Motley Fool Money - Nvidia Sells Chips Like Pancakes

Episode Date: March 19, 2024

The chip designer unveiled a bigger and faster graphics processing unit, but the market isn’t cheering the announcement. (00:21) Asit Sharma and Ricky Mulvey discuss: - The advancements in Nvidia��s new Blackwell GPU. - If any competitors are closing in on the tech giant. - Nvidia’s plays in humanoid robots, self-driving cars, and weather forecasting. Plus, (16:14) Alison Southwick and Robert Brokamp answer listener questions about lending out shares, activist investors, and Social Security. Stock Advisor discount link for podcast listeners: www.fool.com/asit Got a question for the show? Email us at [e-mail] Companies discussed: NVDA, INTC, AMD, GOOG, GOOG, AMZN Host: Ricky Mulvey Guests: Asit Sharma, Robert Brokamp, Alison Southwick Producer: Mary Long Engineers: Dan Boyd, Rick Engdahl Learn more about your ad choices. Visit megaphone.fm/adchoices

This episode is brought to you by Indeed. Stop waiting around for the perfect candidate. Instead, use Indeed sponsored jobs to find the right people with the right skills fast. It's a simple way to make sure your listing is the first candidate C. According to Indeed data, sponsor jobs have four times more applicants than non-sponsored jobs. So go build your dream team today with Indeed. Get a $75 sponsor job credit at Indeed.com slash podcast. Terms and conditions apply.

How many CEOs are doing arenas these days? You're listening to Motley Full Money. Ricky Mulvey, joined today by Asset Sharma, Asset, good to see you. Ricky, good to see you. So, you know, the first question is in the intro of the show. How many CEOs can do an arena? I've got Jensen Wong and Warren Buffett, anyone else on the list, because I don't think so. Yeah, right now I'm down to those two.

Okay. So the big story is this GTC. conference where Nvidia CEO Jensen Wong delivered a keynote address in San Jose at the arena where the sharks play. It had original music composed by artificial intelligence and played

by the London Symphony Orchestra. We had Disney animatronics on stage. We had a wall of partnerships that Nvidia has and a new GPU. How about you? Any high level takeaways before we get into the nitty gritty of this

new GPU? Or the two new GPUs. I thought it was interesting that Jensen Huang disclaimed at the beginning, like, this is not a rock concert, but everyone knows it was a rock concert, because this is the company that has benefited most from the wave of generative AI. Jensen Huang was there to deliver a new product, and all of the developers and other attendees in the audience were there for it. So it did have the atmosphere of a rock concert to me. Yes. I think it's, if you have light up wristbands going in sync to something,

it's difficult to say that it is not a rock concert. There's a new graphics processing unit unveiled. I think this is the big story where we're going to spend some time. It's called the Blackwell. For people who are unfamiliar with this space, why is this chip a big deal? So Ricky,

I'm going to talk a little bit about this and forgive me if you and I go into some detail. There will be a point to this for those of you who aren't familiar with some of these terms. We'll get to the reason we're getting into this detail. But I think the first thing about Blackwell, it's a bigger chip physically. It looks bigger than its predecessor architecture. It purports to have two and a half times the performance in training of large language models, five times the performance in inference. Inference is when you ask a large language model, a question. For instance, you ask chat TPT a question.

That's a generalization, but we'll run with that. What's impressive about it, it uses a new custom 4-nometer manufacturing process, and that's developed with Fab partner Taiwan Semiconductor. The process links to GPU dyes. These are connected by a chip-to-chip link. So the basic architectures is two slabs of silicon with a connect. It also features a new generation of Nvidia's transform. engine, so this is an acceleration library.

It's super tuned for this new GPU. It incorporates also Nvidia's latest network switch to speed communication. Now, Nvidia is taking two of these Blackwell 200 or B200 GPUs along with a Grace Central processing unit or CPU, connecting these together to form the GP200 Grace Blackwell superchews. Super Chip. And if that wasn't enough. No, I need more, Ossid. If that wasn't enough, Nvidia is combining 36 of these super chips with 72 Blackwell GPUs and 36 Grace CPUs to form the Super Pod DGX, GB 2000 platform.

So this is a huge monster of a platform, which apparently weighs 3,000 pounds. It's made of 600,000 parts. if that weren't enough, we move on to the next NVIDIA supercomputer, which is going to be powered by the super pod I just described. So this was sort of the theme of yesterday's keynote presentation. To me, there was one takeaway, bigger is better, and we can dial into what exactly that means for the rest of the industry. Yeah, so I think one takeaway for me is someone who is less advanced in the tech knowledge

of what's going on is the previous chip, which was the hopper, allowed for billions of parameters, which was better for those, the natural language processing for a lot of the chat bots, chat GPT. This one, the Blackwell, is going to have more maybe industrial applications for video and images because you have trillions of parameters that it can work with instead of billions. Yeah, and I think the takeaway here is that Nvidia has been saying for a long time, that Morris Law is reaching its limits. And in fact, when Jensen Huang first started talking about this, that we can only pack so many transistors onto a chip,

and we're reaching the limits of this theoretical doubling every few years of capacity, he was also sort of saying, you're going to have to pay up for this. This was before Chat Chapti burst onto the scene. And lo and behold, now we see that, yeah, like Morris Law only goes so far. We have this new application, which is the transformer technology, in neural networks, which is what powers different large language models. We see a company that's benefiting from all this, but it's really telling us, look, we've hit the theoretical limitations of what we can do.

We can't have these amazing breakthroughs each time with the same size of chip or even the same size of supercomputer or server rack. Things are getting bigger and bigger and bigger. Of course, that's going to come into higher and higher cost, more expensive components and more of them. So I do think there's one potential like caution here. So many hyperscalers and sovereign governments have ordered the previous version of NBIA's latest and greatest at the H-100 GPU series. I wonder what the order book will look at look like as we move forward over the next year or two years and what role the Blackwell series is going to play

in that. Obviously they've got orders on the books. They will sell this, yes, like pancakes, But that whole mystery of Nvidia's supply and demand is going to become more acute with each passing quarter. Yeah, the chips are running into the laws of physics. So the transistors, which essentially are the on-offs, which is making the binary decisions. With this one, it's at four nanometers, which it's difficult to think of that length. One way to do it is that your fingernails grow at a length of about one nanometer per second. So four seconds of fingernail growth for the average human is the tiny, tiny size we're talking about here. And the order book's interesting.

You know, Jensen Wong made a large point that the Blackwell unit can fit in the same slots is the previous hopper unit. So you don't have to change your whole system. We're happy to sell them. I want to go to a different question, though, which is how far ahead is Nvidia's stuff coming from than the other designers? Maybe like Intel, which is also a foundry or AMD. I think it's got a pretty significant lead on both AMD and Intel. AMD is a lot closer. They have their own architecture, which is supported not by a proprietary software acceleration

library like Nvidia's Kuda, but an open source version called ROCM or Rockham. And AMD is a company that has very similar prowess in terms of, linking dyes together. In fact, they pioneered a technology called chip lick technology. And I think that's going to help them sort of keep within spitting distance of Nvidia. They will be able to also take their chips, stitch them together, and show greater processing power and faster inference times as we go forward. And don't forget that AMD also manufactures a lot of chips that go into components like laptops, like mobile devices and the industrial applications that you mentioned, Ricky. So they've got a market there for AI-enhanced circuitry that are

on the chip level, not the GPU level. Intel, I think, is an interesting story. It's backed off a little bit from its challenge to Nvidia and AMD. It's going back to the drawing board. So they may be a bit further out. I think Nvidia is maintaining a lead, but I do see AMD, even after this keynote address yesterday, is taking some share from Nvidia. One thing that's pretty apparent is that all the companies, NVIDIA, keeps mentioning in these announcements, Oracle, Microsoft, Amazon, etc., the big hyperscalers, partners. They also are buying from AMD. They want to support a second source of supply, because over time, that helps them get better prices when there's a competitive product

in the market that can bring down Nvidia's margins and lessen their cost outlay. So I do think we'll see some more orders on AMD's books this year. Yeah, each of these chips is in the tens of thousands of dollars range. And as you mentioned, Google Amazon, ordering from different suppliers, they also, I think, they're working on their own server processor designs in-house. So maybe the next competitor for Nvidia could be one of its current customers. Yeah, from a design perspective, I think that makes sense. The one thing, again, that the companies you mentioned, AMD, Intel, Nvidia, have that

Google, Amazon, Microsoft have less of are these long established relationships within the chip manufacturing supply chain, the ability to reserve capacity at TSM, and to really get a chip from design to mass production. Not that they're really in the business of that. So I don't ever see them becoming these competitors selling GPUs at scale. But in terms of design, yeah, they're all pretty formidable designers themselves. There was a big show for the rock concert that was not a rock concert. The market reaction was fairly tepid. Why do you think Mr. Market is feeling sort of lukewarm about this show, these announcements?

Ricky, I think this goes back to why I asked listeners to bear with us as we went to some detail. When Jens and Huang introduced this new Blackwell chip, he pulled it out of his pocket, and it was much bigger physically than the Grace Hopper. chip we're used to seeing. So that itself was a cue to me that, again, it's not in the design of the chip that Nvidia is creating so much more inference, but it's really in the size. It's stitching previous components together, making bigger racks, bigger supercomputers. So I think the market intuitively understood there's, while there is actually, yes, a very nice power consumption advance here. There is an ROI advance. The chips are more efficient. There's no breakthrough

here that will stave off other competitors. And those who are ordering these chips are really going to have to grapple with the fact that they still have H-100 orders that they're yet to receive. So the backlog is big. The previous generation still is a wait list item, although that constraint has been easing lately. So I think the market's just trying to absorb this and digested and realize that yesterday's keynote didn't bring clarity. There was no true, amazing breakthrough. And I say this like from a jaded perspective, because if you look at this chip and its architecture, yeah, it's phenomenal. I mean, it's crazily good. But we've been a little bit spoiled by this company. And I think that's what's factoring into the stock

price of Nvidia and some other semiconductor stocks. There were some, I would say, interesting pitches that could be breakthroughs. And there weren't definite dates on these, but they did seem, uh, Invidia was excited about it. So maybe we'll do some quick pitches from the conference. I want to see if you're bullish or bearish on these, these takes. Number one is that invidia should be thought of an AI foundry. Invita is to artificial intelligence, what Taiwan's semiconductor is to chips. This is something Jensen Wong was really trying to, trying to communicate with the, with the audience. Yeah, I'm bullish on that. I mean, really quickly,

Nvidia is really providing the whole picture for the data center, you know, from the GPUs to the networking, to the software, and they're creating so many different expressions throughout the industry in design, in execution. So, yeah, I mean, I'm bullish on that. It's a little bit of a mind-stretch, but I'm a bull. Nvidia's weather project, which is developing a digital twin of planet Earth in order to run simulations that will make better weather forecasts. Barish. This is a quantum exercise, and Nvidia certainly is one of the most advanced quantum players, but this is a long-term project. Don't look for Nvidia to come out next year with super accurate weather forecasts.

That's going to be one of the last chips to fall as our artificial intelligence over. overlords takeover. Humanoid robots. We got a very soft, a video of a very soft high five from a live one. It was somewhat, the high five, the robot was standing there and the researcher was gently touching it with a high five. We also got a lot of computer simulations of humanoids. Jensen Wong, big bull on the humanoid robots.

Bullish. I myself am a humanoid. I have had a very fun existence so far. I think we would just need a chatbot model, not a full humanoid model to do a podcast, but I appreciate it. And then fully self-driving cars that are coming in Jensen's pitch, that everything that moves will be robotic. I mean, slightly bullish, just because Nvidia didn't invent this out of the thin air, they've been working on technology for autonomous driving for quite a while, supplying a lot of the underlying infrastructure to companies like Tesla, other major, automotive manufacturers. So, yeah, slightly bullish on that one. And then anything else, as we've talked about, this was a two-hour conference. We've been talking

for a little under 15 minutes. Anything else you want to chat about before we wrap up? I think the thing that stands out to me was just how powerful NVIDIA's partnerships now having the other six of the magnificent seven. Yeah, let's end on that. That's such a fun observation, Riki. If anyone has time, go to Nvidia's newsroom and try to scroll through the press releases that started yesterday on the various partnerships and new product developments. You see you'll be scrolling for quite a while. All right.

Awesome Charma. Thank you for your time on your inside. Thanks a lot, Ricky. This was a lot of fun. 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 and take the lead. You have to adapt to whatever comes your way. When you're that driven, you drive an equally determined vehicle.

the Range Rover Sport. The Range Rover Sport blends power, poise, and performance. Its design is distinctly British and free from unnecessary details, allowing its raw agility to shine through. It combines a dynamic sporting personality with elegance to deliver a truly instinctive drive. Inside, you'll find true modern luxury with the latest innovations in comfort. Use the cabin air purification system alongside active noise cancellation for all new levels of quality and quiet. Whether you prefer a choice of powerful engines or, or the plug-in hybrid with an estimated range of 53 miles, there's an option for you. With seven terrain modes to choose from, terrain response to fine-tuned your vehicle for the roads ahead. The Range Rover event is on now. Explore Enhance Offers at range rover.com. For our next segment, first, a quick ad.

You just heard from Asa Sharma. He's also an analyst on our flagship investing service. It's called Stock Advisor. When he isn't talking stocks with us, he's scouring the world for quality companies capable of beating the market for long-term investors. If you're interested in more analysis from Osset and access to Stock Advisor's full scorecard of stocks, visit fool.com slash osset. That is, fool.com slash ASIT. All right, next up, Alison Southwick and Robert Brokamp

answer your questions about lending shares, where to keep your investments, and what activist investors actually do. Our first question comes from Brooke. I'm getting emails from my broker. Schwab asking me to lend stocks to other investors. They're claiming it's low risk and I'll get paid to do it. Do you think it's a good idea? Well, Brooke, I will start with a rule I have about emails and texts from financial services companies as well as charities and just about anyone else. Don't click on any of the links in the email. Don't call any of the numbers in the email. Instead, contact the company directly

using a number or email that you know is valid. And I say this from experience because my wife once clicked on a link in an email from what she thought was our bank, entered her a little, login info, and then our account was drained by scammers in Italy. That was 20 years ago in the early days of email scams, and they've just gotten worse. And now include phone calls from caller IDs that look like from your bank or your broker or the IRS. So, don't respond directly to any of those. So, all that said, let's get into securities lending. And it happens all the time, usually to facilitate short selling. So someone who sells a stock short is doing so because they think the price is going to go down. But in order to do it, they have to borrow some shares from

someone. And the people who lend those shares get paid a little bit of interest in return. It's all done behind the scenes by the brokerage company. No one's going to contact you to ask to borrow your shares. You don't have to do anything. And if at any time you wish to sell those shares, you generally can. Schwab has a whole page on its website devoted to the program. And in their hypothetical example, you could earn an annualized interest rate of 8.5% on top of whatever happens to the price of your shares. So, that sounds pretty good. Though the actual rate will vary, in many cases, it'll be much lower, especially if you're lending out shares of a widely held company. And you could sign up for the program and your shares never get lent out. It just

depends on the demand for the shares you own. So there are some downsides to think about. One is that you lose your voting rights, and we're going to talk a little bit more about voting rights when we answer a question later in the episode. Another downside is that if the stock pays a dividend, you won't get it. You'll instead receive what they call a cash in lieu or a payment in lieu, which will have the same value as the dividend. However, if the account is a taxable brokerage account, not an IRA, the payment is taxed as ordinary income, and thus taxed at a higher rate than qualified dividend payments.

So this would increase your tax bill a bit. And finally, in a previous bailbag episode, we talked about how most types of investments in a brokerage account are covered by SIPC insurance in the case of fraud, or the broker goes under. However, that coverage does not extend to shares that you have lent out. To offset these risks, most brokerages pledge that they have an equivalent amount of collateral held usually at some third-party institution as extra protection, but just know

that this could still be a potential risk. So you'll just have to decide whether the extra income that you may receive is worth those downsides. All I can say for sure is, don't click on the link in the email. Go directly to Schwab and learn about the program and sign up for it there. Next question comes from Dana. My ex-husband and I got divorced five years ago. He's remarried, but I have not yet. I heard I can get Social Security from him. How does that work? I'd prefer to not reach out to him if possible. So, Dana, you can get Social Security benefits based on an ex-spouse's record as long as you were married for at least 10 years. So if that is the case for you, I have some good news in that you don't have to interact with your ex-spouse at all. It's all handled by the Social Security Administration. Also, what one ex-spouse does with their Social Security has no effect on the other ex-spouse's benefit.

So there's no situation in which your ex-spouse could reduce your benefit or anything like that. So that's the good news. However, the rules can get kind of complicated. So, for example, to be able to claim benefits on an ex-spouse's record, you have to be at least 62 years old, as does your ex-spouse, though he or she doesn't have to actually be receiving Social Security at that time. And in most circumstances, you can't have gotten remarried, though there are actually some exceptions to that. And there are additional rules if your ex-spouse is deceased, because then you may be eligible for survivors' benefits. So as you can see,

this can get pretty complex. So visit the website of the Social Security Administration at ssa.gov to learn all the ins and outs. Next question comes from Dylan. I'm a newer investor, and I was hoping you could discuss how you differentiate your brokerage accounts? When you talk about stocks and ETFs in a portfolio, are these all-in-one-one-account spread throughout a taxable brokerage, IRA, and 401K? Any general guidelines or thoughts to share on setting this up? Thank you again for all the value and insights you provide. You're welcome, Dylan. So a taxable brokerage account, an IRA, and a 401k are all separate types of accounts.

And most investors have multiple accounts at multiple providers. So they may have their 401k, at Schwab, an IRA of Vanguard, a brokerage account with Fidelity, and so on. If you're new to investing, I think the way to think about this is start by thinking about the goal of the investment when choosing the account type. One goal that most of us have is retirement. If your company offers a 401k with a match, that's the first place to start, and you're stuck with whatever the employer offers there. And if you have an outstanding 401k, there's no cost, you like all the investments, you could just stick with that. Most people should be aiming to save 15% of their income for retirement, and that would include

the match. Now, if your 401k is not so great, take advantage of the match, but then go to an IRA, individual retirement account. Just know that any money that you're saving for retirement, you really should leave for retirement. In many cases, if you take the money out of retirement account before age 59.5, you'll pay a 10% penalty. Therefore, if you are investing for a goal in which you'll need the money before your 59.5, that's when you would turn to just a regular taxable brokerage account.

I'll just throw in that if you have kids and you want to save for college, you would choose a 529 plan. Okay, so let's say you've decided that you should open an IRA and or a taxable brokerage account. How do you choose the provider? Well, fortunately, the Motley Fool owns a site called The Ascent, which rates and reviews account providers. So just visit the Ascent, see which firms.

offer the features you're looking for. Our next question comes from Jum. Recently, I've been hearing a lot about activist investors. I just realized, I have no idea what that means. What does an activist investor do? Well, when you own shares of a stock, you are a legitimate part owner of the company. And along with that ownership, usually comes the right to vote on various things, like corporate actions, such as mergers, or you vote for members of the board of directors.

And the number of votes you have is usually tied to how many shares you own. And for most of us, that number of shares is pretty small relative to the total number outstanding. So our vote really doesn't mean that much. However, there are some investors, be they individuals or institutions, that own so many shares, that they can kind of throw their weight around. In fact, they may have accumulated a significant number of shares just for that very purpose. And these folks are often called activist investors. In some instances, these are people who run hedge funds and who think that a company needs to

to make a change or several changes in order to grow the share price. The biggest example in the news these days is Nelson Peltz of Tri-Anne Partners trying to get two seats on the board of Disney. They've created a 133-page document that you can read at Restore the magic.com, a site they created. The document explains all the ways they think Disney is underperforming and the changes that the company needs to make. They're being very public about these ideas because they need other shareholders' votes. Now, Disney is not too happy about this.

So they published their own reports and videos, some of which you can see at VotDisney.com. And the latest rebuttal had a cover with a picture of Pinocchio, which tells you what they think of tri-end partners and their arguments. Other activist investors are large institutions who are just by virtue of their size, they can try to influence companies. An example of this would be the California Public Employees Retirement System, otherwise known as CalPERS. It's one of the largest pension funds in the world.

they use their votes to try to get companies to do more about things like climate change and board diversity. The bottom line is there's a good bit of debate about the effectiveness and value of activists investing. But many of these activists are undoubtedly successful investors. So it's always interesting to hear their arguments. Plus, there's often a good bit of drama involved, which can kind of spice things up a bit. Our last question today comes from Josh. I have only been listening to your podcast for a little over a year, but I've enjoyed all the helpful information you provide each and every day. I never miss an episode. Oh, that's awesome. I'm hoping you can provide me with some guidance like you have for so

many others. When I was two years old, my grandmother purchased some shares of Disney stock for me. Since then, I have received dividend payments when Disney actually pays them in the mail, in the form of a check from a company called Computer Share. My mother was the guardian on the account for all of those years, and now I'm 24 and just want to transfer these over to myself and into my e-trade account. After a lengthy process and lots of frustration, I've discovered I am unable to transfer the Disney shares. Based on what I have seen, it is because they are certificated shares, meaning they are actual paper certificates that represent these holdings and they are not truly held at computer share.

Any advice on what steps to take? So every publicly traded company hires a firm to keep track of shareholder records, send out dividend checks, cancel research shares, things like that. And this company is known as the transfer agent. And in Disney's case, the transfer agent is a company called Computer Share, which is actually the largest transfer agent. And I would bet that many, if not most of the people listening to this podcast, have some relationship with Computer Share,

even though they may have never heard of the company. Unfortunately for Josh, I don't have an easy answer for him, especially if he's already done a lot of the research. I assume that what he's discovered is correct. What I will say is that every brokerage has a department, dedicated to transferring accounts and investments from other brokers and transfer agents. I'm sure there's a group at E-Trade that has dealt with this situation before. Back when I was a financial advisor in the 90s, people would come into our office in Clearwater,

Florida with certificates that they found in like their grandmother's safe. We'd call up to this department in New York City, and they would go through the process of seeing if the shares are worth anything, and then walking us through the process of, Basically, you have to sign a bunch of documents, you have to send in the certificates, they get canceled and reissued. It's a long process. But I'm telling you, there's some group at E-Trade that can do this. So if Josh hasn't already done so, he should contact E-Trade and see if there's someone who can help them. If they can't reach out to any other brokers, you may have to see if they have an answer for you. But in the end, I suppose it's possible that Josh has already done all this, and he's just not able to transfer the shares.

In that case, you'll just have to hold them as is, which I don't think is a horrible option. I have two stocks that I bought directly from the companies through a direct stock purchase plan more than 20 years ago. And Computer Share is the transfer agent for those. And so I get all my statements and my dividends from Computer Share. And it works out fine. Just know that when you actually want to sell the shares, it's going to be a multi-step process that'll take a while. All right. If you've got a question for the show, just like the ones that Allison Southwick and Robert Brokamp just answered, email us at Podcasts at Fool.com.

that is podcasts with an S at fool.com. 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 anything based solely on what you hear. I'm Ricky Mulvey. Thanks for listening. We'll be back tomorrow.