TSMC: Undisruptable And Undervalued!
It's hard to find an undisruptable company trading at a discount, yet TSMC is offering that opportunity now. Given the outlook, the stock can easily double in the next 5 years!
Who will make the best chips 5 years from now? I don’t know. I think nobody knows.
Just look at the chart below and you’ll see why:
In July 2021, Intel had 64% market share in all data center chips while Nvidia had 26% and AMD had 8%.
Two years later, Nvidia’s share jumped to over 66% while Intel plummeted to 25%.
Nobody could see this coming.
What’ll happen 5 years from now?
Well, Nvidia is very well positioned to maintain its lead but AMD also has its strengths. Intel is currently acting like the sick man of the industry but it has the full force of the US Government support behind it.
Anything can happen.
Regardless of who designs the best chip, they will turn to one company to manufacture it: Taiwan Semiconductor Manufacturing Company.
It already makes almost all advanced chips for AMD, Nvidia and Apple.
We don’t know which designer will sell the most chips, but we know one thing: We will need exponentially more advanced chips than we have now to support AI.
Here, Jensen Huang explains:
It’s so simple:
Cloud providers will build new data centers that will take up to a million GPUs.
Those GPUs will need to be renewed once every 4-5 years.
They will spend exponentially more money to buy chips.
Designers will pay TSMC to manufacture those chips.
TSMC is uniquely positioned in the semiconductor supply chain.
It gets all the upside, but it has less downside than the designers.
When the demand is high, TSMC manufactures nearly all the advanced chips; when the demand is low, TSMC still manufactures all the chips required for maintenance so its downside is way more limited than the designers.
It’s even better positioned than ASML because it can increase its own efficiency to the full extent to increase the output before ordering new machines from ASML.
This is the best position you can be at while we are approaching the inflection point in the Artificial Intelligence era, where compute power will be nearly as valuable as water.
What’s even better than all of these is that you can now buy this amazing company at a fair price!
So, let me cut the BS and explain to you why TSMC is a great deal right now!
What are you going to read:
1. Understanding The Business
2. Moat Analysis
3. Investment Thesis
4. Fundamental Analysis
5. Valuation
6. Conclusion
🔑🔑 Key Takeaways
🎯 TSMC has a near monopoly in advance chip manufacturing and it’s entering the accelerated era of growth as the global chip demand is skyrocketing.
🎯 Chip manufacturing is capital intensive, creating high entry barriers. TSMC is the most advanced manufacturer that has way higher yields than the competitors.
🎯 Switching cost in the industry is high, reinforcing the pricing power of TSMC.
🎯 Its business performance has remained solid in the last five years and balance sheet is one of the strongest in the whole market.
🎯 Despite all its virtues, the stock is trading at an attractive price at 20 times forward earnings multiple.
🏭Understanding the Business
“The best victory is the one you achieve without fighting” says Sun Tzu.
How could it be possible?
Simple: You engineer it, just like Pan Wen-Yuan did. He engineered perhaps one of the greatest strategic victories of the 20th Century.
This picture is from 1937. Pan Wen-Yuan is in a ship, going to start his PHD studies in Microelectronics at Stanford University. After graduation, he worked as a researcher in the Radio Research Laboratory in Harvard University until 1945 then he started to work in Radio Corporation of America.
Four years later, in 1949, Nationalists in China lost the Civil War to Communists and fled to a nearby Taiwan, while Communists declared the People's Republic of China in the mainland.
As you could easily guess, the US Government had supported Nationalists against Communists. After Nationalists lost, the US supported Taiwan as China's rightful representative in global diplomacy. But the Mainland was too big to ignore forever.
In 1971, despite the US’ reluctance, the UN General Assembly recognized PRC as the rightful representative of China, replacing the Nationalist Government in Taiwan.
Can you imagine what Taiwanese leaders were going through at the time?
Country had just lost official UN protection.
The US was left with no reason to keep aggressively protecting Taiwan.
Mainland had a population of 800 million while Taiwan only had 15 million.
If the Mainland was to wage a war on Taiwan, it would be a total disaster. This was a war they would lose if they fought. They had to win without fighting.
They came up with a plan called “strategic industrialization.”
Their resources were scarce and there weren’t enough capable Taiwanese on the island who could undertake such projects. They turned to the Chinese in the US, Chinese whose loyalty lied with the Government in Taiwan…
Pan Wen-Yuan was one of them.
He had been working in Radio Corporation America (RCA) since 1945 and he had closely witnessed how America used the newly developed semiconductor technology in the Vietnam War. Government officials reached him in 1973 and he attended a breakfast meeting with several ministers in early 1974.
“We should develop a semiconductor industry,” he said, “if we fight with the Mainland, chipped weapons are the only way we could overcome their superiority in numbers. If we can develop fast enough and far enough, Taiwan can become a major exporter of chip technology to the US, in which case the US would put its full force behind us.”
This was a genius strategy, he was shooting for a victory without a fight.
They had charted the route and found a research institute to develop foundations of the chip technology. For 10 years, Industrial Technology Research Institute of Taiwan (ITRI) imported and then developed chip technology in Taiwan. That wasn’t enough though. Without commercialization, there wouldn’t be much to export.
They started looking for someone who could build the commercial semiconductor industry in Taiwan from the ground. That was hard. There was nobody in Taiwan who had high-level experience in a chip company, who were capable of building the industry from the ground. They were all in the US.
Wen-Yuan knew one of them. The guy had previously worked in Texas Instruments (TI) for 25-years, he even became the head of TI's semiconductor division: Morris Chang.
He had resigned from Texas Instruments in 1983 when he got passed a third time for the CEO position. When the Government officials contacted him, he was working as the COO of General Instrument Corporation but he wasn’t satisfied. He was looking for a new quest.
He accepted the Government’s offer without much hesitation. Government appointed him as the Chair of the Research Institute and tasked him with creating a semiconductor company in Taiwan.
This was the moment that defined the industry.
How can you create a semiconductor company from the ground?
Intel was dominating the advanced chip development and even well-capitalized competitors like AMD were struggling to catch it. Intel had:
Best pipeline of talent.
Dominant x86 architecture.
Full force of the US Government behind it.
Creating a new designer was simply a dead-end.
Chang had an idea though, he had learnt this in Texas Instruments.
TI was one of the leading chip companies in the US back then and the pace of work was frenetic. Employees were often overworked and burnt out. They would occasionally organize game nights and play the legacy game of the state to relieve: Texas Hold’em.
As a TI veteran, Chang had also learnt the #1 rule of winning in poker: Choose your table right.
He knew he couldn’t win in design no matter how well he played. He had to choose another table.
He noticed that the main reason other designers lagged Intel was manufacturing difficulties. Your more developed design was worthless if you couldn’t also develop your fab to manufacture it. This double cost structure only allowed the largest firm to thrive, others were simply feeding on the breadcrumbs.
He thought he could unlock a massive value and fierce competition between designers by doing the manufacturing for them. This way designers could spend all their capital expenditures in design and Chang’s company would take care of the rest.
Thus, the Taiwan Semiconductor Manufacturing Company was born.
This was the solution the chip industry was looking for:
Designers shared their blueprints with TSMC without fearing reverse engineering because TSMC wasn’t in the design business.
Incumbents could spend much more money in design, fueling competition and innovation.
TSMC could jump ahead in manufacturing as all the money it made would be spent to develop the manufacturing processes.
This worked out perfectly.
TSMC started with a 3 micron process in early 1990s, it’s now capable of manufacturing at 2 nano-meter process.
As TSMC became more and more capable, entrepreneurs who didn’t need to think about manufacturing created many more fabless chip companies.
Nvidia exists only because Jensen Huang could think of building a chip company without building a manufacturing plant.
One by one, fabless American chip companies outsourced manufacturing to TSMC. Even those who had fabs like AMD sold their fabs because TSMC was so superior and reliable.
What has this done to politics?
The US semiconductor industry gradually became dependent on Taiwan, warranting full diplomatic support and protection by the US Government over Taiwan.
TSMC was the most genius and elegant national security project that the world has ever seen.
Today, it has over 60% market share in global chip manufacturing. However, when it comes to most advanced chips like Nvidia GPUs, AMD CPUs and Apple’s M series chips, its market share rises above 90%.
Despite this striking background, what TSMC does is simple on paper: Manufacturing chips.
So, an outsider to a company could rightly ask: Isn’t manufacturing extremely competitive?
Well, yes, manufacturers like Samsung, UMC, GlobalFoundries are competing fiercely against each other, but TSMC is in a different league…
🏰 Moat Analysis
If you asked me to define TSMC in just two words, I would say:
“Undisruptable business.”
Together with ASML, they are the two companies that deserve this title most in the world.
When you think about the term “disruption” you instantly understand why it’s undisruptable. Think about software companies, what makes them so invulnerable to disruption?
We can say a few things:
Entry barriers are too low, any college hacker can start a software company from his dorm room.
Copying products is easy. Most senior software engineers can recreate 90% of the existing software. This means even more competition.
Switching costs are too low. Often it costs nothing to switch from one software to the alternative.
Foundry business expresses the exact opposite of these characteristics.
1) Semiconductor Business Is EXTREMELY Capital Intensive
To build a software company you will need:
Some starter servers through AWS or any other cloud provider, free-tier exists.
Stripe to monetize users. You pay nothing upfront, it cuts commission.
Paperwork to create the company, generally costs $300-$1,000.
In theory, you could start your software company for under $1,500.
Everybody can do that.
In semiconductor manufacturing, just the sunk cost of entry would easily reach +$12 billion. This is how much TSMC invested to create its facility in the US in 2021.
You need a huge amount of land, where you can build an industrial complex for you and your suppliers, you must get at least several hundred permissions from the Government, you must have deep enough pockets to fund a 4-5 year development plan only after which the plant will start operations.
Most importantly, you need these machines:
This is ASML’s next-gen Extreme Ultraviolet Lithography (EUVL) machine. Without this machine, you can’t manufacture cutting edge chips.
It costs $380 million per machine and money isn’t enough to buy it.
You have to go through a security clearing to get this machine. They have to make sure that your company doesn’t fall in any category of sanctions that could prohibit you from this beauty. Even after you get cleared, you have to order at least a year before to get it on time.
Even after all that, you will only get your foot in the door.
You will be the least advanced of all advanced chip manufacturers…
For comparison, despite all its expertise, TSMC had to increase its initial $12 million investment in the US fab to $65 billion to turn the fab into a state of art facility.
Well, do you know how many companies in the world have more than a $65 billion market cap?
It’s just 300.
2) TSMC Is Unmatched In Its Expertise
Well, everything TSMC has, Samsung has too:
It’s also a giant company with deep pockets.
It can also afford to hire top talent.
It also gets generous grants from the US Government.
Yet, its manufacturing process is way less successful compared to that of TSMC’s.
TSMC has over 90% yields in its 3nm process and 60% in 2nm while Samsung is struggling to get over 60% in 3nm and 20% in 2nm.
Meaning, over 90% of silicon wafers that got into TSMC manufacturing process come out as working chips while this number is around 60% for Samsung. And designers only pay for working chips.
You may think that this isn’t important, designers will get the working chips anyways. It’s not that simple.
Foundry itself absorbs the losses from lower yields. On scale this forces it to increase its prices which drives the designers away. TSMC, on the other hand, can offer lower prices on scale due to its higher yields.
TSMC could basically offer higher quality service at better prices.
This is not something that could easily be done in the semiconductor industry.
Beyond that, TSMC is the only foundry that has started developing the 1nm process.
As you see, Samsung is working only for 1.4nm, however, it won’t likely start it in 2027 since it has to fix 2nm yields first.
TSMC’s 2-nm yields, on the other hand, are already at the same level as Samsung’s 3nm yields. It’s the only company that can even think of 1nm.
It’s and it’ll remain the only address for manufacturing of the most advanced chips in the world.
3) Switching Costs In The Industry Is So High
Imagine you are running an online t-shirt store.
You are using Stripe as the payment processor but you are thinking about switching to Adyen.
What would be your switching cost?
Well, in the best case scenario it’ll be near 0, you will implement the change in dead hours and will probably miss only a few sales in the process. At worst, you will lose a week of sales.
Not a biggie.
Now imagine you are Nvidia.
What could possibly be your switching cost?
Imagine you could pay Samsung a little less per working chip. What could you lose if you switch?
Well, Nvidia recorded $114 billion revenue in the last 12 months and it has over 90% market share in data center GPUs.
If it switches and there happens to be some issues in manufacturing like yields and delays, Nvidia’s customer’s won't wait for it.
Its customers are cloud giants that are fiercely competing to have more compute power than the peers so their foundational AI model could be more advanced.
If Nvidia delays, they won’t wait, they’ll turn to AMD.
So Nvidia’s switching cost could be as high as its whole business.
Would it switch? Would you switch? Of course not.
As said in the introduction, regardless of who designs the best chips, TSMC will be the one manufacturing them.
It is simply an undisruptable business.
📝Investment Thesis
TSMC is so strong of a company that is positioned so incredibly well that the investment thesis becomes self explanatory:
1) Chip Demand Will Skyrocket
We are truly at a turning point in human history where compute power is an invaluable source.
I am going to be honest with you: I was skeptical of this narrative.
We all witnessed the ChatGPT breakthrough.
Yet, after an initial amazement, most of us quickly got disappointed by the capabilities of tools like ChatGPT.
They were hallucinating.
They lacked conception, they didn’t duly understand your wishes.
Next token prediction (inference) was far away from generating value.
I thought the industry players largely created the narrative to pump the AI stocks. I thought AI would be largely a productivity boost. Lately, I gave such tools another try to keep track of their development, I am amazed by what I ssaw:
Hallucinations are much less.
Conception is way better, you don’t need ELI5 prompts.
Inference is showing some signs of knowledge synthesis and creativity.
What’s most important then all is that agentic capabilities are much better. You can now create basic web tools, games, online workflows etc… They now write better code than junior developers and it’s developing fast.
Take this, Zuckerberg says AI will match mid-level developers this year:
How long do you think it’ll take to match senior developers? Not long.
The only problem is that capable AI is still expensive. It currently costs somewhere between $80-$120 per 100 messages of interaction with the most advanced models.
We will need exponentially more compute power with increased efficiency to meet the demand at cheap enough prices. And cloud giants are doing just that.
As you see, they collectively spent over $50 billion in capex in Q3 2024. Microsoft alone is planning to spend $80 billion this year to develop its data centers.
This is insane.
No matter where they get their chips, AMD or Nvidia, it’ll be TSMC manufacturing these chips.
2) It’s Now Protected From Chinese Invasion Scenario
In the last two years, the biggest factor that depressed TSMC stock was the threat of Chinese invasion. Though I have always stated that this was a very unlikely scenario, it’s also increasingly becoming irrelevant.
TSMC is rapidly building new fabs out of Taiwan now. They have ongoing constructions in the US, Europe, and Japan.
Until now, their global footprint was largely irrelevant for the most advanced chips as the Taiwanese government had mandated it to use at least 2 generations older processes in the overseas countries. Meaning, the most advanced chips were exclusively manufactured in Taiwan.
This is changing now.
The Taiwanese Government lifted this mandate recently, allowing it to manufacture 2nm chips in the US.
I still don’t think China will invade Taiwan in the foreseeable future, but still this is an important relief for its customers.
This means that even in such a scenario, they won’t have to rush to other manufacturers to get their chips made. TSMC will now be able to largely cover the demand even in the most disastrous scenario.
If this isn’t a catalyst, I don't know what is.
3) It Has Immense Pricing Power
If there is one company in the world that can demonstrate pricing power, it’s TSMC.
Its customers don’t care because they sell to the richest companies in the world that also largely don’t care because they are afraid that if they don’t pay, competitors will pay and get ahead of them.
This gives it a huge pricing power that enables it to grow revenues faster than orders it receives. And it’s using this.
It’s reportedly raising prices between 10-20% in 2025.
This gives it an additional revenue boost on top of the booming orders. Nobody has a choice right now. Neither Nvidia nor AMD can afford sliding behind to save a few bucks; nor Amazon, Google, Meta and Microsoft can afford having less compute power.
TSMC sets the price and everybody has to pay.
In sum, chip demand won’t decline anytime soon and TSMC will exploit all the organic growth, possibly more.
This is one of the most “no-brainer” theses out there.
📊Fundamental Analysis
➡️ Business Performance
Solid & steady. These are the two words that instantly appear in my mind when I look at its performance.
It has grown revenue and EBITDA by 80% and 90% respectively in the last 5 years.
You may think this is not a great growth, but note that you are looking at a $1 trillion company that manufactures chips. The business thrived even in the Covid Era between 2020-2022.
No, it won’t grow revenue tenfold in the next 5 years like a software company that has just found a product-market fit. But we know “for certain” that it’ll grow, and grow substantially.
For any investor who is looking for a solid investment, not speculation, this is an impressive track record and great position to be at this time of the history.
➡️ Financial Position
I remember mentioning if there was a “golden ratio” for financial health of the company, it would be:
Total Equity > Total Debt
EBITDA > Total Debt
Guess what, TSMC has this:
As you see, the charts create a perfect harmony: A steady declining line from equity to debt.
It’s absurd to make a lot of comments on the financial health of the company that has 4x more equity and 2x more EBITDA than its total debt.
So, I am going to emphasize something beyond the numbers.
Though magnificent, these numbers don’t capture the full financial strength of TSMC.
Imagine something happened yesterday, don’t get into much detail, just imagine something happened and its equity pool got totally wiped off, including all the cash. On top of that, a very strong earthquake hit its advanced manufacturing facility and it’ll take 4 months to continue operations.
Would it go bankrupt? Of course not.
Investors would rain money on it, banks would kill each other to give it a virtually unlimited line of credit, its customers such as Apple, Nvidia and AMD will invest and buy a piece of the company with a view to secure future supply for themselves.
It’s insane…
It already has a colossal balance sheet, but even that doesn’t truly capture how strong the company’s financial position is.
No further comment required.
➡️ Profitability & Efficiency
Return On Equity (ROE)
Since TSMC uses very little debt to fund its operations, I think ROE is the right metric to assess the quality of its investments.
How does it do here? Well, I will explain it with a book title: “All Quiet on the Western Front.”
Its median ROE in the last 5 years was 29%.
This is beyond amazing for me.
I think this is the first time I have seen a company having this much equity relative to its market cap and still posting above 25% ROE.
Normally, tech companies achieve these numbers by having relatively small equity pools. TSMC, on the other hand, achieves this with a giant equity pool relative to its size.
This is truly impressive.
Gross & Net Margins
Strength & efficiency… This is the story that its margins tell.
Just look at this:
Its gross margin never fell below 50% in the last 5 years despite the skyrocketing cost of raw materials. This means that it’s been able to pass all its costs to customers and perhaps even increased the premiums it charged lately.
What is even more striking is that the net margin is only 16% lower than the gross margin.
It only makes sense.
This is not a consumer company. It doesn’t need to aggressively advertise its services or hire sales teams to spread the word. This keeps the net margin high.
No manufacturing business has ever been able to reach these margins at this scale before.
There aren’t any red flags with its business, not even one.
📈Valuation
Valuation is at its best when you have:
Market leader company.
A fast growing global market.
Durable competitive advantages.
It is simple: If you have a market leader company that has a durable competitive advantage, it’s very well positioned to capture the growth in the market.
This is a very lucky situation as perhaps no company in the world is more undisruptable than TSMC and no already large market in the world is positioned to grow faster than chips.
The AI chip market is expected to grow at 28% CAGR in the next 10 years and reach $750 billion by 2033.
Overall semiconductor market is projected to reach $1.1 trillion by 2033.
TSMC will be the exclusive address for manufacturing most of the advanced chips going forward. On top of that, it can also use its pricing power to grow its top-line.
Given the market’s growth forecast and TSMC’s pricing power, we can conservatively assume that it’ll grow revenues 20% annually in the next 5 years.
This gives us $220 billion revenue for 2030.
I’ll remain conservative and assume that it won’t further expand its net margin and will have the same 40% profit margin.
This gives us $88 net income for FY 2030.
Give it a conservative 25 times earnings, we have a $2.2 trillion company, this is more than double of today's valuation, promising 17% annual return.
Given the undisruptable position of the company, its size and strength, I think 17% annual return is just great.
I would buy this everyday!
🏁Conclusion
TSMC is one of the strongest companies in the world.
It has a near monopoly position in advanced chip manufacturing.
Its customers have deep pockets, allowing it to exercise pricing power.
It’s entering an era of accelerated growth due to exploding chip demand.
On top of all these, it’s a great operation with 40% net margin and 29% 5-year median return on equity despite its gigantic equity pool.
Despite all its virtues, the stock price remained depressed between 2020-2023 as Chinese invasion concerns mounted. Now this risk has largely waned.
It has waned firstly because it’s now able to use the most advanced technology in its overseas plants. Second, the Chinese economy has its own problems. It’s in the longest deflation since the 1990s and record low bond yields show it’s not going anywhere soon. China’s economic situation is now too fragile to even consider such an atrocity.
This is the best setting possible for TSMC to thrive.
Moreover, the stock is currently trading below fair value.
If there is one thing that’s better than a great company trading at a fair value, it’s a great company trading at discount.
TSMC is a great bargain right now.
Have you given any thought to the risk that they are unable to continue decreasing chip sizes which allows competition to catch up over time resulting in a decrease in margins?
Completely agree with everything in your analysis, that is very well written. The only important thing that I would like to add is the black swan risk of a Taiwan invasion by China. This continues to make me uncomfortable to make TSMC a larger position.