원문 보기
Let me tell you something that most people in the technology industry will not say publicly. We are in the middle of the most important shift in computing since the invention of the transistor. And the companies that control semiconductor technology over the next decade will control the future of the global economy.
I am not being hyperbolic. I am not exaggerating for effect. This is simply the reality of where we are. I have spent over 40 years building semiconductor companies. I started in the industry when chips were measured in thousands of transistors. Today, we build chips with hundreds of billions of transistors.
I have watched computing power increase by a factor of over 1 million during my career. And I can tell you that what is happening right now, the convergence of artificial intelligence, accelerated computing, and advanced semiconductor manufacturing. This is unlike anything I have ever witnessed. Today I am going to share with you the three semiconductor stocks that I believe will dominate the next era of computing.
These are not speculative bets on unproven technology. These are companies with real products, real customers, real competitive advantages, and real paths to capturing enormous value as the world transforms around artificial intelligence. But before I reveal these three companies, I need to explain why semiconductors matter so much right now.
Because if you do not understand this context, you will not understand why these investments are so compelling. Every piece of technology you use runs on semiconductors. Your phone, your computer, your car, your refrigerator, your television, the data centers that power the internet, the servers that run your bank, the systems that manage air traffic control.
Everything digital depends on chips. Without semiconductors, modern civilization stops functioning. For decades, the semiconductor industry grew steadily but predictably. Chips got smaller and faster according to a pattern called Moore's law. Prices fell, new applications emerged, but the industry followed a relatively stable trajectory.
Companies knew what to expect. Investors knew what to model. Growth was good, but not explosive. That has changed completely. Artificial intelligence has created a step function increase in demand for computing power. Training a modern AI model requires more computing than all the computers in the world could provide just 15 years ago.
Running AI applications at scale requires specialized chips that did not exist a decade ago. The semiconductor industry has gone from steady growth to explosive growth almost overnight. Let me give you some numbers to illustrate this. In 2018, the market for AI chips was perhaps $5 billion. Today, it is over $100 billion.
By 2030, it could be $500 billion or more. That is 100 times growth in a little over a decade. No other segment of the technology industry is growing anywhere close to this fast. But here is what most investors miss. Not all semiconductor companies will benefit equally from this growth. In fact, most will not benefit much at all.
The AI chip market is dominated by a small number of companies with specific capabilities that are extremely difficult to replicate. If you invest in the wrong semiconductor companies, you will miss this opportunity entirely. If you invest in the right ones, you could generate returns that transform your financial future.
So, let me tell you about the three semiconductor stocks I am betting everything on. The first company is Taiwan Semiconductor Manufacturing Company known as TSMC. This might be the most important company in the world that most people have never heard of. And I am not exaggerating when I say that TSMC does not design chips.
They manufacture them. They take designs from companies like mine, from Apple, from AMD, from Qualcomm, from hundreds of other semiconductor companies and they turn those designs into physical chips. They are what is called a foundry and they are by far the most advanced foundry on the planet. Let me explain why this matters so much.
Designing a chip is incredibly difficult. It requires thousands of engineers working for years. It requires billions of dollars in investment. But manufacturing a chip at the cutting edge is even harder. It requires technology that pushes the absolute limits of physics. It requires machines that cost hundreds of millions of dollars each.
It requires expertise that has been accumulated over decades. TSMC has a near monopoly on the most advanced chip manufacturing in the world. When you hear about 3 nanometer chips or 2nm chips, the most advanced chips on the planet, TSMC makes almost all of them. Apple's latest iPhone chips are made by TSMC.
The most advanced AI chips are made by TSMC. The most powerful processors from AMD are made by TSMC. My company's chips are made by TSMC. There is no alternative. Intel has been trying to catch up for years and keeps falling further behind. Samsung makes some advanced chips but cannot match TSMC's yields or scale.
Chinese foundaries are generations behind and face export restrictions that will keep them behind. TSMC is it. They are the only game in town for cuttingedge semiconductor manufacturing. Now think about what this means for AI. Every advanced AI chip in the world needs to be manufactured by TSMC. As demand for AI chips explodes, demand for TSMC's manufacturing capacity explodes, they have pricing power because there is no alternative.
They have visibility because customers book capacity years in advance. They have a moat that is essentially impossible to cross because catching up to their technology would take a decade and hundreds of billions of dollars. The financial performance reflects this position. TSMC generates gross margins above 50%.
Which is extraordinary for a manufacturing business. They generate tens of billions of dollars in free cash flow every year. They have been growing revenue at double-digit rates. Even as much of the semiconductor industry struggles and with AI driving unprecedented demand for advanced chips, their growth is accelerating.
Let me give you some perspective on TSMC's scale and importance. They manufacture chips for virtually every major technology company in the world. Apple, the most valuable company on Earth, depends entirely on TSMC for its most important product. the iPhone. All those A series chips that make iPhone so fast and efficient are made by TSMC.
Apple has tried to diversify but keeps coming back because no one else can match TSMC's technology. The same is true for advanced AI chips. Every major AI accelerator is manufactured by TSMC. When you use an AI assistant, when you generate an image, when you get a recommendation from an algorithm, there is almost certainly a TSMC manufactured chip doing the work.
The company is invisible to consumers, but absolutely essential to the technology we all depend on. TSMC's technology leadership is not an accident. They invest tens of billions of dollars every year in research and development and capital equipment. They have the most advanced manufacturing processes in the world.
Their 3nanmter technology is in production. Their 2nanmter technology is coming. They are already working on technologies that will not reach production for 5 or 10 years. This continuous innovation keeps them ahead of anyone trying to catch up. The company culture is also a significant advantage. TSMC was founded by Morris Chang, one of the most important figures in semiconductor history.
He created the foundry business model that now dominates the industry. The culture of excellence, of continuous improvement, of obsessive focus on quality and yield that Morris established continues to define TSMC. They attract and retain the best manufacturing talent in the world. Some investors worry about geopolitical risk with Taiwan.
And yes, that is a real concern. But let me offer a different perspective. TSMC is so important to the global economy that every major power has an interest in ensuring its continued operation. The United States, China, Japan, Europe, everyone depends on TSMC. The company has also been diversifying its manufacturing footprint, building factories in Arizona, in Japan, in Germany.
They are reducing their concentration in Taiwan while maintaining their technological leadership. The valuation is reasonable for a company with TSMC's competitive position and growth trajectory. They trade at a premium to the broader market but not an extreme premium given their dominance and the growth they're experiencing for a company that is essential to the entire technology industry that has no real competition at the cutting edge and that is benefiting from the biggest demand surge in semiconductor history.
I think TSMC is one of the best investments you can make. The second company is ASML. If you have never heard of ASML, you are not alone. They are a Dutch company that makes the machines that make chips and they have a monopoly that is even more complete than TSMC's. Let me explain what ASML does. To manufacture advanced semiconductors, you need to print incredibly small features on silicon wafers.
The transistors on modern chips are just a few nanome wide. A nanometer is 1 billionth of a meter. To print features this small, you need a technology called extreme ultraviolet lithography or EUV UV machines use light with a wavelength of just 13.5 nanome to print patterns on chips. Creating this light requires a plasma that burns at hundreds of thousands of degrees.
Collecting and focusing this light requires mirrors that are the most precise objects ever manufactured by humans. The entire machine must operate in a nearperfect vacuum. It is probably the most complex machine humanity has ever built. Azimmell is the only company in the world that can make EUV machines.
Not the best company. The only company, it took them decades and tens of billions of dollars to develop this technology. No one else even came close. Intel tried to develop their own EUV technology and failed. Nikon and Canon, who dominated previous generations of lithography, could not make the leap to UV.
ASML has a 100% monopoly on the most important machines in the semiconductor industry. Every advanced chip made by TSMC, by Samsung, by Intel is made using ASML machines. Without ASML, there are no advanced semiconductors. Without advanced semiconductors, there is no artificial intelligence. ASML is a choke point for the entire technology industry.
The numbers are staggering. An EUV machine costs over $300 million. ASML sells dozens of them per year. Each machine generates recurring revenue from service, upgrades, and spare parts. Customers book orders years in advance. Because production capacity is limited, ASML has a backlog that stretches for years, giving them extraordinary visibility into future revenue.
Let me explain just how complex these machines are, because it helps you understand why ASML's monopoly is so durable. An EUV machine contains over 100,000 components. It requires mirrors polished to an accuracy of less than an atom. It generates plasma at temperatures hotter than the surface of the sun.
It operates in a vacuum more pure than outer space. The light it produces is so delicate that it would be absorbed by the air if the vacuum failed. No single company could build these machines alone. ASML coordinates a supply chain of thousands of specialized suppliers. Carl Zeiss makes the optics. Trump makes the laser systems.
Dozens of other companies contribute specialized components. ASML orchestrates this entire ecosystem, integrating components into a machine that pushes the absolute limits of what is physically possible. This supply chain itself is a moat. Even if a competitor wanted to build EUV machines, they would need to recreate this entire ecosystem of specialized suppliers.
Many of these suppliers have exclusive relationships with ASML. They have spent decades optimizing their components for ASML specifications. Building alternative sources would take years and cost billions. And demand is only increasing. As chips become more complex, they require more EUV layers. A cuttingedge chip today might require dozens of EUV exposures.
Next generation chips will require even more. TSMC, Samsung and Intel are all racing to build more advanced factories and every one of those factories needs more ASML machines. Aziml is also developing the next generation of EUV technology called Hine UV. These machines will enable even smaller features and even more powerful chips.
They will cost over $500 million each and once again ASML will be the only supplier. The monopoly continues into the next generation. The financial performance is exceptional. ASML has gross margins approaching 50%. They generate enormous free cash flow. They have been growing revenue rapidly and have visibility into continued growth for years.
They return capital to shareholders through dividends and buybacks while still investing heavily in research and development. Some people worry that ASML's monopoly will attract regulatory scrutiny or that customers will try to develop alternatives. Both concerns are overblown. ASML's technology lead is so large that developing an alternative would take decades and their customers need ASML to succeed.
Apple needs advanced chips for iPhones. My company needs advanced chips for AI. Everyone in the technology industry has an interest in ASML, continuing to innovate and deliver machines that push the boundaries of physics. The geopolitical situation actually strengthens ASML's position. Western governments are restricting the export of advanced semiconductor equipment to China.
ASML's EUV machines cannot be sold to Chinese customers. This means Chinese competition in advanced chips will be delayed by years or decades. ASML's customers outside China benefit from this protection and they are willing to pay premium prices for access to the machines that keep them ahead. The valuation for ASML is premium but justified.
This is a company with a literal monopoly on the most important machines in technology. growing revenue at double-digit rates with a backlog that stretches for years and with the next generation of technology locked up. You pay up for quality like this. And over the long term, I believe ASML will reward investors who recognize what they have.
The third company is my own company. Now, I know that might seem self-serving, but let me make the objective case for why this is one of the three best semiconductor investments you can make. We make the chips that power artificial intelligence, not some AI applications, almost all of them. When the largest AI labs train their models, they use our chips.
When cloud companies deploy AI services, they use our chips. When enterprises run AI applications, they use our chips. When researchers push the boundaries of what AI can do, they use our chips. We have a dominant position in the most important computing market of our generation. Let me explain why this dominance is so durable.
AI computing is fundamentally different from traditional computing. Traditional processors, CPUs, are designed to execute instructions one at a time, very fast. AI requires something different. AI requires doing millions of calculations simultaneously. This parallel computing is what graphics processors, GPUs were designed for.
We have been building GPUs for over 30 years. We have more experience in parallel computing than any company on Earth. But hardware is only part of the story, maybe 20% of it. The real mode is software. We have spent decades building a software platform called CUDA that makes our chips programmable.
Every AI framework, every AI library, every AI tool works with CUDA. Millions of developers know how to write code for our chips. Billions of lines of AI software have been optimized for our architecture. Switching to a competitor would mean rewriting all that software, retraining all those developers, recreating all those optimizations.
That would take years and cost billions of dollars. And by the time you finished, we would be another generation ahead. The software ecosystem is the real competitive advantage. And it compounds over time. The more developers use CUDA, the more software gets written for it. The more software gets written for it, the more valuable our chips become, the more valuable our chips become.
The more developers use CUDA, this flywheel has been spinning for over 15 years and accelerates every year. I watch competitors try to break into this market. They raise billions of dollars. They hire talented engineers. They build impressive chips. And then they struggle to gain traction because they cannot replicate the software ecosystem.
The hardware might be competitive on some benchmarks, but customers need software that works, developers who know how to use it, and an ecosystem of tools and libraries that make development efficient. We have that. Competitors do not. The financial results speak for themselves. Our data center business has grown from almost nothing a decade ago to tens of billions of dollars per quarter, and we are still in the early innings.
AI adoption is concentrated in a relatively small number of hypers scale companies and AI labs. As AI spreads to every industry, as every enterprise deploys AI applications, as AI becomes embedded in every product and service, demand for our chips will continue to grow. Let me share some perspective on the scale of opportunity ahead.
Today, most AI workloads run in a handful of large data centers operated by cloud companies. But that is changing rapidly. Every major enterprise is now developing an AI strategy. Every industry is exploring how AI can improve their operations. Healthcare companies want AI for diagnosis and drug discovery.
Financial firms want AI for fraud detection and trading. Manufacturers want AI for quality control and optimization. Retailers want AI for personalization and inventory management. Each of these companies needs computing infrastructure to run AI. Some will use cloud services. Others will build their own systems.
Either way, they need chips. And not just any chips. They need chips designed specifically for AI workloads. That is what we provide. The total addressable market for AI computing is expanding from billions to hundreds of billions to potentially trillions of dollars over the coming decades. We are also expanding beyond just training chips.
Inference, which is running AI models rather than training them, is a huge and growing market. We have chips optimized for inference. We have chips optimized for edge computing. We have chips optimized for automotive applications. We have networking products that connect AI chips together in data centers.
We have a comprehensive platform for AI computing. not just a single product. The networking opportunity deserves special attention. When you build a massive AI system with thousands of chips, connecting them together is a huge challenge. Data needs to move between chips at incredible speeds. Any bottleneck in networking reduces the efficiency of the entire system.
We acquired a company called Melanox that specializes in high performance networking. This acquisition has been transformational. Our networking products are now essential components of major AI data centers. We are not just selling chips. We are selling complete systems that include chips, networking, and software.
The automotive opportunity is particularly exciting to me. Autonomous vehicles require enormous computing power. They need to process data from cameras, LAR, radar, and other sensors in real time. They need to make life ordeath decisions in milliseconds. This is exactly the kind of workload our chips are designed for.
We have been working with car manufacturers for years building the computing platforms that will power the future of transportation. As vehicles become more autonomous, they will need more powerful computers. We are positioned to provide those computers. Some investors worry about competition. What about AMD?
What about Intel? What about custom chips from the cloud companies? What about startups with novel architectures? I take competition seriously, but I also see the structural advantages we have, our software ecosystem, our scale, our research and development investments, our relationships with customers, our track record of execution.
These advantages compound over time and create barriers that are extremely difficult to overcome. The market for AI computing is also growing faster than any competitor can capture share. Even if a competitor takes some business from us, the overall market is expanding so rapidly that we continue to grow.
We are not fighting for a fixed pie. We are capturing a large share of an exponentially expanding market. That is a wonderful position to be in. Let me talk about how these three companies work together because understanding the semiconductor supply chain is important for understanding these investments.
ASML makes the machines that print the patterns on chips. Without ASML, you cannot manufacture advanced semiconductors. TSMC uses ASML machines to manufacture the world's most advanced chips. Without TSMC, you cannot turn chip designs into physical products. My company designs the chips that power AI and has them manufactured by TSMC without our chips.
You cannot run advanced AI applications. Each company is essential. Each company has a dominant or monopoly position in their part of the supply chain together. They form the foundation of the AI revolution. Investing in all three gives you exposure to AI infrastructure at every level. If AI chip demand increases, all three benefit.
If a new AI application emerges, all three benefit. If a new company disrupts one part of the market, the other two still benefit. You do not have to pick which AI applications will win or which chip architectures will dominate. You win regardless because you own the picks and shovels. This supply chain concentration is also a source of risk that investors should understand.
If any one of these companies faces a serious problem, it affects the entire AI industry. A natural disaster in Taiwan could disrupt TSMC. Export restrictions could affect ASML. Competition could emerge for any of these companies. Diversifying across all three helps manage company specific risk while maintaining exposure to the AI semiconductor theme.
Let me address the question of valuation because I know these stocks are not cheap by traditional metrics. All three trade at premiums to the broader market. All three would look expensive if you use traditional price toearnings ratios. But let me offer a different perspective. These companies are not traditional businesses.
They are not growing at traditional rates. They are not facing traditional competition. Traditional valuation metrics assume that growth rates will revert to average, that competition will emerge, that margins will compress. For most companies, those assumptions are reasonable. For these three companies, they may not be.
TSMC has a near monopoly on advanced chip manufacturing that will likely persist for a decade or more. ASML has a complete monopoly on EUV lithography that will likely persist even longer. My company has a dominant position in AI chips with a software mode that grows stronger every year. These are not normal competitive situations.
The usual assumptions about mean reversion may not apply. Furthermore, the growth opportunity is enormous and extending further into the future than most analysts model. AI is not a one-year trend or a five-year trend. It is a multi-deade transformation of the global economy. Every industry will be affected.
Every company will need AI capabilities. The demand for AI infrastructure will grow for decades. Companies that are positioned at the foundation of this transformation deserve premium valuations. I would rather own exceptional businesses at premium valuations than mediocre businesses at discount valuations.
Over longtime horizons, business quality matters more than starting valuation. And these are exceptional businesses with structural advantages that are unlikely to erode anytime soon. Now, let me talk about risks because no investment is without risk and you should understand what could go wrong. The most significant risk is geopolitical.
TSMC is in Taiwan. ASML is subject to export restrictions that could tighten or ease depending on political decisions. My company sells globally and could be affected by trade tensions. A serious conflict involving Taiwan would be catastrophic for the entire semiconductor industry and for these investments specifically.
This is a real risk that cannot be diversified away entirely. There is also execution risk. Technology leadership requires continuous innovation. If TSMC falls behind in manufacturing technology, if ASML fails to develop next generation machines, if my company loses its edge in AI chips, these investments could underperform.
So far, all three companies have executed well, but past performance does not guarantee future results. Demand risk is another consideration. The AI boom could slow down. Customers could pull back on spending. The economy could enter a recession that reduces technology investment. These cyclical factors could affect near-term results even if the long-term thesis remains intact.
Finally, there is regulatory risk. Governments are increasingly focused on semiconductor supply chain security. This creates both opportunities and challenges. It could mean subsidies and support for domestic manufacturing. It could also mean restrictions on where companies can sell and who they can work with.
The regulatory environment is evolving rapidly and could affect any of these companies. Despite these risks, I believe the riskreward is favorable for long-term investors. The companies have dominant positions in the most important technology market in the world. They have structural advantages that are extremely difficult to replicate.
They are benefiting from demand trends that are likely to persist for decades. And while the valuations are premium, they are not unreasonable given the growth trajectories in competitive positions. Let me close with some thoughts on how to think about building a semiconductor portfolio. First, position sizing matters.
These stocks are volatile. They can move 20 or 30% in a matter of months based on news about AI demand, geopolitical developments, or earnings reports. If you cannot stomach that volatility, you need to size your positions appropriately. These should probably not be the only stocks you own. Second, time horizon matters.
The AI transformation will take years to fully play out. These companies will not realize their full potential in months or quarters. They will realize it over years and decades. If you need the money in a year, these might not be the right investments for you. If you can hold for a decade, the odds of success are much higher.
Third, diversification within semiconductors makes sense. Owning all three of these companies rather than betting everything on one reduces company specific risk while maintaining exposure to the AI theme. Each company has different risk factors. TSMC has Taiwan exposure. ASML has regulatory exposure.
My company has competitive exposure. Owning all three provides balance. Fourth, stay informed. The semiconductor industry moves fast. New developments can change the competitive landscape. Read the earnings reports. Follow the technology news. Understand what each company is doing and how the industry is evolving.
Informed investors make better decisions. These three companies, TSMC, ASML, and my company, represent the foundation of the AI revolution. They make the machines that make the chips that power artificial intelligence. Without them, there is no AI industry. Their positions are dominant, their moes are deep, and their growth trajectories are compelling.
I have been building semiconductor technology for over 40 years. I have never seen an opportunity like this. The convergence of AI demand, accelerated computing, and advanced manufacturing is creating a once- in a generation investment opportunity. The companies that control semiconductor technology will capture an enormous share of the value created by AI.
These three companies are positioned to be among the biggest winners. The numbers I mentioned earlier bear repeating. The AI chip market has grown from $5 billion to over 100 billion in just a few years. It could reach 500 billion or more by the end of the decade. The companies at the foundation of this market, the ones with dominant positions and insurmountable moes, will generate enormous returns for patient investors.
This is what I am betting everything on. the AI semiconductor stack, the companies that make it all possible, the picks and shovels of the most important technology revolution of our lifetimes. I have never been more confident in an investment thesis, and I have never seen a clearer opportunity for investors who understand technology and have the patience to let it play out.
The future of computing is being built right now by these three companies. The chips they design and manufacture will power everything from the smartphones in your pocket to the autonomous vehicles on the road to the AI systems that will transform every industry on earth. Owning a piece of that future seems like one of the best investments anyone can make.
That is why I am betting everything on these three semiconductor stocks. That is why I believe they represent the best opportunities in technology investing today. And that is why I wanted to share this perspective with you. The opportunity is here. The companies are identified. The thesis is clear.
What you do with this information is up to you. Let me leave you with one final thought about timing. Some investors wonder if they have missed the opportunity in semiconductor stocks. The stocks have already gone up substantially. The AI story is already wellnown. Is it too late to invest? My answer is no.
We are in the early innings of a multi-deade transformation. Yes, these stocks have appreciated, but the growth opportunity ahead is even larger than what has already been captured. The AI chip market in 2030 will be many times larger than it is today. The companies that dominate that market will be worth many times what they are worth today.
When people asked if it was too late to invest in Microsoft in 2000, the stock was trading at around $50, they thought they had missed the opportunity. Today, it trades at over $400. When people asked if it was too late to invest in Apple in 2012, the stock was trading at around $20 adjusted for splits.
Today, it trades at over $200. Great companies continue to compound over time. If the fundamentals are strong, if the competitive positions are durable, if the growth opportunities are real, it is rarely too late to invest. These three semiconductor companies have exactly those characteristics. Strong fundamentals, durable competitive positions, real growth opportunities that extend for decades.
The stocks may be volatile in the short term. They may have periods where they underperform the market, but over the long term, I believe they will generate exceptional returns for patient investors. The semiconductor industry has always been at the heart of technological progress. From the first computers to smartphones to the cloud, every major advance in technology has been enabled by advances in semiconductors.
Artificial intelligence is the next great chapter in this story. And these three companies TSMC, ASML and my company are writing that chapter. That is why I am betting everything on them. That is the investment thesis I have the highest conviction in. And that is what I wanted to share with you today.
Comments 0