NCI: Nvidia Breakdown & AI Trends

NCI_Nvidia Breakdown & AI Trends

Speaker1: [00:00:00] All right. Good afternoon. Welcome to the next episode of Lumida Non-Consensus Investing. I am pleased to welcome again Michael p PreK, or a Pete Guest who probably needs no introduction, but I'll give one anyway. He's really a polymath as I've gotten to know Michael. And back in the day, he was the Goldman Sachs Internet Research Group, founder and partner peer of folks like Mary Meer and Henry Blot.

And he runs one of the most thoughtful substack on ai. So highly recommend it every quarter, usually after NVIDIA earnings, we get together, we talk about all things ai. We'll be talking about NVIDIA earnings or jenssen's guiding towards a trillion dollars in CapEx by 2030. We'll talk about sovereign ai.

And what was Trump's objectives in the Middle East visit around ai. And what does this mean for the sector? We'll talk about Google [00:01:00] Agentic ai, Google Smart Glasses, Waymo, what Elon's doing with Tesla and Xai and the combination. So there's really quite a bit to get to keep questions coming as we discuss as well, we'll do our best to get through it.

I do also want to take a philosophical detour towards the end. Michael and I both have interest in these kind of topics and how different generations are using AI and Gen Z. So there's so much to get into. With that said welcome back Michael. It's good to see you again. 

Speaker2: Thank you. Good to be back.

Hello everybody. 

Speaker1: Let's start with Nvidia earnings. So Nvidia beat on earnings on revenue despite the four and a half billion dollars write off related to the export controls on China. As I mentioned, they're projecting a trillion dollars in CapEx in 2030. Their data center revenue is 39 billion.

That's 73% year over year growth. And the Blackwell GPU ramp [00:02:00] is the fastest new product ramp in the company history. So Blackwell, of course, is their leading framework providing AI inference and training solutions. And yeah, China didn't really make a dent at all in NVIDIA's earnings. So what are your thoughts and reactions?

Surprises? 

Speaker2: It was fascinating to watch Jensen really execute yet again. We've been watching and discussing their quarters now for a few ti, a few now, a few quarters now. And this one, the general theme is that, look, there's an insatiable demand as far as the, I can see the next few years for.

Infrastructure for the world is now convinced that this thing is real. We need huge amounts of AI data centers, AI factories, as Jensen calls them, and half his customers are the top four or five multi-trillion dollar companies who are buying every AI GPU. They can get handoff they're on. And [00:03:00] now the rest of the world is also in on this buying spree.

There's a whole trend now that Jensen dubbed appropriately the sovereign AI category, which is already in the tens of billions of dollars over the next makes few years. He just came back after accompanying the president in the Middle East, along with Sam Altman from OpenAI and others.

Came back with, over a million Blackwell GPU sold at 30,000 plus a pop to Saudi Arabia and a million Blackwell units sold. Correct those things. A million. Now that is a remarkable number when you consider that two years ago most AI data centers, and if you had 5,000, 10,000 chips back then it was the Hopper chips, which was the previous generation.

That was a big number. And these things, remember, 30,000, 40,000 are pop just for the chips, not for all the other equipment that you need for that. The point I'm making here is that the [00:04:00] Middle East was allotted under the Biden diffusion rules, maybe a few tens of thousands of ships. President basically said, okay for all number of reasons, we're gonna up that and it's a million trips.

The reason this is important for Nvidia is to the quarter is that their quarter was spectacular on almost every dimensions that meet and beat peoples expectations. They their margins were higher than people thought in, 72.5%. Their growth was great, and all these categories that you mentioned, especially the data center category, and we link their, the communication components that were doing great.

But as you mentioned, China is the biggest headwind that they've got for all the geopolitical reasons that are going on and addition all the tariff issues that they and the whole industry is going through. Apple in particular is on the other side of this, of that trade. As and it's like watching a formula F1 race and all the cars are lapping around and [00:05:00] NVIDIA's up in the front, literally, there's a huge pile up, explosion, everything, Nvidia just goes, the car goes through and it really comes out almost unscathed on the other side, as you said, they took a five and a half billion dollar charge.

It's gonna cost about 8 billion. Their market share in China, which is the second largest market for ai in the world, measured at least 50 billion a year. They had a 95% market share. They're going in, going in about a year and a half ago. Right now they're at 50%, and that'll probably go down to maybe 5%.

And yet the company is beating everyone's estimates. They've got more demand and they can shake a stick at, and the street goes up, upping their numbers. You can barely imagine what would've happened if there was no China issue. This company, this stock would've been, at new highs and beyond.

So great execution, not easy to do. It sounded, makes it looks easy. There's a great quarter from all of those perspectives. Having said that the [00:06:00] issue or the long-term issue over the next couple of years, RAM, is simply that, and I, and we've been talking about it for a while, is that the current policies of the US against China in particular are basically leaving, closing them off, giving them no options, and essentially compelling them to come up with their own software and hardware architecture.

Yeah. Now to compete with Nvidia and everybody else's, they won't do it in a year or two, but they will do it in five to 10 years if we let them. And that is the biggest threat to Nvidia. If people have a chance, read Jensen's official comments after. Earnings results on the, in the transcript. He lays it out as clear as he can.

We'll listen to, in a nondramatic way. You should talk about that. We'll, 

Speaker1: to his comments. We'll listen to his comments right here. I'll play a clip from Jensen yesterday after earnings. 

Speaker3: You just can't, you can't underestimate the importance of the China market. This is the second largest AI market.

This is the home of [00:07:00] the world's largest population of AI researchers. And we want all of the world's AI researchers and all of the world's developers to be building on American stacks. And it's irrespective of the near term revenue success that we have. We can't ignore the fact that the Chinese market is very important.

Speaker1: Yeah no, I agree with that statement too. The United States has dependencies on China. One would think that we would want to cultivate a China dependency in the United States, especially if AI is that important. 

Speaker2: You 

Speaker1: can, 

Speaker2: and he starts saying, half of the world's researchers are in China.

Which means we want them to be totally familiar with ai. Kuda frameworks from from Nvidia, which are the open source frameworks that, once you get to use it you are familiar with them and then basically spend your career around those tools. It's switching from Windows to Mac, et cetera.

You don't want them to switch. And that's the thing that Jenssen's talking about, 

Speaker1: right? And the more dependency China has in the us, the more the United States can shape [00:08:00] growth and, 

Speaker2: Continue to be the leading technology platform for the whole world. Whether it's business customers, government customers, and then remember, these platforms are what, hundreds and then thousands and millions of applications and services are being built.

Once you have that's the lock-in. And that lock-in can keep going for a very long time, and that's very important for an infrastructure provider. Last thing I'll just say is we've barely begun in terms of rolling out AI data centers in the world. Data centers in the industry has been around for 25 years.

I took some of the first ones public at Goldman Sachs way back in the nineties. Companies like Exodus, Equinix, which is still around, et cetera. Today's data centers are entirely different constructs. They consume 10, 15, 20 times the energy, 30 times the energy. They require billions in investments, not tens of millions of investments [00:09:00] their data centers are in all but name and the hardware and soft infrastructure that's going into them is essentially mountains compared to the ones that we had Jensen yesterday in the call talked about how just an average rack that they've got has 1.2 million components, not just from Nvidia, but from whole bunch of vendors.

Dell's storage memory, all these subsystems, all of that. But right now is all American. Not just Nvidia, but everybody else. Yeah. And we are facing a world where a lot of that could be non-American or Chinese in about five or 10 years. And that's the issue that he's really re ringing the alarm bell on.

Speaker1: Yeah. No, I, look I agree. I think future civilizations, if they discover an ancient lost Western civilization, they'll say, how do they build the pyramids? And how do they build the Blackwell GPU in terms of the complexity of all the parts? Let's have a listen to this clip from Google, CEO at the Google AI [00:10:00] Product Day.

Speaker4: Yeah. That was a great developer conference. We constrained this year in our cloud business. And when we are, all of us are simultaneously looking to scale up data centers. I. So we are running into real constraints. The way the constraints play out today is delays in projects because of permitting or not having access to electricians.

All of that is realities all of us are dealing with. If this trend line continues, the pace at which we are all ramping up, and obviously for it to continue, we all have to generate the returns on it and so it has to really impact the economy in a more substantive way so that they go hand in hand.

If the trend continues, these constraints will be much more visible. I think today we are all working through these constraints, so I think there are real constraints today, but I expect it to, for us to be competitive with China, et cetera, I think we have to solve the, these constraints in the near future.

Speaker1: The big debate around NVIDIA's earnings has been, Hey, look, we know [00:11:00] 2025 will be good, but what will 2026 look like? The evidence is consistent. Every quarter the hyperscalers come out and say, we are constrained by GPUs, we're constrained by energy and premise to build data centers.

‍