Nvidia (NVDA) stock has been running and seems unstoppable as it in on the threshold to a new all-time high. The AI chipmaker announced that its Blackwell graphics processing unit (GPU) is currently sold out for the next 12 months, while Advanced Micro Devices' (AMD) newly unveiled Instinct MI325X AI chip was met with underwhelming enthusiasm by investors. Bernstein managing director and senior analyst Stacy Rasgon sits down with Julie Hyman and Josh Lipton for a conversation around the deepening race between AI chip makers and hyperscalers. "The gap in terms of timing between what Nvidia is delivering, what AMD is delivering is still very, very wide. I think it's still Nvidia's game to lose," Rasgon says. "People are foreseeing demand continuing, and they're actually making investments to enable that. And with with all these stocks, like there's a big worry that just the numbers have gotten so big so quickly. They're just worried about sustainability. "But, clearly everything we're seeing from the supply chain and the customers, it doesn't suggest that there's any sort of faltering of demand anytime soon."
All right, here's another thing we're watching, NVIDIA jumping today set to finish at an all-time
high as it closes in on reclaiming the number one spot atop the market cap mountain. This coming as
the chip makers seem to have cornered the market with Blackwell demand soaring and rival AMD's
latest chip disappointing investors. So how should you be positioned among these big chip names? For
more, I want to welcome in Dr. Stacey Razgon, Bernstein Managing Director and Senior Analyst.
Stacey, always great to see you. So let's talk NVIDIA here. This stock, Stacey, is up about 14%
already this month. I'm sure it helps CEO Jensen Wang when he's out there, Stacey, saying,
you know what, demand for my Blackwell chip, it's insane. That's his word.
He's not holding back, is he, right?
He is not, but what do you make of this move, Stacey?
I'm sorry, what was that?
I just said he never does.
No, no, he doesn't. Sorry, what was your question?
Walk us through the move, Stacey. Is it justified in your opinion?
Yeah, look, so if you go back a couple of months ago when they reported earnings,
the stock sold off a bit after that print. It actually wasn't really because of the demand,
it was because of the gross margins. They had suggested margins come down a little bit into
the end of the year, primarily on mixed, by the way, and it's because a lot of these new
products are starting to ramp and they're very complex and they're not cost optimized yet.
As people have done more of their checks as you go through the quarter and we met with them and
everything else, it's very clear for a few things. One is that the margins, they're not coming down
a lot. They're still well into the 70s, they're fine. The demand, I guess with every day that
goes along, it just looks like it's getting better and better and better. You're right,
Jensen was out, I think on CNBC or wherever, talking about it was insane and they're looking
for material amounts of new products to ramp as we go from Q3 to Q4 as that product hits the market.
But even the older generation stuff, the older hopper chips, our demand is still off the charts,
that's still growing. Then you couple that, AMD had their event last week.
Again, I always say this, I don't want to knock AMD, they're doing very well for where they were
coming from. This business for them might do $5 billion this year, it was zero a year ago,
that's fine. But if you look at the roadmap that's going forward and you look at some of
the specs they put in for their new products, it's called the MI325 is the new one. It barely
stacks up against the hopper chips that NVIDIA has been shipping for a year and a half, two years.
The new NVIDIA chips, the Blackwell chips, the new stuff, it's night and day in terms of the
capabilities of that. The gap in terms of timing between what NVIDIA is delivering,
what AMD is delivering is still very, very wide. I think it's still NVIDIA's game to lose.
Then I think finally, it even gets with some of the comments you just made about
Alphabet and nuclear reactors. People are foreseeing demand continuing and they're
actually making investments to enable that. With all these stocks, there's a big worry that just
the numbers have gotten so big so quickly, they're just worried about sustainability.
Clearly, everything we're seeing from the supply chain and the customers, it doesn't
suggest that there's any faltering of demand anytime soon, quite the opposite.
That's what's been taking it up. Well, I think that's a really interesting
and important point that you make, Stacey, because I think part of the pullback
in NVIDIA stock, perhaps, in addition to there being questions about supply and demand, etc.,
was this idea that all of the demand is being soaked up by the hyperscalers,
they're that middleman, if you will. It seemed like there were questions about where it goes
from there, what their customers are doing, how their customers are using it. It doesn't seem
like that question has necessarily gone away, but maybe it doesn't matter if hyperscalers are just
buying everything they can. It's not all hyperscalers. I think
NVIDIA themselves, I can't remember the number, 40, 45% maybe of their revenue right now was what
they called large cloud vendors and consumer internet companies, those guys. It's not all
going to them, but a lot of it is. Clearly, they're the ones with the deepest pockets,
and they can place the biggest orders, and it can be lumpy. For now, at least,
they can continue to spend. They're almost like an existential arms race. I think Google CEO put
this. He suggested that the risks of underinvestment were much bigger than the risks of
overinvestment. As of right now, they can continue to spend. Now, certainly, over time,
there needs to be a return. Clearly, if it turns out there's no return, all this stuff,
then the whole thing comes crumbling down for NVIDIA, and frankly, probably for everybody,
both in and out of studies. However, though, I don't think that that is really in question.
I think that there is a return. Return can come two different ways. It can be driving revenues,
new business models that drive revenues. It can also be driving productivity.
At least in the near to medium term, the large hyperscalers can spend even if the return is not
there. They generate hundreds of billions of dollars of operating profit every year. They
can afford to spend if they think it's existential. Over time, clearly, you do need their
customers, enterprise customers to adopt. We're already seeing data points. You've had companies
come out and talk about productivity savings. We've got other ones. It's still early days.
We're still in the experimentation phase. I think the general signs that people are actually
building with this and working with it, I think, are still there. I think it's early.
I think if we're asking ourselves this question three years from now, what is return? I think we
have a problem, but I don't think we're anywhere near that. For now, I think the demand is still
off the charts. They're going to keep selling everything they can make, at least for now.