2024-01-12 11:45:42 ET
Summary
- TSMC's revenue declined 5% in FY 2023 YoY, according to the last Q4 2023 results, beating expectations and being an improvement given the guidance in Q2 2023 of -10% YoY.
- AI processors only represent a small portion of TSMC's total revenues but have a CAGR growth prospect of around 50%.
- TSMC faces competition from Intel and Samsung in the most advanced nodes segments, which could impact its market position over the long term.
- Geopolitical risks should be incorporated into any investment analysis of the stock.
- I calculate the intrinsic value by taking Warren Buffett's view of the stock. I rate the TSMC stock as a hold.
I rate Taiwan Semiconductor ( TSM ) (TSMC) as a hold as there are important factors beyond the company's excellent fundamental prospects, adding that there are underlying assumptions that need to be assessed if an investor wants to consider TSMC as a good investment opportunity now. TSMC's technological leadership and its well-deserved market position in the foundry segment might mislead an investor who is just solely focusing on its growth prospects associated with the long-term tailwinds of artificial intelligence. In this article, I will focus on certain critical aspects that should be considered to balance the pros and cons of TSMC, taking into account the current stock price. It seems that the current stock price is already incorporating the company's growth prospects for the next few years.
This is my second article about semiconductors, and if you've read my previous articles, you'll notice that my horizon is long-term; in other words, buy at a good price and hold the stock for years. On many occasions, good opportunities in the short term do not mean good opportunities in the long term. I will offer my intrinsic value, taking Warren Buffett's view of the stock as he invested in it in the last quarter of 2022 and sold most of his position in the first quarter of 2023.
Context
In the call for the Q2 2023 results, CEO C. C. Wei of TSMC said the following:
Thus while we maintain our forecast for the 2023 semiconductor market, excluding memory, to decline mid-single digit year-over-year, we now expect the foundry industry to decline mid-teens and our full year 2023 revenue to decline around 10% in U.S. dollar term. With such inventory control, we also forecast the fabless semiconductor inventory to exit 4Q '23 at a healthier and lower level as compared to our expectation 3 months ago.
In that call, the TSMC CEO gave guidance of -10% for FY 2023 YoY, most likely due to macro headwinds in the global economy and weaker demand from China, which impacted the end market of smartphones and PCs, both of which are the most important segments for TSMC. However, TSMC's performance improved in the third quarter, which helped to have revenue growth accumulated as of September 2023 of -6.2% thanks to strong AI demand while other key markets started to recover. The last Q4 2023 results beat the management's expectations, reaching a revenue of $20.1 billion in Q4 when the guidance was between $18.8 billion and $19.6 billion.
Some people might think that one of the most important TSMC's revenue drivers is the demand for AI processors, which could be defined as CPUs, GPUs, and AI accelerators; nevertheless, AI processors only represent 6% of total revenues, although with a CAGR growth prospect of around 50%, which means an impact on TSMC's total revenues of around the low teens. As such, it is expected that AI systems will represent a higher portion of revenues in the next few years, as they will grow faster than total revenues.
However, I need to see if most of that growth is already incorporated into the current price, which is something that I will show later. On the other hand, there is no information released about the net income as of 4Q 2023 YoY yet, but as of 3Q 2023 YoY, net income dropped 16.9%, which is explained not only by the higher fixed costs associated with a lower use of capacity due to the difficulties in the PC and smartphone industries in the first half of 2023 but also by the high costs associated with the development and ramping of the 3-nm and 2-nm process nodes.
TSMC's Market Position
TSMC has been reinforcing its already strong market position in the foundry industry in the last few years, using most of its free cash flow to reinvest in its business, demonstrating excellent execution, which is shown by its market share of more than 50%. The information I will present is until Q3 2023 since it has not been updated until Q4 2023 yet:
Furthermore, the semiconductor industry experienced difficulties in 2022, but TSMC's revenues fell more smoothly, showing an interesting resiliency.
There is no information about the revenue breakdown by node as of 4Q 2023 yet, but there will not be material changes if I take the revenue breakdown by node as of 3Q 2023. I noticed that the 3-nm, 5-nm, and 7-nm process nodes represent almost 60% of the total revenues, giving TSMC strong pricing power, which is reflected by its very high FCF margin average in the last 5 years of 61%:
In the right chart, it's shown how the 3-nm is beginning to be reflected in TSMC's revenues, so it's very positive to see how the most advanced process nodes, which have a long way ahead in the next few years, represent a large portion of TSMC's total revenues. In addition, I should point out that TSMC has 90% of the global market share in the 7-nm and 5-nm nodes.
Thus, it's understandable that many analysts are bullish about the company's prospects given the long-term tailwinds of the several platforms where TSMC is involved, such as high-performance computing ((HPC)), smartphones, the internet of things (IoT), automotive, and digital consumer electronics ((DCE)).
Competition: A Factor to Bear in Mind Even For a Company Like TSMC
There are two other important competitors, particularly in the most advanced nodes segments: Intel (INTC) and Samsung (SSNLF). I wrote about Intel in June 2023, showing a bullish thesis when the stock was trading at $33 per share, mentioning the probabilities of a successful turnaround for the company. In fact, in that article, I showed that capital allocation was one of the main drivers that put TSMC in a way better position than Intel in the last decade, so maybe it would be interesting that you read that article too.
DIGITIMES Research, January 2022
In the table above showing the roadmap of the three companies for the next few years, I can see that TSMC and Intel will use the architecture Gate-All-Around Field-Effect Transistor (GAAFET) for their respective 2-nm production since 2025, whereas Samsung is already using it for the manufacturing of its 3-nm process node after replacing the FinFET architecture, which is the architecture that Intel and TSMC are using currently.
The FinFET architecture was very important in reducing the process nodes from 22 nm to 16 nm , as it enables less leakage current, enhanced drive current, faster switching time, and superior scalability while reducing the transistor's size. Most of the problems related to the previous architecture, Planar FET, were solved with the FINFET.
With the last advances in the process nodes of 7-nm and 5-nm, the FinFET architecture showed signs of burnout as the increase in efficiency was slowing down. In the process node of 3-nm, TSMC is still using the FinFET architecture, but it is struggling as it has only gotten a yield of 50% in three out of the five different types of 3-nm process nodes as of October 2023; any yield lower than 70% is considered poor. Remember that two of those 3-nm are already in production.
On the other hand, Samsung has three known 3-nm process nodes, of which one is already in production. However, Samsung is not performing better than TSMC in this process, as it only reached 60% yield using the GAAFET architecture. Probably the main advantage of Samsung over TSMC is that Samsung is already familiarizing itself with this new architecture to face the next lower nanometers, while TSMC and Intel would start using that architecture in the next nodes.
Now, I do not think that Samsung could represent a very strong threat to TSMC since the latter has a better reputation, and most likely, Samsung could be seen as a second source for several clients prioritizing TSMC's foundry services. In 2021 , Nvidia (NVDA) experienced a shortage of its GPUs because of the poor yields of Samsung's 8-nm process nodes. That encouraged Nvidia to switch its RTX 40 series to the TSMC's 4-nm process node. In 2023 , Samsung improved its yields to 70% for the 4-nm process node.
Intel could be an interesting case to follow since it's known that the company experienced delays in its 7-nm, and most analysts do not think that the company could catch up with TSMC. However, I would not disregard Intel since there are situations that might favor its long-term prospects; for instance, in December 2023, Intel purchased 6 out of the 10 high NA-EUV lithography tools that will be manufactured by ASML (ASML) for 2024.
With these machines, Intel could follow its ambitious roadmap for the next few years, as these high NA-EUVs would enable it to produce process nodes lower than 2-nm. TSMC CEO thinks that even in that scenario, TSMC would still be the leader, as he said in the last call for Q3 2023 results:
In fact, let me repeat again, our 2-nanometer technology without backside power is more advanced than both N3P and 18A, and will be semiconductor industry's most advanced technology when it is introduced in 2025.
N3P is one of the 3-nm types from TSMC, whereas the 18A is Intel's 1.8-nm, which is set to be manufactured by Intel in the second half of 2024.
TSMC is the dominant foundry now, but the company might face a hard battle with other competitors in the next few years, particularly Intel. This is important for investors who have very long horizons in their investments, such as 5 to 10 years ahead. Even when I think that there is long-term growth for the three players, it's important to be aware of the possible factors that might slow down TSMC's growth in the future to pick a better price for the stock now.
TSMC's Secret Sauce Might be Weakened
I read an interesting article about the beginnings of Taiwan and how it became a global powerhouse in semiconductors; for instance, in the 1970s, TSMC was already able to achieve yields of its chips of around 80%, whereas US companies had yields as low as 10%. In the 1980s, Japanese companies could average 60% yields.
Dr. Chintay Shih was one of the pioneers in Taiwan who studied his Ph.D. in Electrical Engineering at Princeton, working some years at Burroughs Corporation in St. Louis, Missouri, before finally returning in the late 1970s to Taiwan to be one of the leaders of the Industrial Technology Research Institute that set the foundations for Taiwan's semiconductor industry.
He implies that there is an ecosystem in Taiwan that is related to the educational system and idiosyncrasy that enabled the country to have brand new facilities with the most updated equipment; over the years, it was possible to recruit the best engineers, and even machine operators were highly skilled. The semiconductor industry in Taiwan attracts the most talented professionals because it's the sector with the highest salaries. Dr. Shih stated that Taiwanese professionals absorbed knowledge from their American teachers while applying continuous improvements.
So, I imagine that it's very difficult to replicate that ecosystem in other places outside of Taiwan. TSMC is building fabs in Germany, Japan, China, and the US, but according to the management, those fabs would not focus on the manufacturing of the most advanced chips, as it was clearly stated in the last call for the Q3 2023 results that these fabs would be only focused on the 12-nm to 16-nm process nodes and the 22-nm to 28-nm process nodes.
As such, the strategy of expansion geographically clearly does not solve one of the main problems of TSMC: The concentration of its most important factories in Taiwan.
Taiwan's Geopolitical Risks: A Factor to be Considered For a Long-Term Investor
US intelligence services believe that the worst scenario might happen in the next few years. In my own view, I am not sure about this possibility since an event like that would cause so much damage to international trade, including that of China.
It's hard to imagine all the possibilities about the final result of this event, but in any circumstance, I think that TSMC's main ecosystem for semiconductors would disappear. I think this event is very unlikely now as China is in the middle of economic, political, and financial challenges to be solved, but I cannot make a decent prediction about Taiwan's future in the next 10 years or even in the next 5 years.
There are two factors that made me put this issue as a variable to be considered in my investment thesis: i) Warren Buffett's divestment of his TSMC stock, and ii) Beijing's regulatory crackdown that started with Ant's IPO in 2020. In the first case, Buffett invested in TSMC stock in November 2022 at around $60 per share, selling his shares in the first quarter of 2023 at around $100 per share.
Buffett stated that TSMC is an incredible company and that there's no other company in the industry in its league, but he would prefer deploying that capital in Japan or in the US. In other words, he felt uncomfortable with the geopolitical risk associated with the TSMC's location. Buffett is an investor who likes to gauge the probabilities of different risks associated with a certain stock in which he is interested. He only invests in companies for a buy-and-hold strategy when he can calculate those probabilities, and if those probabilities are low enough, that makes him feel comfortable. Apparently, Buffett was not able to calculate those probabilities associated with TSMC's geopolitical risk for the next few years, so he divested most of his position.
The stock price Buffett paid for TSMC stock was very attractive, so we will discover if Buffett was right when I calculate an intrinsic value later. The second point that made me think about the invasion of Taiwan was that Xi Jinping in 2020, unlike the West, was not prioritizing support for the Chinese economy in a very hard scenario like COVID-19.
Indeed, we should remember that, in 2020, after Jack Ma's speech at a conference, Xi Jinping halted Ant's IPO while establishing new regulations for the whole tech sector in China, in the middle of the serious economic consequences of COVID-19. When the West was dictating policies that favored consumption while working to reduce disruptions, China was setting strong regulations for its tech sector, weakening its private sector in an "inadequate" time, creating even more disruptions for themselves. That made me think that factors associated with the economy might not be the most important for Xi Jinping since an invasion of Taiwan could punish the Chinese economy enormously.
Now, probably you are thinking that an invasion of Taiwan would impact very negatively companies like Apple (AAPL), Nvidia, AMD (AMD), Broadcom (AVGO), etc., or, in other words, TSMC's clients. Nevertheless, I think that it will depend on when that scenario happens; for instance, if that scenario happened in 10 years, companies like Intel might be in a good position at that time to be an alternative to TSMC; if that scenario happened in 5 years, there could be problems associated with Intel's capacity. That might cause temporal disruptions for all these TSMC's clients, but those could be solved over time. But, in any scenario, if Taiwan is invaded, TSMC's ecosystem of its most important nodes would disappear forever.
If you are an investor who holds your stocks for 1 or 2 or even 3 years, this scenario would sound very unlikely. I think that it's very hard to make such a prediction, as global trade could be seriously affected, including China itself. But, in my view, this scenario should be incorporated in some way into your investment decision at TSMC.
Valuation Using Buffett's View
Even when I see that TSMC stock appears to be attractive according to several multiples, I would say that most of these metrics do not incorporate geopolitical risks. As such, I will offer another method to see TSMC's valuation in a more conservative way.
Assumptions:
Let's make some assumptions:
- Outstanding shares: 5,185,800,000 (as of September 2023).
- FCF margins: 60% (average of the last 5 years).
- Exchange rate: NTD1 = 0.032 (in the last 10 years, the exchange rate has been stable, fluctuating between NTD0.030 and NTD0.035 per dollar).
- I assume that TSMC stock is held by the investor until 2026.
- Revenue growth for FY2023: -5% (according to the last Q4 2023 results ).
- Revenue growth for 2024 and 2025: According to consensus .
- Revenue growth beyond 2026: I assume 20% is aligned with the previous 2 years.
- P/FCF 2030: 11x (the average of the last 5 years was 12%).
- Discounted rate: 13% (a high discounted rate assumed to incorporate a risk premium associated with TSMC's location; as a benchmark, I assume a discounted rate of 15% for Chinese companies and 8% for strong companies in the US).
Based on my assumptions, I make a projection of revenue growth of 20% for 2024, 2025, and 2026. So, I take the revenues projected for 2026, and then I multiply those revenues by FCF margins of 60% of my assumptions to get an approximate FCF of NTD2,209,843 million for 2026.
Now, I take that estimated FCF and divide it by the outstanding number of shares of 5,185,800,000 million to get an FCF per share for 2026, getting NTD426.1 per share. In this sense, I take the FCF per share of NTD426.1 and multiply it by 11, which is the TSMC's average P/FCF of the last 5 years. Then, I get a target price of NTD4,687.47 per share in 2026, so I calculate the present value to bring it back to 2024.
Thus, finding the intrinsic value in dollars:
Intrinsic value = 4,687.47*exchange rate/(1+discounted rate)^3
You can see that I am considering 20% revenue growth for the next 2 years , which takes into account the ramp-up of two different types of 3-nm nodes, N3P and N3X, and the ramp-up of 2-nm for all these periods. I assumed an P/FCF of 11x for 2026 since, in the last 5 years, that average multiple was 12x.
Now, considering a discounted rate of 13% and FCF margins of 60% for 2026, which has been the average in the last 5 years, I would say that an intrinsic value could be around $105.32 per share. Buffett sold at a price of around $100 per share after holding it for some months, which means that he estimated an intrinsic value very similar to that I calculated.
Of course, the intrinsic value could be higher if I take, for instance, an P/FCF of 14x at a discounted rate of 10%; I could get an intrinsic value of $145 per share. Bear in mind that the stock was trading at a multiple P/FCF of 14-17x in 2020 and 2021, when the market was bullish with all the tech sectors and nobody was talking about a potential Chinese invasion seriously.
Buffett is very conservative in his calculation of the intrinsic value, putting himself in the worst-case scenario when he is trying to make a forecast. That's how he builds up a margin of safety in any of his bets. So, it's likely that when he saw TSMC stock at a price of around $60 per share in the last months of 2022, he saw a very good opportunity, knowing that TSMC's intrinsic value was around $100 in his worst-case scenario. Once the stock reached $100 per share, he could have decided to hold it as he does with any other outstanding company, but he revalued his "buy and hold" strategy considering the geopolitical risk as he mentioned that he did not like TSMC's location, ending up selling almost his entire position.
TSMC is an outstanding business, but if Buffett is not able to calculate his own probabilities about the negative events a company might face in the future, he prefers to get out of that stock. However, in my calculation, I am assuming that the probability of an invasion of Taiwan is very unlikely in the period of time between 2024 and 2026.
Risks
I think that the most critical is the risk associated with a potential invasion in the future. It's possible that Xi Jinping is only making a rhetorical statement about his plans for an invasion of Taiwan, which would mean that the stock would experience an important surge in the next few years, even beyond 2026. Another scenario is that Xi Jinping just decides to give up on his plans to invade Taiwan, which would break my thesis too.
Nevertheless, I do not know for sure about the real intentions of Xi Jinping, but as disciplined investors, we need to incorporate in some way that scenario into our investment thesis to buy at a price that could mitigate, at least partially, the risk of that event. Some investors might sell their entire position; others might adjust TSMC stock in their portfolios. Depending on what you decide to do, it's important to be aware of all the risks associated with any stock in which we want to invest. In my case, I am assuming that until 2026, that scenario is very unlikely, but taking a higher discounted rate, I am punishing the intrinsic value to have a reference while building up a margin of safety.
Final Thoughts
The purpose of this article was to evaluate the pros and cons of TSMC, which is, undeniably, an outstanding business with an amazing culture and a strong moat. The company has a strong ecosystem in Taiwan that gives support to the company to have highly skilled employees in different areas, combined with an educational system that supplies those employees to the semiconductor sector.
However, I reviewed different factors that any investor should be aware of if he or she wants to invest in TSMC, particularly those investors whose horizons are long term, 5 to 10 years. One of the things that I showed in this article is that the company has very good long-term growth prospects, but apparently, those positive factors are already incorporated into the current stock price; in addition, that price might be incorporating a geopolitical risk and the potential difficulties of fierce competition in the next few years. In any case, we should not forget that the geopolitical risks might be low now, but those risks could increase over the years.
For further details see:
TSMC: Not All That Glitters Is Gold