Semiconductor factory - Can Terafab unseat TSMC?

Kind: captions Language: en an iPhone, a Mac, [music] a PS5, a Tesla, an Nvidia server running the AI you use every day. What do these devices have in common? The chips that power them, the electronic brains that allow all of this to exist, are manufactured by the same company on the same island, just a few miles apart. You've probably never heard of this company in the mainstream press. It doesn't advertise. [music] It doesn't sell a single product with its name on it. And yet, if it disappeared tomorrow, your phone, your car, the data centers that run the internet, everything would come to a halt gradually, then suddenly. This company is called TSMC, Taiwan Semiconductor Manufacturing Company. According to the latest data from Trendforce, it controls 71% of the global semiconductor foundry market. More than seven out of every $10 spent on manufacturing advanced chips goes through them. It is the quietest and most powerful SH monopoly in the global economy. Last week we talked about Elon Musk and his Terrafab project. That crazy bet to build a factory capable of competing with TSMC. Today we're going inside the opponent to understand [music] what Tesla really wants to dethrone and why this is perhaps the most difficult industrial challenge ever attempted. In this video, how TSMC was born from an idea everyone thought was [music] absurd. its numerical dominance over competitors, what it knows how to do that no one else masters, its geopolitical Achilles heel, [music] and an honest verdict on the Terapab. Please subscribe now if you're discovering Blueprint Hub with this video. The button is right below. And thank you to those who made the Terapab video take off. To understand where TSMC stands today, why it is so strong, why no one can copy it, you first have to understand how it [music] was born. The story begins with one man, one idea, and an island the whole world was underestimating. Part one, the brilliant idea of 1987. The year is 1987. The semiconductor world is dominated by American giants, Intel, [music] Texas Instruments, Motorola, and Japanese ones. Taiwan is a developing economy known for cheap products, not technological prowess. A 56-year-old engineer, a former vice president at Texas Instruments, has just been recruited by the Taiwanese government for one mission. Build a semiconductor industry on the island. His name Morris Chang. Chang could have created just another classic chip maker, an IDM, or integrated device manufacturer, designing, manufacturing, and selling [music] its own chips as Intel and Samsung did. But he had a radically different idea that his contacts, [music] including Intel and Texas Instruments, refused to fund. He had to turn to the Taiwanese government and Dutch giant Phillips to raise the initial $220 million. The idea, never design your own chips, never compete with your customers. Be only the manufacturer. The Pure Play Foundry model was born. From day one, TSMC [music] committed to never having its own interest in the designs it manufactures. This model triggered a revolution. Suddenly, any engineer with a good idea could start a chip company without raising hundreds of millions to build a factory. Nvidia, Qualcomm, [music] and Broadcom all started with TSMC. Apple has fully entrusted its [music] chip manufacturing to TSMC since 2016. As Espectrum [music] noted, Morris's foundry model unleashed an army of engineers. Without TSMC, neither Nvidia nor [music] Apple Silicon nor the current AI revolution would have been possible. The result 38 years later, 465 active customers, 12,682 [music] different products manufactured using 300 five different process technologies. From iPhone [music] chips to Nvidia's AI accelerators to the processors in your Mac, your PS5, and SpaceX satellites. Morris Chang was 56 when he founded TSMC. He had just created the invisible infrastructure of the modern technological world. But a great founding story isn't enough. What has kept TSMC at the top for 38 years, far ahead of all its competitors, is a numerical dominance that when you really look at it, is dizzying. Part two. 71% of the global market. Domination by the numbers. In Q3 2022, TSMC holds 71% of the global foundry market. For the full year 2025, its revenue reached 122.54 billion, up 36.1% year-over-year, an increase of more than a third in a single year. Its closest competitor, Samsung Foundry at 7.2%, [music] 2% 10 times less than SMIC at 5.3%, UMC at 4.35%, Global Foundaries at 3.87%. If you gathered all of TSMC's competitors in one room, they still wouldn't equal half of what TSMC does by itself. What is perhaps even more striking is the trend. TSMC's market share isn't stable. It's increasing every year. Start of the decade in 2020, 64%. 2024 approximately 67% 2025 71%. This isn't a monopoly maintained by inertia. It's one that is actively deepening. While Samsung slips, global foundaries lose a share and Intel foundry remains marginal despite tens of billions invested. The reason technological lead nearly 3/4 of TSMC's revenue now comes from seven nanometer nodes and below. the most advanced, most profitable, where competition is virtually non-existent. Apple, Nvidia, and AMD compete to secure production slots. Apple has publicly admitted to being constrained by TSMC's supply, not its own demand. TSMC sets the pace, not its customers. Samsung, the most serious competitor on paper, struggles despite massive investments in 3 nanometer and 2nmter class. Its yield rates remain below TSMC's. Broadcom tested Samsung wafers and chose TSMC. Qualcomm did the same. It's not a matter of image. It's measurable, costly reliability. And TSMC wafer prices have been rising steadily for years. Yet, the customers keep coming. When you can raise your prices and keep your customers, you're no longer a supplier. You are infrastructure. [music] But behind these impressive numbers lies a technological reality that is even more fascinating. The real question isn't how much TSMC [music] sells. It's why no one else can do what they do. Part three in what TSMC knows how to do [music] that no one else truly masters. In December 20 to 25, TSMC confirmed the start of volume production on its N2 [music] 2nanometer process, the most advanced node ever reached in commercial production using nano sheet [music] transistor architecture for the first time. The biggest transistor evolution since finfets in 2011. Compared to the previous three nanometers, [music] 10 to 15% more speed or 25 to 30% power savings. Capacity ramps from 35,000 wafers per month [music] in Q4 2025 to 140,000 wafers per month by [music] end of 2026, well above initial projections. Apple has already secured more than half of this initial capacity for its A20 and M series processors. But the most strategic battle is being fought in advanced packaging. COS chip on wafer on substrate. The technology that assembles multiple ultra- high performance chips [music] and HBM memory sticks into a single ultra dense package. Without co-was, the Nvidia GPUs powering every AI data center in the world don't exist. As AMD stated in a rare moment of cander, everyone in Taiwan knows about co-was. That is the real bottleneck of global AI. Nvidia [music] has secured over 70% of TSMC's co-well capacity [music] for 2025 with 800,000 to 850,000 wafers reserved for 2026. Demand is growing at 113% per year [music] and lead times remain fully booked. But TSMC's true competitive moat, the one hardest to bridge, is neither 2 nanometers nor coowwas. It is the ecosystem, what its SVP calls the 1h hour semiconductor ecosystem. All strategic suppliers, [music] equipment manufacturers, chemical suppliers, EDA software are reachable within one hour of TSMC's [music] Taiwanese fabs. When there's a problem on a production line, the supplers's engineer is on site within the hour. A density of expertise [music] and responsiveness that no other place in the world can replicate. Compare this to Arizona. suppliers scattered across thousands of miles, permits taking twice as long, engineers driving across cities to reach partners. Morris Chang himself called the US effort a very expensive exercise in futility. This ecosystem was [music] built over 38 years. You don't duplicate it in three, no matter the budget. TSMC [music] doesn't sell chips, it sells trust. And in an industry where one bad batch of wafers can cost hundreds of millions of dollars, trust is priceless. If TSMC is so extraordinary, how can we talk about an existential vulnerability? The answer comes down to one word, Taiwan. Part four, the Achilles heel. Everything rests on an island 100 miles from China. Here is the most staggering number in this entire video. Not the 71% market share, not the 122 billion in revenue, not the two nanometers. That number is 100 miles, 160 km, the distance separating Taiwan from the Chinese coast. On this 14,000 square mile island, more than 90% of the world's advanced chips, those at 7 nanome and below are concentrated. The chips running your iPhones, Amazon's and Google's data centers, the most advanced US weapon systems, and Nvidia's global AI infrastructure. All of that on one island 100 miles from China. Bloomberg economics modeled this scenario. A major disruption to Taiwan and TSMC would cost the global economy approximately $10 trillion. More than total CO 19 economic losses in a single disruption. Prediction markets currently estimate a 22% probability of conflict by 2027 in [music] the Taiwan Strait. And global stocks of advanced chips represent an average of only 4 [music] to 8 weeks of consumption, not months, weeks. This is what's known as the silicon shield. Taiwan is so indispensable that no one, US, China, anyone has an interest in causing a disruption. Global dependence is [music] a form of protection. But this shield has a cruel paradox. The more TSMC invests outside Taiwan, the more the shield erodess. In March 2025, TSMC [music] announced $100 billion in US investment, five new fabs in Arizona, plus a fab in Japan with Sony and Toyota, and one in Germany with Bosch and Infinian. But the most advanced chips, current two nanometers, won't be produced in Arizona until 2028 [music] at the earliest. Until then, Taiwan is the only place in the world where they exist. And this is where it all connects. The Silicon Shield paradox is precisely what gives [music] Elon Musk's terapab its strongest strategic logic. Washington does not [music] want to depend on an island 100 miles from China for its most strategic chips. The chips act has already distributed 6.6 billion to TSMC Arizona. An American company saying we can build a 2 nanometer fab on US soil is exactly [music] the message Washington wants to hear. And Washington has the money to reward that message. Now that we know TSMC from the inside, its strength, its ecosystem, its dominance, and its vulnerability, let's ask the real question. What does all this reveal about Elon Musk's real chances? Part five, Terraab versus [music] TSMC, an honest comparative verdict. Here is the honest verdict. No fan service, [music] no denial, just the facts side by side. On experience, TSMC has 38 [music] years of continuous production. Tesla has zero. On technology, TSMC has been producing 2 nanometers since [music] December 2025. Tesla targets mid 2028 at best, according to Morgan Stanley, a minimum 3-year lag. On supplier ecosystem, [music] TSMC has 38 years of integrated relationships. Tesla must build everything from scratch. On funding, total investment is estimated [music] between 35 and $45 billion by Morgan Stanley. Fundable but challenging against Tesla's $6.2 billion annual free cash flow. On geopolitics here, Tesla has a real advantage. US soil, chips [music] act, perfect timing to secure massive Washington subsidies. The honest conclusion, on a pure technological level, [music] the Terapab starts with a three decade disadvantage. Intel, once the global leader, is still struggling [music] to catch up. Despite $100 billion invested and thousands of experienced engineers, Tesla wants to go from zero to the world's largest 2nanometer [music] fab. Literally unprecedented in the industry's history. But the most serious analyses nuance this picture. Trendforce notes that Tesla's [music] most realistic entry point would be advanced packaging rather than two nanometer logic manufacturing where the learning curve is less steep. And Morgan Stanley considers the Terrafab the right strategic move despite the ambitious [music] timeline. So the real question isn't can Tesla beat TSMC? The answer is no. Not in the next 10 years. The real question is can Tesla become capable enough to no [music] longer depend on TSMC for its most strategic components. Those are two radically different challenges. And the second starting with packaging building a base of experience gradually reducing dependence [music] is perhaps attainable especially with Washington behind it. So when after everything we've just seen what's the real takeaway conclusion the world needs TSMC [music] and that's exactly the problem. Let's go back to the beginning. An iPhone, [music] a Mac, a Tesla, an Nvidia data center. They all share one thing. A company founded in 1987 by a 56-year-old engineer on an island the world underestimated. A company that invented a model no one wanted to fund and which 38 years later controls 71% of global advanced chip manufacturing with a technological lead that even Intel with 30,000 engineers and 100 billion [music] has failed to close. TSMC is currently irreplaceable. Rarely has that word been so literally accurate. But irreplaceable today has never meant irreplaceable forever. Nokia was irreplaceable. Kodak was irreplaceable. What makes companies invulnerable one day often makes them blind [music] the next. And TSMC's vulnerability, that island 100 miles from China, is real, known to everyone, and [music] impossible to resolve quickly. What Elon Musk understood and what many of his critics have yet to realize is that the Terraab is the most [music] ambitious semiconductor bet in history. But wild ambition is sometimes right where reasonable pragmatism is wrong. The Terapab won't beat TSMC in 2028. Probably not even in 2030. But if it succeeds in producing AI5 chips in small batches, [music] mastering advanced packaging, and building a base of experience, it will have changed something. It will have created real pressure on TSMC, reduced US strategic [music] dependence on an island under geopolitical tension, and proven that Morris Chang's pure play foundry model can be challenged. To understand TSMC is to understand why the Tarap is so hard. And to understand why the tariff is so hard [music] is to understand why it is necessary. The two stories are inseparable. The question I'm dying to ask, and I really want to read your answers. In 10 years, will TSMC [music] still have 71% of the advanced chip market? Or will the world have succeeded in creating a real alternative? Whether it's Tesla, Intel, [music] Samsung, or someone we don't even know yet. Let us know your [music] prediction in the comments with your arguments. If this video taught you something, share it. It only takes 2 [music] seconds and it helps Blueprint Hub grow. If you aren't subscribed yet, the button is [music] right below because in our next video, we're going to talk about Samsung Foundry, the company that had everything it needed to compete with [music] TSMC and yet didn't make it. The most valuable lesson there is for Elon Musk. [music] Subscribe so you don't miss