Nvidia earnings: What to expect from the AI darling

Data and Analytics 03
6 followers
1 Video View·Oct 29, 2024
Nvidia (NVDA) is set to report third quarter earnings for the fiscal 2025 on Wednesday, November 20. Goldman Sachs managing director Toshiya Hari joins Seana Smith and Madison Mills on Catalysts to discuss what to expect from the key artificial intelligence (AI) player. The analyst highlights two things that he’s keeping in mind heading into the earnings results: the demand environment around Nvidia’s products and Nvidia’s strength beyond AI training in inferencing that support the chipmaker's growth is strong and sustainable.
And last but obviously not the least to report, we've got NVIDIA. The AI chip giant with a market
cap greater than three trillion dollars will close out the Magnificent Seven reporting
towards the end of November. Joining us for more, we've got Toshia Hari. He is Goldman Sachs' head
of U.S. semiconductor research and he holds a buy rating on NVIDIA. Toshia, it's always great
to speak with you because you have such great insights into NVIDIA. I know that you recently
met with Jensen Huang. You also had him speaking at the Goldman Tech Conference earlier this year.
What is the biggest thing that you learned from those conversations that is going to impact your
thinking heading into NVIDIA's print? Of course. Thank you so much for having me, Madison.
You're right. We did have Jensen attend our conference in early September. We were also
on the road with him and the rest of the team very recently. Look, there are a couple of things. First,
in terms of the demand profile out there, the environment continues to be very robust.
If there is a problem, it's a supply problem, not a demand problem.
You've got the large hyperscalers, which we will hear from next week. From a couple of them,
they're still in the early stages of the buildout. From an AI infrastructure perspective,
the customer profile is broadening beyond the hyperscalers into enterprises and also sovereign
states. There is demand across the board there. Secondly, I think there has been this perception
historically that NVIDIA is very strong in training, yet their competitive position could
potentially dilute as we transition to inference. As you look at the most recent models that have
come out from OpenAI, the complexity from an inference workload perspective is going up,
and as that proliferates to other models, we think that will only grow NVIDIA's growth
opportunity set and potentially expand the competitive moat that they have.
So again, demand very strong, competitive moat very sustainable as well.
Yeah, so Tushita, it sounds like you're saying that any concern almost of the fact that NVIDIA
could be slightly losing their dominance or the fact that some of these smaller players are
gaining maybe more of a competitive edge or better able to compete with NVIDIA doesn't
sound like that's worrying you at all. It's something to definitely keep in mind and be
cognizant of. And there are a handful of startups that are making really good progress. There are
other merchant companies as well that compete effectively with NVIDIA. But when you think
about NVIDIA's ability to really innovate across multiple components, if you will, it's not just
the GPU. They've got their internally designed CPU, the gray CPU. They've got NVLink, which is
a switch that essentially connects multiple GPUs. They've got Ethernet connectivity. They have
InfiniBand connectivity. And obviously, they have a very broad suite of software that augments their
hardware capability. So when you think about their breadth and their ability to innovate at a 12,
18 month cadence, I think it's going to be very difficult, very challenging and expensive
for others to compete head to head with NVIDIA. I want to get your response, Tushita, from two
things that analysts we've spoken with have called out in a somewhat bearish case on NVIDIA,
though you can't really be bearish on NVIDIA right now. But they've talked about the potential
competitors, something like a Cerebris, for example. They've also talked about the potential
slowdown in demand if we do have a macro challenge moving forward. Which of those two do you think
is more likely, or do you think that neither one is actually going to be that big of a headwind
for NVIDIA? Yeah, it's a great question. Between those two, Madison, I would say the latter
probably comes up more and more in our conversations with investors. Again, I don't
think the company is complacent in any way. They are very paranoid when it comes to competition.
But again, the bigger debate is really how to think about sustainability of demand going forward.
The majority of investors, we believe, are fairly comfortable with the outlook going into
calendar 25. The big question mark is calendar 26. To the extent CapEx continues to go up
at the hyperscalers, potentially negatively impacting their gross margins,
their free cash flow, what does that mean for 26 and beyond? So between those two, Madison,
I would be a little bit more concerned about the latter. But again, at this point,
we remain pretty positive on the stock and on the trajectory going forward.
Talk to me about what you've heard from those hyperscalers about their CapEx when it comes to
NVIDIA. There's been this big question about whether or not we're going to see a return on
that investment from hyperscalers. What are you hearing about that?
As of now, Madison, I think, again, as a group, as a cohort, the hyperscalers are very much in
build mode. Our colleagues, Capital Spending Estimates for calendar 25, the bottoms-up number
is suggesting sort of mid-teens to high-teens growth. We do think based on what we're hearing
from customers and partners and the supply chain in Asia, there's probably more upside risk to
those numbers. So again, we'll find out as we progress through next week and
a couple of the big customers report. We do think there's potential upside there.
Again, the big debate is not so much calendar 25, it's really calendar 26. At this point,
it's very difficult to prove or disprove where things are headed beyond 25. To the extent we're
sitting here in 6-12 months and the monetization is not there, the ROI is underwhelming.
Increasingly, we would potentially become a little bit more worried about not just NVIDIA,
but the companies that really make up the ecosystem from an AI infrastructure perspective.
But at this point, again, we remain very constructive. So Tashia, what does that tell
us about where the stock's headed? I'm guessing it's headed higher, in your view, and you see
new all-time highs on the horizon. That's right. We do expect upside from here. When we compare and
contrast how we're thinking about the 25 earnings profile, the margin profile for the company
relative to street consensus, we're modeling about $4.65, give or take, in EPS for calendar 25.
As of this morning, I believe we're about 15 percent above street consensus. So again,
we have about a month left until the company reports. But as we've progressed through
earnings season, we hear from the hyperscalers, we hear from NVIDIA in a month's time,
we would expect street numbers to move up and for that to drive the stock higher as well.
All right. Tashia Hari, Goldman Sachs Managing Director. Thanks so much for taking the time to
join us.
