Transcription
Over the past few years, the stock market has been on an absolute tear. And as we know, it's largely been on the back of seven companies that are all making moves in the world of AI. Profits have been rising, and investors have been flooding money into these companies, pushing up their stock prices and their valuations. But there's a pretty big problem, and the more I see these companies rise, the more nervous I get.
Now, to be clear, the share prices rising is not necessarily a problem. That can go on a very long time before it leads to issues. The problem I see is a bottleneck. It's one single choke point that has the potential to really hurt all of these AI companies. And with these seven stocks now accounting for over a third of the S&P 500 index, I think it's something that deserves a lot more attention because a problem affecting these seven companies could effectively bring the whole market down like one big house of cards.
So to explain this problem we're facing, we need to look at one of these seven companies in particular. That company is Nvidia. Now, as we know, Nvidia has performed exceptionally well over the past few years on the back of insatiable demand for its AI chips like the H100 or the B200. They're now the largest company in the S&P 500, accounting for over 7% of the index just by themselves. They're a $4.5 trillion company, and investors are expecting this growth to continue. They're giving them a PE ratio of around 50. Aka after 50 years of current performance, the business will earn enough for you to make back your money as an investor.
Now, that's a high valuation, and it means investors expect this company to grow, not to stand still. But fair enough, the company is growing quickly because Nvidia's chips are in hot demand from all the big companies around the world. In fact, out of the other Magnificent 7 companies, the only one not buying up Nvidia chips is Apple, who designed their own. But the overarching point is everybody wants Nvidia's chips.
Now, here's the thing. When you ask people what Nvidia do, they'll probably say, "Oh, Nvidia, they make chips, right?" When you listen to news reports or YouTube videos, that's also what you'll hear. Nvidia make chips. They're making the chips that are fueling the AI boom. But that's actually not true because Nvidia is not actually a chip maker. They are a chip designer. Nvidia engineers design the architecture of its GPUs. Everything from how cores process data, memory layout, power efficiency. They even design the software stacks that make their hardware useful. But they do not manufacture their own products. In other words, Nvidia's value is in the intellectual property. They draw the allimportant blueprints.
It's the same thing with Apple. Apple love to say that they now make their own chips, the A19 chip in the new iPhone or the M5 in the latest generation of MacBooks. But in reality, they are also just the designer. The company that actually manufactures the chips for both Apple and Nvidia is TSMC. That's it. Just one company. And this is where you start to see a potential bottleneck. The M series for MacBooks, the A series for iPhones, Nvidia's A100, the H100, the H200, the B200. All of the world's most desired chips. And the ones that are almost completely powering this AI revolution are manufactured by one company. That in itself is a really big risk.
But right now, you've got that risk compounded because what does TSMC stand for? It isn't Tennessee semiconductor. It's Taiwan semiconductor. And if you've been living under a rock, there's quite some concern that China will look to take control of Taiwan in the not too distant future. In fact, US intelligence believed that Xiinping is preparing the Chinese forces to invade by no later than 2027. We've seen China building and testing specifically designed new landing ships for just such an occasion. They're building 70 new commercial fairies capable of transporting troops and armored vehicles scheduled to be completed by the end of 2026 and in the last few months have been conducting large-scale military drills in the waters around Taiwan.
Now, I'm not necessarily saying that China is going to actually follow through and do it, but I do find it a little concerning that the entire AI industry is propped up by this one company that manufactures 80 to 90% of its chips in Taiwan and has even admitted that moving its fabs out of Taiwan would be basically impossible.
Now, I know what you're thinking. Okay, that sounds like a big risk, but I'm sure if something happened to TSMC, then Nvidia and Apple could just use another company to make their chips, right? Take their designs to another manufacturer. This sounds logical in theory, but the reality is it's a much tougher problem than people think because the types of chips that Apple and Nvidia are designing, the 3 and 4 nanometer chip generation, can really only be manufactured by two companies in the world, TSMC and Samsung. Why? Not to get too into the weeds, but the 3 and 4 nm generations are the most advanced chips in the world. They will soon be overtaken by the 2nm generation. It's a bit of a mess because the naming convention used to be about transistor size. Now it's kind of morphed into just marketing speak. But the idea is the smaller you can make the transistors, the more you can fit onto a chip and the more powerful and efficient the chip becomes.
Now for scale, a human hair is around 70,000 nm thick. So, these transistors are 3,000 to 4,000 times smaller than that. To make transistors that tiny, you need an insane level of precision, and there's only one company on Earth that even makes those machines, ASML, based in the Netherlands. Each one of these machines costs around $300 to $400 million, and there are only a few hundred out there in existence. These machines are owned by various companies around the world, but mostly by TSMC and Samsung. But the kicker is, even if you do have one of these $400 million ASML machines, you still can't make the 3nm chip unless you have the decades of experience, the software, and the supply chains that go with it. And then to make these chips at volume, only TSMC and Samsung can do it.
Okay, so send the orders to Samsung instead. The problem is Samsung, literally the next most advanced foundry, still struggles to match TSMC's production quality. Their 4 nanometer processors had big yield issues, meaning a big chunk of their chips come out defective. And honestly, when you're producing a $30,000 GPU or a $200 iPhone processor at scale, a 20% yield loss is just too painful. Nvidia learned this the hard way during the pandemic because it actually did use Samsung's 8nm processor for its RTX 30 series GPUs, but the yields were so poor that Nvidia went straight back to TSMC for the next generation.
So to sum it up, if something happened to TSMC, you can't swap them out. There's no viable backup plan. There isn't a ready to go fab to hire in Texas that Nvidia can rent out if the Taiwan situation escalates. And that's what makes this whole AI boom so fragile.
And you might say, but hold up, isn't there a massive push for chips to be made in the US? And isn't TSMC building a chip fab in the US? Yeah, that is correct. TSMC is building chip fabs in Arizona to be specific. But the thing is, just because you build the factory doesn't mean you can instantly start pumping out the world's most advanced chips. For starters, those new Arizona fabs are years behind schedule. The first one was supposed to start producing chips in 2024, but it's now been pushed back to late this year or even next year. And the second site might not be ready until 2028. These fabs are extremely complex. I don't know if you've ever seen footage of them, but they are not like a car assembly plant. These places look like something straight out of a sci-fi. The air is cleaner than an operating room. Workers wear full body bunny suits to avoid shedding a single speck of dust because even a single particle can destroy an entire wafer worth millions of dollars. We're talking about facilities that require investments of tens of billions of dollars, require thousands of specially trained workers, and need ultra specialized supply chains of gases, chemicals, and precision machinery that just isn't established locally in the US.
So, that's hurdle number one, and that's not even mentioning the talent problem. TSMC's engineers in Taiwan have decades of experience working up to the most advanced 3nm and 4nometer nodes. But when they brought that operation to Arizona, they discovered that the local workforce just doesn't have that level of expertise yet. Engineers even had to be flown over from Taiwan just to get things working. So you really can't replace that technical knowhow overnight.
And then of course the final problem, even if these fabs do finally come online, they won't be building Nvidia's top tier AI chips right away. The plan is for the Arizona plants to make chips on TSMC's 4nanmter process, while the latest and most in demand chips like Apple's A19 and Nvidia's B200 use 3 nanometer technology soon moving to 2nm technology. So, it's going to take quite some time before that true state-of-the-art manufacturing is happening at scale in the United States.
So, yes, there's progress. The US is absolutely investing in chip manufacturing through the chips act and companies like Intel and Samsung and TSMC are all expanding on American soil, but it's not something that can happen in a couple of years. Building up that ecosystem, the people, the materials, the suppliers, the precision tools. Ultimately, it's going to take many, many years, potentially a decade.
And that's why when I asked Steve Eisman about this, a man pretty good at predicting stock market bubbles, he said this. Do you think the biggest thing we have to watch out for at the moment is the geopolitical? >> Yes. Okay. >> Definitely. >> The reality is there is a massive dependency on Taiwan and TSMC.
But what happens if China does invade Taiwan? This is where things get very interesting. Now, of course, we are just speculating at this point as the truth is we don't know what would happen. But TSMC has said for years that if China invaded, they wouldn't be able to continue producing chips for the West under Chinese occupation. even if they wanted to. It's just not that simple. As I was saying, their facilities depend on global inputs, high-end machinery, specially designed software from the US, and so on. If China attacked, those supply lines would be instantly cut off. So, even if TSMC's fabs weren't destroyed, they'd basically be dead in the water. They wouldn't be able to get the parts, the tools, or the support they'd need to keep running.
And this is why the US is watching so damn closely. And it's why America would almost certainly get involved if China invaded Taiwan. In fact, the US has made it very clear that they would rather see TSMC's equipment disabled than fall into Chinese control. There are even whispers that Washington has contingency plans to evacuate key Taiwanese engineers or remotely disable the factory tools so that they can't be used by China.
And from China's perspective, if they took control of Taiwan, it's unlikely they just keep exporting the world's most advanced chips to American companies. Wishful thinking, but I think it's highly unlikely. I mean, you think about it. China already controls over 90% of the world's rare earth refining. And they have not done that by accident. They've done it for power and control. It gives them massive negotiating power and massive leverage. They control a critical ingredient in EVs, batteries, and defense equipment. If they also controlled TSMC, that's a huge economic move and it would really weaken the West. It's really quite crazy. They take this one little island off their coast and the West is in big trouble.
And that's why I think we as investors need to be really careful about this one choke point of the whole AI economy. remove one tiny little island off the coast of China and Nvidia's chips stop, Apple's chips stop, and then you have to deal with the flow and effect into Amazon, Microsoft, Tesla, Meta, Google to an extent, and they're just the big ones.
But let me know what you think. Again, I don't want to come across as fear-mongering. I'm not predicting the end of the world here, but I think it's certainly something that should be thought about probably a little bit more than the euphoric market is currently thinking about it today. So, let me know what you think down in the comments section below. Am I crazy? I'd love to hear your thoughts. But apart from that, please leave a like on the video if you did enjoy it. And with that said, I'll see you guys in the next one.