Semi-conductors: the value chain and exploring the opportunity

I wrote about the semi-conductor sector a couple of years ago (here).  Since anything and everything related to semi-conductors has been red hot as of late, I thought it might be time to take another look at this sector.

Similar to the last time I looked at semi-conductors, I was relatively intimidated by the complexity across the value-chain, so much so that I almost decided not to write this article.  Moreover, the volatility of the sector – both in terms of company share prices and news flow – is such that the passage of even a single day can change things quickly and markedly.  Look no further than this past week, with some of the chip companies experiencing 10%+ swings in prices in a single day!  Most of the prices and other metrics used in this article are from around mid-week (June 24^th), and some are already out-of-date.  Such is the nature of sentiment shifts related to semi-conductors, and for that matter, anything even remotely related to A.I. 

Semi-conductor manufacturing, and all things related (meaning the so-called “semi-conductor value-chain”), includes several categories of companies.

• Only a handful of companies manufacture semi-conductors as their sole business, and these are referred to as “foundries”.  Perhaps the best known foundry is Tawain Semi-Conductor Manufacturing Company, or TSMC (ADRs are TSM).

• Many companies both design and manufacture semi-conductors, but are generally not as efficient at manufacturing semi-conductors as are the pure foundries.  These companies are known as “Integrated Device Manufacturers”, or “IDMs”.  Examples include Micron Technology (MU) and Intel (INTC).

• Some companies only design chips and rely fully on foundries to produce them.  These companies are “fabricators”, or Fabless Semi-Conductor Companies (“Fabs”).  One of the best know Fabs recognised by investors globally is NVIDIA (NVDA). 

• Some companies are world-renowned for their involvement in the semi-conductor value-chain, although they are neither foundries nor fabricators.  These companies produce machines that are critical to the design and production of semi-conductors, the best known of which is ASML (Dutch company that makes lithography equipment).

In my personal portfolio, I own NVDA and ASML, but no other companies in the semi-conductor value-chain.  This includes both chip makers (foundries) or integrated device manufacturers.  And I can’t help but asking – am I too late to look at other opportunities since the run in semi-conductor stocks seems over-baked?

The sector is complex

Relatively speaking, the two companies I own are easy for me to understand.  However, when I started digging deeper into foundries, IDMs and fabricators, I quickly discovered that the sector is very complex.  For this reason, it would be difficult for me to properly evaluate the opportunities unless I spent hours and hours exploring the intricacies of the sector and the various companies involved.  One thing is clear – not all chips are created equally.  The quality of a semi-conductor can range from commodity-like (low value added, sluggish growth and low operating margins) to specialty (high value added, more robust growth, higher operating margins).  The combination of Street analysts’ research, along with materials from rating agencies and other specialists (like trade associations), will fill many hours of your time, should you really want to roll up your sleeves and do a deep dive.  I have to admit – I did not have the patience to do this.  As a result, take my commentary with a grain of salt as I admittedly am no semi-conductor expert.  Even so, I was able to conclude based on a cursory review that if I were to invest more in the semi-conductor value-chain, I would choose one or possibly two targets: TSMC (“TSM” ADRs) or the SOXX ETF.

Why are semi-conductor stocks so popular now?

Propelled by growth in data centres and significant forward investment by hyper-scalers and others to capitalise on the growth of A.I., anything semi-conductor related has been on a tear recently.  For example, the price of the semi-conductor ETF SOXX has doubled this year, and perhaps the best known foundry – TSMC – is up 46.4% YtD, more than triple the return of the NASDAQ 100 over the same period.  The enthusiasm for the sector has collateral beneficiaries, too.  With much of the global semi-conductor manufacturing located in Asia, the stock markets that have foundries/IDMs like the KOSPI (South Korea) and the TWSE (Tawain) have also soared in the last 12 months, as you can see in the table below. 

You might recall that there was a fair amount of focus in 2022 on semi-conductors, around the time that A.I. was starting to move into the forefront of investors’ minds.  The pandemic exposed supply-chain risks for the US, meaning that the US had insufficient on-shore semi-conductor manufacturing capabilities.  You might wonder how this could be the case when the most well-known semi-conductor company – NVDA – is an American company. However, as I mentioned already, NVDA is a designer of cutting edge semiconductors (a fabricator), not a manufacturer.  NVDA relies on foundries mainly located in Taiwan or South Korea to actually manufacture its cutting-edge chips.  The pandemic exposed the supply-chain vulnerability of the US as far as domestic chip manufacturing, an issue that the Biden Administration tried to tackle via the “Chips and Science Act of 2022.”  President Trump lambasted this act, but the Republican Party overall was not keen to repeal the act given the obvious vulnerabilities of the US in terms of ramping up chip production on US soil. 

Ultimately, the Trump Administration eventually grasped the severity of the situation by investing in Intel (INTL), a beleaguered IDM.  On August 22, 2025, the Trump Administration announced that the government would purchase $8.9 billion of common stock (9.9%) of Intel (INTC) to help the company boost its semi-conductor manufacturing capabilities on US soil.  By doing so, the administration hoped to reduce the country’s reliance on foreign chip manufacturing.  Since the investment was made, INTC’s shares have nearly quadrupled, very much in line with the acceleration in stock prices across the entire value chain. 

A.I. has been the driver

The growing focus on A.I. has increased sharply since the pandemic, most pronounced in the last six to 12 months.  This has driven a surge in stock prices across the entire A.I. value chain.  The insatiable need for semi-conductors initially started – at least the best I recall –when hyper-scalers began to significantly rachet up their expenditures on A.I. data centres.  Of course, the opportunity in A.I. has had plenty of other collateral beneficiaries in indirect sectors too, like utilities/power companies.  However, the biggest beneficiaries recently have been foundries and IDMs, which investors view as “picks and shovels” companies in the value chain.

What’s the downside of investing in a foundry or IDM?

The asset base of foundries is large because there are significant fixed assets associated with manufacturing facilities.  Based on what I read, a semi-conductor plant costs between $11 billion and $20 billion to construct, and can take three to five years (source here).  Foundries and IDMs have to spend significant amounts to add capacity in order to meet growing chip demand, driven more heavily than ever by the explosive growth of A.I.  Comparing asset intensity between a fabricator and a foundry, NVDA (fabricator, no manufacturing) has net fixed assets of $16.6 billion, and TSM (foundry, no design) has net fixed assets of $126 billion.  From a return on asset perspective, NVDA’s ROA is nearly three times that of TSM’s (52.7% vs 17.3%).  And keep in mind that TSM is by far the largest and most dominant foundry globally, meaning it is probably the most efficient. 

The current A.I. euphoria can cause investors to lose sight of the fact that manufacturing semi-conductors is an asset rich business that is vulnerable to business cycles.  The lead times exacerbate the boom and bust cycles of foundries, a fact that seems like the distant past in today’s booming global A.I.-driven economy.  In fact, until the last 12 months or so, investors wanted to be anywhere along the semi-conductor value chain except in foundries.  This is because the business is extremely cyclical, and most investors (this one included) generally leaned into fabricators like NVDA, which are deemed the “smart guys in the room” because they do not suffer as much from cyclical downturns.  In fact, any lingering concerns about cyclicality largely faded into the background in the second half of last year, when foundries came into focus of investors who recognised the potential bottle neck caused by limited production capacity of chip manufacturers that sit at the very beginning of the A.I. value chain.

Playing the sector (aside from NVDA and ASML)

There are two ways to get involved in the semi-conductor value chain: either through ETFs or through one or more of the individual names in the chip sector.  Let’s take a quick look at both.

ETFs: One of the best known and tracked ETFs is iShares SOXX, which is a passive ETF that tracks the performance of the NYSE Semi-Conductor Index.  It is comprised of roughly 78% semi-conductor companies and 22% semi-conductor equipment companies.  Another slightly broader index is the Vanguard VGT, which is a broader ETF that tracks the Information Technology Sector.  The VGT’s exposure to the semi-conductor value-chain is around 43%, so it is a less pure-play than the SOXX.  Note that both ETFs are diversified across many semi-conductor companies, reducing returns (vis-à-vis single names) but also substantially reducing single-name risk.  The table below looks at the attributes and performance of these two ETFs, compared to similar attributes for the benchmark S&P 500 and the NASDAQ 100 (tech-heavy).

Individual foundries and IDMs:  With a fabricator (NVDA) and an equipment company (ASML) already in my portfolio, I have been focused mainly on whether I should add an IDM or a foundry.  As I started to delve into some of the companies, it became clear that there were such significant differences in terms of their focus that to decide on a particular name would take many hours of research, and I would probably still get it wrong.  I had Perplexity look at the fabless producers and IDMs, to see which foundry they use (right most column) and their product focus.

To further whet your appetite, the table below looks at the financial metrics of some of the participants across the semi-conductor value chain.  I like to pay particular attention to EBITDA growth and EBITDA margins.  Note that NVDA, AVGO, TSMC (even though a foundry), SK Hynix and Micron Technology all have EBITDA margins above 50%, a form of moat from an investment perspective. 

The table below looks at the performance and valuation metrics of the shares of the companies listed in the table above.

The financials, share performance and valuations of these companies vary widely, mainly reflecting the differences in focus and market share in sub-markets across the participants.  The valuations are rather frothy in many cases to say the least.  Investors have most recently been focused on SK Hynix and Micron Technology (MU), both IDMs that produce much-in-demand “value-added” memory chips like DRAM, NAND and high band-width memory.  From my perspective, it would take too much time and effort to choose one of these companies (aside from the two I already own) because I would need to invest more time in fully understanding their business models.  Having said that, it is clear that these companies are both extremely well positioned in the semi-conductor sector, and their valuations are toppy but perhaps not ridiculous.  For momentum investors, one or both might be good plays on weakness, and you will have opportunities because the stocks are extremely volatile.  For what it’s worth, the other side of the very strong demand for value-added memory chips is that their prices are increasing quickly, with the latest casualty being APPL which has been forced to raise the price of its laptops and iPads by $200 (#FT story here for subscribers). 

If I were to add a third company related to semi-conductors on weakness, it would most likely be TSM because it is the easiest for me to understand.  The reasons I would choose TSM are:

• It is the largest and most dominant foundry, with a 71% market share globally of pure foundries, and 35%-39% market share globally of the semi-conductor manufacturing market if IDMs are included,

• It has high operating margins that act as a moat in difficult periods,

• It produces all types of chips, ranging from commodity chips to the most specialised chips,

• It has a rich valuation, but not as crazy as the likes of some of the IDMs that have been runners.

The downside is that TSM is in Tawain, which does raise some potential risk around “its address” given China’s focus on re-unifying with Tawain.

Conclusion

I am watching the semi-conductor sector carefully.  However, I am not keen on adding a third investment in the A.I.-driven semi-conductor sector until valuations moderate.  Were they to do so, I would consider the SOXX ETF (diversified sector exposure) or TSM (dominant foundry by a wide margin and well diversified).