Deep Dive
Shownotes Transcript
I'm Ricky Mulvey, joined on the internet today by Tim Byers. Tim, good to see you. Fully caffeinated, ready to go, Ricky.
Today's a caffeine day. Okay, I'm good. I got lucky. Today is a caffeine day. All right, let's start with Taiwan Semi and see how caffeinated you are about this one. All right. We can talk about the business results, but I thought there was some swagger in this earnings report because here comes a large semiconductor company that says, we're still expecting mid-20s growth in our outlook. We're maintaining the goals we set in January, never mind the tariffs.
What do you make of management over at Taiwan Semiconductor maintaining this much swagger in a trade war?
Yeah, big earnings energy. In some ways, you are seeing what it looks like when a well-positioned company has a hammerlock on an important and emerging market. That's what this is. Taiwan Semi is still far and away the leader when it comes to global chip manufacturing, Ricky. This is especially true at the nanoscales where the highest performing chips are made. If we want to be more specific about this, just going to the conference call,
CEO C.C. Way said in that call that TSMC essentially has seen no impact
from tariff policy. So it's a little early and there is some swagger in this, but I just want to quote him here because I think this is an important quote. So we understand there are uncertainties and risk from the potential impact of tariff policies. However, we have not seen any change in our customer's behavior so far. Therefore, we continue to expect our full year 2025 revenue to increase by close to mid-20s. So like, yeah,
Like, no change in behavior. And the translation I have for that, Ricky, is just the tech world needs our production capacity and expertise, and we're going to keep charging a premium for access to it. So, it's a heck of a position to be in. So, no change in behavior yet. Still a relatively fresh trade war. We're not that far from Liberation Day. But in the meantime, we're going to focus on the business results. This is really impressive to anyone. When you're a company that's worth more than, what is it, $600 billion, right?
It's a lot to move the needle. Net income rising for Taiwan Semiconductor by 60% from the same period last year. Its high-performance computing division is now about 60% of total revenue. That's a lot of cheddar. What's behind these business results?
Well, I think we know the answer, and it is AI. Demand for hyperscale AI compute continues to be robust. And Wei said this, he said that recent breakthroughs in introduced efficiencies, so we're talking about things like DeepSeq, only serve to democratize AI and increase the need for chips and compute that TSMC manufactures. By the way, he is also sort of
Not saying exactly what Jensen Wong has said, but it's awfully close to what Jensen Wong has said. He is parroting that message. But again, I think he's got a point here. You do not need to take my word for it. We can look at the numbers. So in Q1, 58% of wafer revenue. So this is revenue directly from manufacturing the wafers that hold chips. You have wafers and on those wafers are a bunch of chips. And when you separate the chips out of the wafer, that's how semiconductors are manufactured. So,
58% of that revenue was sourced at 5 nanometer or below. So that's the scale at where Taiwan Semi is producing the ultra-high performance chipsets. And many of these are sure to be purpose-built for AI training and inference. So that's what's going on here. This is an AI tailwind. So one of the benefits of these very small chipsets
three nanometer, five nanometer chips, it means that you can get the same performance for a lot less power or more performance for the same power. And it's worth noting just the tech we're talking about here. When we're talking about one nanometer, that is about the length that your fingernail grows in one second.
And these chips are so high. And if you think about it, like you're not noticing how much your fingernails are growing in one day. Hopefully you're noticing how much they're growing in one week. But this is insanely small stuff we're talking about. A lot of transistors going on to one chip.
Battery life is one example. So if you have a newer iPhone, you may notice your battery goes on a lot longer than older iPhones. That's because they're packing on more transistors onto one chip. But whenever we're talking about Taiwan Semi, it's probably worth bringing NVIDIA into the conversation. That's a big customer for Taiwan Semi. What can NVIDIA do with these super small chips?
You've got this right. More transistors in the space that you are producing the chip. And the multiplier effect here is, so a GPU, a graphics processing unit, which is, that's NVIDIA's core business. You are pulling together quite a lot of relatively dumb stuff.
high horsepower chips. And so, a more efficient GPU built at smaller nanoscale is going to pack more of that dumb processing power into the same relative space. So, it's just like, if we go the fast and the furious analogy here, Ricky, you got a whole lot more nitrous.
is what you get here. So the technology is pretty complex here. And I asked, so like, how much efficiency are we talking here? So I asked Jim and I to describe the difference between three nanometer and five nanometer, just to kind of put a finer point on this. And what it said was a three nanometer chip
It'd have a transistor density of nearly 300 million transistors per square millimeter. A 5-nanometer chip has a density of 130 to 230 million transistors per square millimeter. It's not quite a doubling. Let's say it's like 70% better, and that's just going down two nanometers. You can see how this scales up pretty quick.
Last thing I'll note about Taiwan Semiconductor, what stood out to me, this is a company that maintains close to a 50% operating profit, which is a lot for a manufacturing company. Tells me, Taiwan Semiconductor, still a lot of pricing power there.
And we're going to talk more about semiconductors later in the show. Tim and I have a deep-ish dive on what's going on with the NVIDIA export controls. But for now, actually, we're still talking about pricing power because we're going to talk about Netflix, which reports later today. But...
It hosted its annual business review meeting just a few days ago. This is what the Wall Street Journal has reported on and what I want to talk to you about because they're really zooming out on the business here. Management basically telling its investors, this is what we plan to do by 2030. We are going to triple our operating income. We are going to double revenue, and we will achieve a $1 trillion market cap.
Right now, Netflix is about a $430 billion company. It needs to more than double by 2030. That's a lot. I want to start from a place of good intentions, Tim. What needs to go right for this to happen? Global ARPU, that's average revenue per user, needs to go up as the ad tier gets successful in other territories around the globe. They need to get footholds in other places.
Another great sign would be more content finding its way across borders. Essentially, in order to do this, Netflix doesn't need to do new things. That's the point I want to underscore here. The reason I believe them is they don't have to do new things. What they have to do is take their global customer base and get more from that global customer base and then continue to grow that global customer base
as they have been. The ad tier is going to be essential to this, Ricky. I really believe that. But I will say this idea of more content finding its way across borders
Do not underestimate how important that point is. Every time content grows -- let's take The Queen's Gambit, one of my favorite recent Netflix shows. Well, it's not that recent anymore, but it's relatively recent, the last few years. Great show, has gone around the world. Every time content goes multinational, it creates a massive ROI on the original spend.
That's really important. The internal return on capital employed is really good for Netflix. This is why their margins can go up. Every time this happens, they just get fantastic
internal returns on that content. If that continues, they can keep this up. And so you would expect to see along the path here, Ricky, more global content, more global content that they can market across borders. So you could have
Chinese language content that doesn't just appear in China, but appears in Chinese speaking territories, right? You know, so you would expect things like that. But yes, they don't do a lot of licensing and I'm not really expecting them to, but that's something that is open to them later on in their life. They just don't need to do new things to do what they're talking about.
To do what they're talking about, though, is pretty incredible. To join the legion of...
trillion dollar companies we're talking about like Apple Nvidia Microsoft I think Berkshire Hathaway is up there right these like at the end of the day Netflix is an entertainment company with a subscription business it's it's consumer spending it's not a large conglomerate it's not doing it's not the backbone of cloud computing this is while pretty recession proof this is a cost that people can cut and has some pretty fierce competitors
Sure. I'm a Netflix shareholder, but I don't know if I'm buying what management is selling here. How about you? No, I am. And I think the reason for this is because Netflix is the one streaming service that's been profitable and consistently so for years. As others have tried to scale up, Netflix has just grown more and more profitable. They've been doing this better over
than everybody else for a long time. The global infrastructure that everybody else needs, Netflix already has. That is a huge advantage for the company. I don't think they ever get enough credit for that. Let's be clear, Ricky, unless it is something that is marketed here in the United States, we don't really see
all of the amazing stuff they're doing in other territories. Like we see Narcos, we've seen squid game, but yeah,
Generally, the hits that they have in other territories, we don't really see that. But they do have hits in other territories. I was in Costa Rica in January and pulled up Netflix there. The movie selection when you get out of the United States, absolutely incredible. If I were outside of the United States, I could not imagine ever canceling Netflix if I had the means to have it.
Right. I think this is important. I also will say this, Netflix gets the benefit of -- they get benefits from their competitors. Here's the primary benefit. Every time you have a money-losing streaming service, and that is essentially all of them, not named Netflix, feeding the habit of streaming, what you're doing is saying,
Look, streaming is the default and Netflix is the big dog. You are advertising for Netflix by doing that, by creating and helping feed the paradigm shift. So yes, I think they are the big dog. I think they have a significant advantage in global distribution that their competitors are still a little, a little ways from matching.
One thing I've noticed about Netflix right now and the commentary around it, no one seems to be bearish on it. Just a few years ago. Yeah, that's a worry, isn't it?
Netflix lost some subscribers, and there was a huge sell-off in the stock. Now, management, a few years later, very politely has said, we're no longer reporting subscriber numbers. Times are good. This is a pretty safe haven from the tariff madness going on. A Netflix subscription is one of the last things to go in a recession for people.
But because times are so good, because people are so bullish on this company, you can't find an analyst with a bearish thing to say about it.
Let's play the bear case. What would break your thesis for owning shares of Netflix? Yeah, the thing that would really screw it up for me, Ricky, is what Peter Lynch famously called diversification. That's where you are taking the core business and saying, yeah, this core business is great, but there are other things we could do, and you layer other things on top of it.
So, for example, they are experimenting with games, but they are not presently over-investing in games. I find that to be good. If that changes and they start over-investing in games, I'll get a little nervous. Or if they build Netflix land, and just because they have the cash on their balance sheet to create a theme park, that doesn't mean they should do it. I don't think they will. If they did, I would find that really, really worrying.
Netflix is fantastic at greenlighting small budget shows around the globe, a whole lot of them. And then nurturing, there's a lot of losers in that group, but they nurture just enough of the winners into viral hits that cross borders, that produces very high internal returns. That's what they need to do. Just keep doing that, and I think they'll be fine. Let's leave it there. Tim Byers, appreciate your time and your insight. Thanks for joining us on Motley Fool Money. Thanks, Ricky.
Today's show is brought to you by the Range Rover Sport. The old adage goes, it isn't what you say, it's how you say it. Because to truly make an impact, you need to set an example. You need to take the lead. You need to adapt to whatever comes your way. And when you're that driven, you drive an equally determined vehicle, the Range Rover Sport. Blending power, poise, and performance like you, it was designed to make an impact. With seven terrain modes to choose from, Terrain Response 2 fine-tunes the vehicle to take on challenging roads ahead.
A force inside and out, the Range Rover Sport was created with a choice of powerful engines, including a plug-in hybrid with an estimated range of 53 miles. Defining true modern luxury, the Range Rover Sport includes the latest innovations in comfort and convenience. Use the cabin air purification system alongside active noise cancellation for all new levels of quality, comfort, and control. Free from unnecessary details, raw power and agility shine in the Range Rover Sport.
Build your Range Rover Sport at rangerover.com/us/sport. All right, up next, Tim is sticking around to talk about the export ban that Nvidia is facing and the tech giant's current valuation.
All right, Tim, before we get to the ban, I think it's worth discussing what is being banned here? What is the H20 chip and where does it stand in sort of Nvidia's lineup of offerings? Yeah, so Ricky, the H20 looks to me like a customized for the Chinese market version of their...
advanced AI chipsets. I think you could think of this as the knockoff that you would buy at a dollar store. It's for the lower income, can't afford the full brand, but you can have the
low-brand equivalent that you could pick up at your local dollar store. I think in this particular case, it's because the American government has been very clear for a while now -- this is not just the current administration, this is the prior administration -- they just said, "Look,
China can't have the most advanced American chip technology, so you have to give them something else. But now, it does appear that there are new licensing requirements. These licensing requirements mean that all of those H20 chips that were going to be sold, now it appears $5.5 billion worth.
are now not going to be sold into China. And they are the low-budget, doesn't really work anywhere else type of chips. And so, it becomes a write-off for Nvidia. I mean, that's a pretty big hit.
Yeah. The U.S. government, even through the Biden administration, as you mentioned, has restricted, or I should say tried to restrict, the sales of advanced chips to China. This is big news for Nvidia. It's a big write-off. What's so new about this band versus what had come previously? I think the licensing requirements.
I have not seen anything that shows what these licensing requirements are. That's what the Trump administration has introduced, and it's affected not just NVIDIA, but it's also, in fact, affected AMD. So AMD now is going to write off $800.
million worth of inventory and product that it can't sell now because of these new licensing requirements. What it looks like, Ricky, I can't say that I know exactly what it is, because I'm not so sure it's been fully defined yet, but it looks like new requirements, new hurdles that the
these chip manufacturers are going to have to go through in order to sell to China. Or, it may just be that this is effectively a ban, and there are going to be no sales to China. I wouldn't predict that. I would rather say that
Whatever these new hurdles are, AMD and Nvidia will figure out how to meet those hurdles and then provide some kind of product into the Chinese market. I know you said this was the dollar store version of what Nvidia provided. The dollar store version of what Nvidia provides is still probably really good. I shop at dollar stores, Ricky. The dollar store versions are still good. But why can't Nvidia sell this H20 chip to other customers?
Well, I think because it's more customized and also because, just technologically, the power requirements, just some of the ways that the chip is limited, just isn't going to work for markets where they are looking for a much more advanced chipset. I mean, I think it's more complicated than that. But bottom line, I think there are other markets that have moved on to more advanced platforms. These are not highly power-efficient chips.
They are memory hogs. They just aren't designed to perform in the way that the more advanced chipsets are. Other markets are going to be like, "Yeah, thanks, but no thanks." This is more of a take, but I think it's important to bring in anytime we're discussing export bans. If the Chinese government really wants these chips, if they really want these
AI chips, they can still probably get them. They might have to do some work with shell companies. They might have to buy them outside of China, but then bring them into China. But this is something they've been dealing with for a while. Anna Swanson at The New York Times has reported in the past about how many of these super advanced chips and GPUs are still able to leak through export bans. Even though you set the rule, it may not be followed in the way you intended.
I think that's exactly right. And I think just so we aren't too naive on this show and so that we don't, you know, let members believe that we are too naive and,
The most advanced chips will still find their way into China, because that has been true for generations. How that happens is something that the U.S. government is going to continue to worry about and continue to crack down on. But yes, we should not pretend that the Chinese will not get access to sophisticated AI chipsets.
Meanwhile, they are trying to build up their own industry. Huawei is trying to build up its own supercomputer. You shared with me a newsletter that described where things stand. In your view, where does Huawei's supercomputers stack up to what Nvidia is providing right now? Full credit to Ben Thompson at Stratechery and his newsletter. He was talking about this, about the 384 supercomputer, which is
highly inefficient, but also highly powerful supercomputer designed off of their own chipsets. But I think the key here is that the Huawei supercomputer is something that we would not tolerate here on American shores because it is an absolute power hog. It is absolutely abusive in its use of memory. You
You would essentially be draining extraordinary amounts of power in a data center off Lake Erie somewhere, and you'd have houses flickering in the background because it was sucking so much power just in order to power these supercomputers. It just doesn't work for us.
But in China, they have different needs, their pollution laws are different. It works for China in order to just be a raw horsepower type of inference machine. For their purposes, they can do that. It's just not something that is likely to exist in a lot of economies outside of China.
One of the moats that Nvidia has with its chips and the CUDA system that it runs on is that they only play nicely together. If you want access to the system, you have to use Nvidia chips, which is where a lot of AI architecture is built.
Do you see competitors, maybe not Huawei, but other competitors that are going to build a more open offering in the near future? It seems like tech platforms are moving more towards that -- not moving towards, there's a lot of open-source AI types of offerings, but not really within the architecture. Yeah, in this particular case, Thompson makes the argument, and I think he's right about this, is that the export bans have created a counter-incentive
So, at some point, you do want to procure the most advanced American chips where you can get them, whatever ways you can get them. But you would also like to have your own alternatives. And in particular, you would love to have
an open-source alternative to CUDA that would allow you to program any NVIDIA chips you could get your hands on, and any other chips you could get your hands on. You're right, NVIDIA is highly vertically integrated. In order to best orchestrate and take advantage of NVIDIA GPUs, you use the CUDA low-level programming tool to do that.
That had been a real strategic advantage for Nvidia. If you want to have Nvidia chips, you got to use Nvidia CUDA. If you could do anything to break that hammerlock, that is good for China. But also, the real judo throw here, Ricky, which Thompson makes in his newsletter, and I think he's right about this, is once they do that, they will do what they did with DeepSeek. As soon as DeepSeek came out and they said, "Hey, look,
We did something. By the way, one of the most interesting parts of DeepSeq was that they bypassed CUDA in order to make the DeepSeq model really super interesting and build it on top of NVIDIA chips by building down a level below where CUDA was and dropping into something that was equivalent to a
assembly code. You could think of it like programming directly into the silicon. Really hard things to do, but to make a more efficient model. Similar idea. Could you create something that is an entire alternative to NVIDIA and then just throw it out into the world and say, "By the way,
We just dramatically weakened NVIDIA's moat. Here's a free tool that's exactly like it, and you could do it to program any GPU, including NVIDIA GPUs. There is a real incentive to do that now. Is it going to happen tomorrow? Is it going to happen within three years? I have no idea. But this is an example of...
when you have an economic force in place like export controls, you create incentives to get around those export controls any way possible. And in this case, it makes me worry
A little bit, not a ton, but just a small amount about Nvidia's moat. To be clear, this race towards AI supremacy was going on long before these export controls. The point you seem to be making is that maybe this speeds up, let's say, the shipments of sand going into Nvidia's moat to build islands there. I think that's right. I also think that the tougher the administration gets,
at trying to crack down on China and prevent it from getting the most advanced AI equipment, the greater the incentive there is for the Chinese government, Chinese entrepreneurs, and entrepreneurs around the world to figure out how to get around those restrictions. I actually like hearing your concerns about this, because Nvidia is still the top dog. It is a dominant player. As I'm looking at the price tag on this stock right now,
Nvidia is at about 42X cash flow. It had flirted in the past with about 80X cash flow. The PEG ratio, which is a price-to-earnings multiple that accounts for growth, is now at about 0.2X. For those investors who have been waiting for the dip to pick up some Nvidia shares,
I remember when it was the hottest thing, it was trouncing the market cap of McDonald's in one trading session. Now investors are getting much more pessimistic. That might be a good thing if you've been waiting to pick up some shares of this super dominant tech stock. What say you, Tim? I do think Nvidia is going to continue to be dominant for quite some time.
What this does is it introduces the first real grain of uncertainty in Nvidia that I think we've seen in a while. There's going to be some real movement to see what
other countries and what other companies can do to break the hammerlock that Nvidia has had on this market. And that uncertainty creates downward pressure. So, yes, if you haven't ever bought any Nvidia shares, it may be an interesting time to at least think about it. What I will say, though, Ricky, is, those multiples you just introduced there, there's still that 0.2 peg ratio.
That only matters if the growth rate continues as it has. And if the growth slows, then that kind of invalidates the peg ratio. So, if you're going to speculate, please just be careful. I hate to sound like your old granddad here, but be careful, son. This is still a speculative stock.
I've got other companies on my watch list. But, yeah, Nvidia is there, and probably rightfully so. Tim Beyers, appreciate you being here. Thanks for your time and your insight! Thanks, Ricky! As always, people on the program may have interests in the stocks they talk about, and The Motley Fool may have formal recommendations for or against, so don't buy or sell stocks based solely on what you hear. All personal finance content follows Motley Fool editorial standards and are not approved by advertisers. The Motley Fool only picks products that it would personally recommend to friends like you. I'm Ricky Mulvey. Thanks for listening! We'll be back tomorrow.