How hyperscalers like Oracle and Meta are driving the AI arms race

0:05 spk_0

Welcome to Stocks and Translation, Yahoo Finance’s video podcast that cuts through the market mayhem, the noisy numbers, and the hyperbole to give you the information you need to make the right trade for your portfolio. I’m Jared Blicky, your host, and with me is my co-host, Yahoo Finance senior reporter, Brooke De Palma, who’s here to connect the dots and to be that bridge between Wall Street and Main Street.Today we’re going straight into the AI arms race, and we’ve got a former chip designer turned Wall Street analyst to help sort things out. Our word of the day is hyperscaler, because in AI a handful of giant buyers can set the pace for everyone else. We clear the air on exactly what these giants do. And for today’s market show and tell, we’re looking at Nvidia through a fundamental lens. We’re breaking down gross margin to show how it can swing the stock price when the cycle shifts.And this episode is brought to you by the number 53%. Remember smart glasses, they’re back, and that’s how much demand is expected to grow this year. And today we are welcoming David O’Connor. He is a senior semiconductor research analyst at BMP Paribas and a former chip designer who spent a decade building silicon.With companies like NXP and Texas Instruments from the clean room to the street. David, it’s great to have you here.

1:18 spk_1

Yeah, thanks so much for inviting me guys.

1:20 spk_0

So let’s begin, uh, with your big picture overview of how you see the AI trade playing out right now. Oh.

1:26 spk_1

Yeah, straight into it. No, no, no, no softball there. Uh, I would say on the, um, on the semiconductor side of things, uh, kind of, uh, sentiment is quite tricky at the moment.A lot of wall of worries, I would say for the, uh, for a lot of investors, you know, worry on the kind of the disruption AI is causing. You know, we’ve seen that more in the software side or software stocks in the last couple of weeks, but also on the, uh, the hyperscale, you know, concern around the ROI, concern around monetization, also in the capic side of things as well. You know, we’ve got like 700 billion this year for just hyperscale or capex.Uh, up 70% year over year, but the question is, you know, how much higher that can go. Uh, and there’s no, uh, no good answers, I would say. So certainly, uh, right now we need better answers to these questions to kind of get us through this wall of worry and, uh, onto the promised land.

2:20 spk_2

Let’s stick with that theme because our word of the day is hyperscalar, and that’s a giant cloud operator that runs.Massive data centers and networks at a global scale. In the US, think Amazon, Microsoft, Google, Meta, and Oracle. And the reason this word matters right now is that in AI, a handful of these giant buyers can set the pace for everyone else. When they speed up spending, it pulls the whole supply chain with it, chips, power, networking, all of it. And when they tap.The brakes, it can hit hardware demand pretty fast. The misunderstanding is that hyperscaler sounds like a cloud buzzword, but it’s really a market power word. It’s about who has the checkbook, who gets priority, and who can turn AI into a real business. So David, from your seat on the sell side, what’s the most important thing that people need to know or get wrong about hyperscalers right now?

3:12 spk_1

Yeah, I, I think the, uh, most important thing for people to understand because they are representing 70% of the AI market. So they are, as you, as you, as you mentioned, writing the checks. Um, I think the uh thing to realize is that they’re well funded, uh, so they can afford, and that’s what is a bit different now versus if you look back at previous cycles back to the tech bubble in 2000.There, there was, uh, no cash being generated. So the guys who are actually buying all the infrastructure, uh, buying all the chips, as you talk about the connectivity, the power, they were not uh funded. They had no real business model. So, that’s very important to understand here that this uh spending can continue and as I mentioned,You know, this year, 2020 calendar 26, probably 90% of, uh, you know, that 700 billion represents 90% of their cash flow. So, quite elevated, but at the same time, you know, they can afford it, uh, they can afford it, uh, you could say. You

4:07 spk_2

mentioned earlier.Or that ROI for these hyper scales that’s return of investment. When you think about these names, where are we at now? Because that was really the fear back in the fall. And so has that fear eased at all heading into this year? Yeah,

4:20 spk_1

um, so it’s the same concern we had 3 years ago, 2 years ago, last year, uh, in September, as you mentioned. Uh, it hasn’t gone away. Still the same questions, uh, but now we need better answers and, uh, those answers aren’t really forthcoming.When we look at the hyperscale, you know, they, they are monetizing initially to a certain extent through better click through rates, better advertising. So there are levels of monetization there. But I think, uh, we need to see a lot more, like 700 billion in capics this year. Wow, that’s

4:53 spk_0

such a big number

4:55 spk_1

that, uh, and of course we see open AI.Uh, you know, let’s talk about their kind of revenue run rate and tropic as well, but there’s a big gap there between these hyper scale investment levels and what these guys, their annual are their run rate basically. So yeah, so it’s for sure, better answers, not new questions, but certainly we need better answers now given that the, um, you know, we’ve had 3 years of infrastructure build out so far, you know, is what’s going to drive that.Forward? Is it going to be just one more year of infrastructure build out, 6 months? You know, there’s uncertainty there. Um, and to kind of make that bridge from infrastructure to actually monetization, yeah, we need answers there. We need to see business models and, uh, yeah, that’s the big concern at the moment.

5:36 spk_0

Sounds like we just have to wait a bit of time for these answers, but the market’s always gonna front run something. Let me, let me, maybe this is unfair, because I know you’re a chips guy, but I want to ask you about the software side, just because it’s gottenUh, decimated over the last few months, and we saw this huge washout just a few weeks ago. Um, and I guess the big overarching theme is software as a service is going to be disrupted by AI. Elon Musk is out there saying that coding is dead, you know, you’re just gonna have one software app to control everything, who knows? But, uh, do you see a bottom in software coming and do you think that’s important to uh the AI rally?

6:14 spk_1

I spent 10 years in the industry, and 15 years on the uh the cell side. So, seen a lot of new technologies over the year, over the years. And I would say, you know, there’s parallels what we’re seeing at the moment where you have all the tools built out for a new kind of technology, which is AI.And, you know, as those tools are built out, now we’re in the stage where we’re getting more of these AI agents come to the fore and yet, to your point, uh, every week now, we seem to have a new AI agent that’s coming out and causing a lot of questions around existing business models that have been there for many, many years. Either tooling, workflows, um, all these AI agents are pretty, uh, disruptive. Now, is there value in them?You know, someone working on an agent over the weekend and coming out Monday morning with something that publish a

7:04 spk_0

paper, you know, er, erase a $1 trillion in market cap value. But you know, does the rubber meet the road there? Yeah,

7:10 spk_1

exactly. And to your, um, you know, to, to your question, where does this end? Uh, I would say, you know, so there’s, we’re at the stage where there’s a question on incumbents across the space, you know, obviously sour impacted at the the moment.I think, um, you know, as we go through this year and we see more and more agents, there’s going to be more questions asked. But I think all of this as well as as as I mentioned, you know, how much value is actually attached to these agents. It’s not very clear.And how disruptive they’re going to be long term and can they kind of really build a big ecosystem around that. There’s question marks there. So, yeah, like every, like we’ve seen previously, it’s up to the incumbents to actually go out there and prove and disprove these uh these new um disruptors and to actually show that AI is actually an accelerant to their businesses and uh, but that takes time.It’s easier to show your disruptor and uh to disprove that. So yes, I think that’s going, this is going to remain, this concern is going to remain with us through this year until those big incumbents can show that actually they’ve adopted AI and it’s driving, you know, new revenue growth for those businesses

8:16 spk_0

big overhang for the rest of the year.For market show and tell, and today we’re breaking down gross margin, using Nvidia as an example. Gross margin is just the share of revenue a company keeps after it pays direct costs to make and deliver its products. So when gross margin is rising, it usually means stronger pricing power, better product mix, or lower cost per unit.And we’ve got a chart of Nvidia stock, uh, price versus gross margin going back to 2019. And for our audio-only listeners, I’ll describe it. Both are up over the last seven years, but with a couple of detours. In 2022, gross margin took the down elevator because of gaming hangover that was inventory.Piling up in the channel, prices resetting and charges tied to clearing that inventory. Then in 2025, you had a different kind of pressure, a big new product ramp for Blackwell chips, where early units started out with lower margin. Plus China export restriction, restriction headaches that were forcing write-downs on certain chips. And a quick note on the charts, we get Nvidia stock price in real time, but we don’t get the final metrics like a financial metrics like gross margin until after each quarter ends.And in general, you don’t really use company fundamentals to time the market. Having said that, David, how do you use gross margin in your analysis for companies? How important is it to you in your models?

9:37 spk_1

Yeah, great question, Jared. Uh, you know, gross margin is a fundamental to any company really. Uh, you know, it’s actually shows how much value they’re delivering.Um, so taking their, their, their basic raw materials, adding value, whether that’s hardware, whether that’s software, and charging for that, and, um, you know, that’s where companies will pay. When you look at Nvidia, um, you talk about Nvidia there, you know, that mid-seventies gross margin there we’re at at the moment, uh, obviously, uh, you know, quite a, a very um strong gross margin relatively to the.Semiconductor industry. In general, I would say semiconductors you’re probably going to attain maybe 50% gross margins. So in AI, obviously it’s a new technology. It’s a hot technology. There’s a scarcity element as well. Um, you know, you can charge for all that. There’s a bigger software element as well. And, um, Nvidia is doing, uh, you know, tremendous job by capitalizing on that, I would say.Uh, now, there’s always puts and takes to that. Uh, but right now, you know, in the space, uh, some investors are concerned that kind of the sustainability of that level of gross margin in that kind of mid-seventy range. I would say from my perspective, you know, Nvidia is the fastest runner here in this AI chip game. And there’s value in being the fastest runner. Um, and that’s why I would see it as sustainable going forward, given the breakneck pace that kind of AI is progressing.Today

11:05 spk_2

And really quickly, when you think about where Nvidia sits, it sort of has like become the poster child of this AI arms resolution, as the Web Bush’s Dan Ives have put it. What would you, what, what would you tell investors who are maybe only thinking of Nvidia when it comes to this AI arms race?

11:22 spk_1

Well, like any technology, there’s probably room for just a couple of players. Uh, there’s room for the first guy who’s uh the fastest runner and he typically attains 70-80% of the market. You know, conceptually, number two guy, then maybe 10, 20% market share. And then the third guy, you know, just uh less market share.And anyone else behind that is out of business. So, you know, it’s quite, uh, any, and, you know, we’ve seen this over multiple generations of technology. So it’s very important to be the first guy and to be the fastest runner to kind of secure, uh, secure those big design slots that, um, at hyperscalers.

12:01 spk_0

Real quick, what makes Nvidia such a fast runner here?

12:06 spk_1

I would say they started sooner than everyone else. Uh, absolutely. They engaged, you know, in video when, um, you know, you talked there about the gross margin pressures that we had many years ago due to kind of crypto and maybe more gaming, but back then they actually engaged.AI developers, AI labs sooner than anyone else, and they actually kind of created that software ecosystem around kind of AI and they, you know, in one way as well, they had a product that was perfect to run AI workloads, right?Um, which happened to be gaming chips. So, you know, there’s a, there’s an element of uh being first, development, the foresight of the company to actually kind of say, hey, this could be a huge market and support that. So, so yeah, there’s, um, you know, multiple things went into it, but essentially.At the end of the day, it’s, uh, they got their 1st, fastest runner.

12:59 spk_2

So all eyes on who’s in 2nd and 3rd place, but we do need to take a short break. Coming up, we’re talking the latest AI tech, and we’ve got a runway showdown that takes a bite out of the AI spending race. Stay tuned.This episode is brought to you by the number 53%. That’s the projected year to year growth in AR glasses shipments this year, which is kind of wild because smart glasses were supposed to be the next big thing last decade, and instead they turned into a very expensive science project. But here’s why they might finally have a brighter future. Apple’s reportedlyPlanning up work on AI focused wearables, including those smart glasses, as part of a push towards AI hardware built around Siri and visual context. So David, how are you thinking about these next gen glasses evolving and what sort of features do they kind of need to have in order for users to think that they’re worth buying here? Yeah,

14:00 spk_1

it justSo moving over to the consumer side, uh, you know, uh, we were at CES in January, the big, obviously electronic show happens every January in Las Vegas, and I would say every second booth at the show was pretty much around AR glasses. Um, lots of Asian suppliers, um, you know, showing their wares, basically, uh, lots of US companies involved in this. So certainly you can feel that the technology is now on that cusp of adoption.Uh, and for, for sure, I think this year we’re going to see it as a, as probably the number one holiday product would be my guess. Um, you know, the, uh, AR glasses are not new to your point. You know, they’ve been around many years, but the form factor wasn’t there, the technology wasn’t.Quite ready. The use case as well. I think now, you know, the uh the form factor is there. They’re now light enough with the technology that we can actually wear. So the tech is there, the form factor, and, uh, from here it’s just all about the features. So still maybe lagging a bit on the features I would say.Um, it’s a bit more space invadery still, I think, um, you know, to draw an analogy from gaming, uh, but they’re progressing very fast, and I think, uh, you know, they’re, um, going forward, I would see AR glasses as kind of, yeah, on that kind of the knee of the curve here in terms of, uh, where they’re about to, about to take off. So very, uh,I must say I came away from CS uh pretty excited around uh AR glasses.

15:27 spk_0

What aboutsome of the other wearables like, uh, there, there have been pins that you just wear on your vest, your, your shirt, your lapel, and I’ve, I’ve seen these over the last few years, and the first ones just didn’t get any traction, but now you’ve got Apple supposedly getting into the game. Uh, where do you think that space ends up?

15:45 spk_1

Yeah, I think, uh, you know, there’s multiple other accessories that are kind of come and go. Uh, but I think that the big guys, guys like Apple, guy like guys like Meta, you know, are really focused on kind of, uh, things like AR glasses, lots of news out there around that. So it’s more on those devices than kind of more than the accessory side of things, uh.Yeah, the, the accessories, it’s, uh, it’s tricky. I, I, I would say, uh, you know, depending on the different, different gadgets, um,

16:13 spk_0

anything excite you at CES besides what we talked about?

16:16 spk_1

I would say outside of Aglasses, robotics was the other kind of application that’s really on the cusp of adoption here.Uh, it’s more on the kind of B2B, more on the industrial side of things, but having these humanoids kind of uh robots, um, you know, the

16:31 spk_0

Olympics, yeah.

16:34 spk_1

Exactly, the, uh, helping us, uh, in different, different applications on the factory floor. So again, that’s another application that’s been around for many years, but I think really now as we go into more 2027.Uh, the technology is ready. The, uh, um, the application is there. Maybe there’s still a bit of work to do on the acceptance side. Also they accept these robots, these humanoids wandering around, learning, following us around and learning what we’re doing. But again, can deliver massive value. So, humanoids is the other, uh, area and technology they’re pretty excited about, uh, coming out of CS this year. And a lot

17:08 spk_2

of investors around Tesla.Big bet on what exactly that could turn into,

17:12 spk_0

yeah, which is a good quote. What what companies are really investable in this? I mean, uh, Boston Scientific comes to mind, but I don’t know many others.

17:21 spk_1

Yeah, there’s, um, you know, there’s some that are directly involved in that obviously Boston Dynamics, uh, you mentioned Tesla as well. Uh, all these guys are pretty, uh, excited around around those, and they have their own offering.Uh, from a semiconductor perspective, I would say, you know, again, play the, uh, the picks and shovels guys. Uh, so that would be the analog names, analog semiconductors. So in my coverage would be guys like Texas Instruments, uh, analog devices, microchip, uh, NXP, all these guys who have the, all the sensors, all the, uh, compute chips, uh, that you need to actually kind of.Um, make these robots work basically. So, pretty, uh, exciting from the, uh, analog names that I cover. They have the kind of key technologies there is that, uh, any, any humanoid needs. Um, I would say just to give you one data point, on average, kind of about as much semiconductors in a humanoid, about $500 as there is in a car today. So that’s the kind of level of semiconductors per humanoid going forward.So pretty exciting once you get the volumes

18:25 spk_2

and big opportunity. It sounds like big road map,

18:27 spk_1

big, uh, big opportunities, uh, going forward. Uh, again, once the volume is here and we’re on the cusp of adoption, it’s really just about volume and ramping it up. Um, so that’s pretty exciting, uh, for 2027. We got, we

18:38 spk_2

got hyper scales, we got humanoids. Things are changing pretty fast,

18:41 spk_0

guys. They are. And now we’ve got a market stake on a classic Hollywood gap show staple, who wore it better.And today’s runway is all about the AI race, but with a twist. Do you win it by spending the most or by shipping the best experience to the most people? On the left catwalk, the Capex sprinter briskly struts. This look is pure arms race energy. An oversized utility jacket with server rack quilting, a heavy belt that looks like a power cable, and boots so thick they feel like they’re built for a data center floor. It’s loud, it’s modular, it’s built it bigger, build.Faster. On the right catwalk, saunters the design purist. This is a quietly expensive fit, clean monochrome, a crisp black turtleneck under a razor tailored coat. The only flexes are a few subtle, subtle wearables, sleek AR glasses and an AI pennant pin, discreet and effortless. So David, when you look at this AI cycle, who ends up wearing it better in the long run, the Capex sprinter or the design purist?

19:42 spk_1

Uh, great question, Jared, uh.My money is on the uh those who wear it best, I would say. So, for those who can deliver that experience basically to the consumer, I think they’re the one who kind of went out in the end. And you know, it’s it’s interesting right now we can kind of even start tosee a kind of move towards the edge, basically. So, you know, to what we talked about earlier, the monetization. You know, the monetization does happen at the edge. And um, yeah, for there, for, for those models to go into edge devices, whether that be the AR glasses, your smartphone.Um, and if you can deliver that user experience, you know, that is where the, uh, that’s where the money is. And in the long term, as I say, right now, the money is on the infrastructure side, but you know, the kind of that pots of gold shift as you kind of move through the technology evolution. And I think long term, for instance, guys like Apple are in a perfect position there

20:35 spk_0

that aren’t spending billions or 700 billion on Capex. Yeah. And what’s

20:39 spk_2

interesting is Apple kind of has been the one to fall behind, but could this new technologies, new products sort of put it ahead?

20:46 spk_1

Yeah, I think, uh, you know, we’re entering the kind of uh AI agent basically era now and we’ve seen the disruption to the software models we talked about.You know, also kind of as those agents get deployed more on the inference side of things, you know, they end up on our smartphone, and that’s where the kind of value, um, is to be realized. So, yes, Apple has been kind of behind, I would say. Um, I mean that’s a fair comment. It’s done recently the uh the deal with Google and Gemini, um, to have it via Siri, to have it exactly for for for for for Siri potentially.And you know, once uh that can drive a big upgrade cycle as well, uh, for someone like Apple. But again, you know, Apple is very focused on the user experience. And once it gets that right, then it’s in a perfect position to leverage that massive install base it has. Um, so, yeah, I think on AGI we’re pretty excited about Apple long term.

21:38 spk_0

I wonder, we’ve talked so much about AI we’ve also folded in robotics. Uh, anything else in chip talk that we kind of left out here that’s exciting to you?

21:48 spk_1

Yeah, there’s uh, you know, AI itself is happening, is, is, is accelerating the whole, uh, chip design process. So, you know, chip design itself is embracing AI, uh, just like on the coding side of things, coding obviously has uh embraced significantly AI and, uh, you know, even by the end of this year, about 50% of coding, software coding is going to be done by AI basically. And

22:11 spk_0

it’sgoing towards machine language instead of all these structured languages, it seems like.

22:15 spk_1

Absolutely, absolutely. I mean, it’s still tech kind of based, but uh a lot of these models have got to a level now they can, um, you know, really accelerate product cycles. And that’s pretty exciting for the tech industry in general. You know, when I was designing chips, it probably takes, you know, could take you anywhere from18 months to 2 years to really get the product right and and get it kind of out the door. And I think AI is uh enables us to accelerate that going forward. So faster design cycle is, you know, better for companies, better for businesses, and better for consumers.

22:48 spk_2

Really interesting. I’m looking out for the holiday. I want to know the price point on those glasses because I’mscared. We got a few months.

22:53 spk_0

I feel like I just got over the holiday season 125. All right, we gotta wrap.Things up here, but just a quick review, we began with the word of the day, hyperscaler, of which Nvidia is not. You gotta think the cloud companies, so we’re talking Amazon, uh, Microsoft, Google, and a few others there. And, uh, just thinking about fundamentals, we did talk about gross margin where Nvidia is the leader, they’re the first to the market, the fastest runner as we’ve been talking about, and, uh, in the mid-seventies there, they’ve got quite the lead, and we’ll have to see if they can hold on to that.And then we talked about smart glasses, uh, finally coming back here. I’m excited about the next generation of technology. I don’t know if I’m gonna be wearing it on my face, wearing it on my shirt, but I’m gonna be wearing it somewhere at some point. I will be testing this out. So, uh, we’ll have to figure that out as it comes.Translation, but be sure you check out all our other episodes of the video podcast on the Yahoo Finance site and mobile app. We’re also on all your favorite podcast platforms, so be sure to like, leave a comment, and subscribe wherever you get your podcast. And we will see you next time on Stocks in Translation.