception of semicaps, obviously) spends out of the value chain's operating cash flow (for which EBITDA is a proxy here).  My calculation does not support the bearish notion that the world is over-spending on semicap equipment or that the value chain is overextending itself by buying too much equipment.   My argument can be sense-checked.  E.g. by zooming in on TSMC - the world's #1 leading edge foundry and the biggest spender on semicap equipment – you will find that even with TSMC's substantial step-up in planned capex over the next few years, as long as TSMC meets the bottom end of its projected revenue and implied margin targets (and substantially falls short of consensus), TSMC's % ROE and % ROA will still remain above 2016-2019 levels; and if TSMC matches consensus revenue and margin expectations, % ROE and % ROA will return to 2020 levels by 2024.   

An alternative way to frame the bearish view on semicap would be to state that the logic industry's capital cost-per-gate used to steadily decline but this stopped around 2010 (when the industry moved to 28nm), and if some new technological (or process-oriented) breakthroughs manifest themselves such that cost-per-gate will start declining again, this will mean lower required capital intensity and thus a big reduction in semicap demand.  I struggle to see where this kind of trouble for semicaps can come from.  Every comment from a semicap management team or foundry / logic management team that I've seen is pointing to the opposite.  Net capex required per 100,000 monthly water starts is increasing node-to-node (and this is why virtually everyone is saying that capital intensity is increasing, even though I've explained in the previous paragraph that when judged against value chain EBITDA or foundry ROA %, there are far fewer reasons for concern).  The most reasonable explanation for all of this is as follows.  Semicap companies are fundamentally able to get away with the rising cost-per-gate and rising capex-per-wafer-starts simply because semicaps are providing lots of value for the value chain (to foundry customers directly; to fabless designers more indirectly; and then to ultimate household and business consumers even more indirectly), and are being paid a defensible 'rent' for the value they deliver.  These kinds of dynamics are supported by the oligopolistic nature of the semicap industry, both overall and within each stage grouping (etch, dep, litho, process & inspection, etc).  Even if you look at the least concentrated, most competitive stage groupings (etch and dep), you'll find that the 3 key players (AMAT, Lam and Tokyo Electron) in etch & dep have enjoyed solid revenue growth and profitability.  Combined, in 2011-2021e these 3 companies' revenue CAGR is 9% and EBITDA CAGR is 14%.   Going forward, it will be hard for semicap to substantially outgrow or undergrow the overall semi GDP by too large a margin, but my base case expectation is for low double-digit EBITDA CAGR.  Combined with the starting point of c.16-17.5x forward EPS and semicap's consistent track record of reasonable capital allocation (including c.100% of FCF returned as dividends and buybacks), even if we ignore the prospect of any multiple expansion, I struggle to see how being long any of the 3 large American semicap companies (AMAT, LRCX or KLAC) will deliver an equity IRR of less than 15% over the long term.

Recent industry developments further improve the backdrop for semicap

During the 20-year period from the start of this millennium til the onset of covid,  I think it's fair to say that the electronics value chain has been trending ever closer in the direction of greater concentration and consolidation.  This happened pretty much at every layer of the value chain.  The number of foundry players capable of producing leading edge logic chips has shrunk.  The number of NAND producers has also shrunk.   The semicap industry has become more concentrated, with Lam acquiring Novellus (and significantly increasing its share in the deposition market, as a result), Applied Materials merging with Varian, and ASML organically gaining a dominant share of litho while Japanese vendors stagnated.  And at the end-customer level, we can see that Apple has grown to control a huge portion of smartphone spend (and is now TSMC's #1 customer, with 25% share) while datacenter & infrastructure spend is increasingly getting concentrated in the hands of AWS, MSFT's Azure, Google, and Facebook (to power its own ever-growing needs).  All these developments have coincided with greater globalization (ie greater % of global semi output moving to Taiwan, China, Korea and Japan).   

Some of these trends are now moving in the other direction, with a big assist from geopolitics and developments at Intel.  While specific policy support details are still a bit hazy, US and EU policymakers appear to have an increasing appetite to incentivize leading-edge manufacturing to remain onshore as much as possible. We know that when the US government decides something is truly strategic and vital to America’s defense, it is willing to unleash a budget of gigantic proportions to support it.  We might be in the early stages of moving towards this from a semis and semicap point of view.  The US government’s increasing distrust of mainland China's government is a big factor that is contributing to this.  There's also some renewed general desire for there to be more 'buffer' and redundancy in the value chain - this likely means lower utilization and thus more semicap tools required.  And at Intel, CEO Pat Gelsinger is increasingly likely to have genuinely gone all-in on his bid to restore manufacturing process leadership and 'catch up' to TSMC.  Whether or not Intel will ultimately be successful in its strategy is unclear, but for the foreseeable future, it is very clear that Intel will spend a lot more on capex.  I believe failure to spend is unacceptable to Gelsinger personally and to the US politically (and the more things like AI’s applications and impact on the everyday lives of ordinary people begin to crop up in the news, the more likely that the political pressure will only increase; I think this is secular).  Prior to Intel's most recent quarterly report, consensus was for Intel to spend around $20 bil in capex pa in the next few years.  The real number will be $30 bil, in my view.  Intel's newly confirmed $25-28 bil capex plan for 2022 will step up in future years, and has been released on a net-of-subsidy basis (meaning that either the gross receipts to semicap makers will be higher, or WFE as % of that capex spend will be higher, or a mix of the two).  An unrestrained Intel that’s willing to spend a lot and show minimal FCF for the next few years  (instead of prioritizing the dividend and stock buybacks, like they have for the past few years) is a positive for semicap companies.  And this is not to even say that before this, TSMC's large (and increasing) market share resulted in a capex monopsony that was so bad that semicap players didn't earn any money because TSMC beat them down on pricing so much that semicaps earned poor spreads vs their cost of capital.   Here, I am referring to semicap's concentrated customer base; I think this aspect of the business scares many potential investors away, and contributes to semicap companies trading cheaply.  This is even though semicap players have proven - beyond all reasonable doubt – that they are a strategic, symbiotic partner to their customers, both in logic and in memory.  Semicap companies have extensive information on customers' roadmaps and the qualification process to win an application at each node is lengthy.   Another reason why the "concentrated customer base = dealkiller" argument is silly: semicap's installed base of equipment (# chambers) is always growing and is attracting a decade+ stream of revenue from servicing, upgrades and parts - the customer is quite dependent on the semicap provider.  Lam Research said they expect Lam's service revenue to grow every single year.   As long as the overarching markets – technology more broadly, and semis GDP more specifically - continue to grow, it is hard to see the consolidated customer base hurting semicap too much.  Now that Intel is gearing up to compete more aggressively, semicaps should get an extra boost (even though they prospered without that boost in recent years), as Intel tends to spend more equipment dollars per unit of monthly wafer capacity than TSMC.  I believe Intel's capex hike should have been predictable for Pat-watchers, but for the most part they did not predict it, and I think people are still adjusting their expectations.  Heading into the capex hike, many semicap-observers and sell-side analysts believed 2022 may be a down-year for WFE, and if 2022 won't be a down year, then they thought that 2023 surely will be.  This fear was noticeable, despite growing indications that TSMC's "$100 billion over 2021-2023" plan increasingly is looking conservative.  I think 2022 and/or 2023 being down years should no longer be the base case.  If 2020 WFE was $61 bil and 2021e WFE is c.$85 bil, I currently expect at least low-% DD WFE growth in 2022.  I also expect 2023 to be at least flat vs 2022.  I think there is a real prospect of WFE reaching $100 bil in 2023.   I also don't believe that 2021 is way above mid-cycle WFE but more likely to be below mid-cycle.   If I am right, the short-term (ie next c.2 years) setup for semicap is quite good.  The starting point is favorable.  The big 3 American semicap companies are on 19x-24x trailing earnings and will grow EBITDA by a 10%+ CAGR in the next 3 years and beyond, while using relatively little capital to do this (ROEs % in excess of 50% on average).  Each year, these 3 companies are spending 100% of their FCF on buybacks and dividends.  

Putting all this together, we have relatively clean balance sheets, a 5%-6% starting FCF yield (all returned via buybacks and dividends), and a base case of 10%+ EBITDA growth annually.  This is enough for a 15% unlevered IRR, assuming no multiple expansion.  If there’s multiple expansion and greater use of leverage, the equity returns will be better.  I expect very similar LT returns from all of the 3 major American semicap companies (Applied Materials, Lam and KLA) and I think holding them as a basket (potentially also adding the large European litho monopoly and Tokyo Electron) is a good way to go for most investors because it removes much of the individual business risk.  Market share %s fluctuate and there are periods where one or two players significantly outgrow the other players.  If I was forced to recommend only one of these companies to someone, I’d probably go with Applied Materials (AMAT) simply because AMAT is the most diversified and likely to deliver the investor the “smoothest” return and the investor would thus be most likely to hold on to the investment for the long run.

Risks / where can I be wrong

The biggest area where I can end up being wrong is the economy.  An unforeseen shock to consumer and business confidence can result in people buying a lot fewer smartphones / computers and businesses spend less on IT (which translates to lower spend on servers). Semicap company short-term results tend to suffer in these circumstances.  Even though semicap still generates +ve FCF in such years, the sentiment tends to still overshoot to the downside when economic conditions are weaker and multiples can go as low as HSD P/E.  Semicap managements can also suddenly be too timid in buying back shares in such an environment.  Even though I think the equity % IRR generated by the semicap sector will be quite attractive LT, it will probably come with substantial volatility, and many investors have limited ability to tolerate that.  Because of the volatility, I think holding on to the shares will remain an ongoing challenge to my own psychology, too.  The other main risk is that semicap shares will be extra-sensitive to any newsflow relating to US-China foreign relations as well as any tension relating to Taiwan.  This is even though any increase in hostilities on the latter front has potential to create new vectors of semicap demand (and I’ve alluded to that in my prior write-up on AMAT).  

Catalyst

Semicap quarterly results and commentary

Key customer (TSMC, Intel, Samsung) quarterly results and commentary

Details on US / European government support packages for domestic semiconductor manufacturing

Compounding of intrinsic value / passage of time