저자: DanIves—자막내명시소속미확인,제목·설명란기준FinancialFiles채널출연자로추정 · 외국인
요약 보기
엔비디아 실적 호조(매출 78% 성장, Blackwell 110억 달러)에도 관세·중국 수요 불확실성은 존재하나 AI 수요가 더 강해졌다고 평가. Blackwell 구성 변화로 마진은 압박받지만 매출 성장이 이를 상쇄해 이익 성장은 지속될 것으로 봄. 2025년 데이터센터 매출 2000억 달러 전망, 2026년 주당순이익 6달러 전망, 현 주가는 저평가라는 주장. 쟁점은 2026년 이후 AI 지출 지속 여부와 AMD·구글 등 경쟁 심화.
다듬은 글
There's definitely nervousness in terms of tariffs and what that's going to mean, but I actually believe it's really about demand. I mean, demand is more robust when it comes to chips, AI revolution, Nvidia than it was four months ago last. Does Bush model out a 5% China GDP as President Xi talked about?
Do does that diffuse across the Pacific Rim? Look, I think ultimately a stronger China, more stimulus is more bullish for Apple. It's more bullish for Tesla in terms of everything we see in China as just how important those markets are. Tariffs on China create uncertainty for companies like Nvidia, Apple, and Tesla.
China's 5% GDP growth target, is it enough to boost global markets? AI demand has increased over the last four months. Nvidia and chipmakers benefit from AI expansion despite macro concerns. Apple and Tesla rely heavily on China for sales and supply chains. All right, let's talk a little bit about Nvidia because Nvidia beating fourth quarter expectations, revenue surging 78% from a year ago on strong AI demand.
Now the chipmaker also issuing strong guidance for the current quarter. Joining us now we have Antoine Chaban. He is a new street research technology infrastructure analyst. Antoine, it's great to have you back here on Yahoo Finance. So we are seeing some gains here ahead of the open. Of course, this comes following the initial drop that we saw on the results yesterday.
I'm curious how you're making sense kind of of that knee-jerk reaction maybe that we did see and your assessment of the quarter. Thanks Jenna. Thanks Brad for having me. Great to be back on the show. Um I would say you know our key takeaway here is that Blackwell is ramping uh very fast. You know like 11 billion sales in the quarter.
Uh that's well above our expectations. That's well above uh what the company you know guided for last last quarter of like several billions. Um and I think you know that's what the the street you know likes in the print. Um it means that hopper revenues are declining uh overall uh sequentially. But the most important takeaway here is that it dismisses concerns of a slow um of a backloaded deer.
Um uh you know there were a lot of concerns in recent months in recent quarters um around you know Blackwell being slow to ramp around like uh some challenges in ramping the the rack level system that's really pushing the boundary of innovation. We're talking about really complex systems here and I think investors were worried about that and now you see the company printing 11 billion sales of Blackwell in the quarter.
um that completely dismisses you know uh those those concerns because now we expect in the in the next quarter probably sales of of Blackwell have approaching 20 billion revenues um and and that's going to more than offset of course Hopper coming down. So I would say that's the key takeaway uh from the quarter here.
Key takeaways from Nvidia's earnings are that revenue increased up to 78% yearover-year driven by strong AI demand. Blackwell sales hit 11 billion and the results are faster than expected adoption. Next quarter forecast for Blackwell sales is around $20 billion. Hopper revenue is declining, but Blackwell's growth offsets it.
Concerns about slow Blackwell adoption are dismissed. And it seems like they're really leaning on these different configurations of Blackwell to drive some of the margins going forward. If if we can get real nerdy with it for a second here, what does that configuration set and and the broader kind of uh ecosystem for Blackwell look like here, especially as they're trying to make sure that they can extract more margins to impress investors long term who may not have been impressed on the initial release of the results.
So as you know uh up to now a GPU system you know consisted in like eight GPUs a GPU server what Nvidia has been calling you know like a DGX and HGX comprised up to eight GPU um and as we're moving to Blackwell we still have the option to to have like an 8GPU server but now you also have a 36GPU server you also have a 72GPU server um and you can you know use different types of networking to connect this 72 GPU or this 36 GPU use and you can water cool them, you can um air cool them.
And so that makes of course the ramp of the of the of that new generation much much harder. And so it's no surprise you know that gross margins are coming under pressure and we expect you know things to improve throughout the year as the product ramps. But actually you know over the longer run uh these chips are not getting uh any less complex.
uh as you as you think you know about Rubin and like its successor, we expect a lot more high bandwidth memory, you know, which is like this memory where the chip can really um uh exchange data with with the memory at very very fast bandwidth, very very low latency. We expect more of that in the subsequent generations and that's going to keep driving gross margins down in our view.
Um, that being said, you know, revenues are going to grow so much and Nvidia is going to be able to extract leverage on OPEX uh in a way that's still going to drive very very strong earnings growth. So, we're not concerned about any of that. Key changes in Blackwell configurations include custom networking and cooling options, both air and water, higher complexity, and GPU setups, which are increased from 8 to 72 GPU configurations.
Margins are under pressure due to more high bandwidth memory, more complex networking and cooling solutions, and larger GPU clusters. However, revenues will outpace margin decline. Operating leverage will drive earnings growth. Even with short-term margin pressure, Nvidia is still growing rapidly. And Don, talk to me a little bit more about just the earnings growth that you are expecting in the quarters to come.
And and I want to play a quick sound bite that we heard from CEO Jensen Wong. Nvidia CEO Jensen Wong last night talking about Blackwell Ultra and the impact that he sees that having here in the second half. Blackwell Ultra is second half. Uh as you know the first Blackwell was uh we had a hiccup. Uh that probably cost us a couple of months.
We're fully recovered. Of course, uh the team did an amazing job recovering and all of our supply chain partners and uh just so many people helped us uh recover at the speed of light. And so now we've successfully ramped production of Blackwell. But that doesn't stop the next train. The next train is, you know, is on on an annual rhythm and Blackwell Ultra with new networking, new uh new memories and of course new new processors and all of that is coming online.
An I bring up that that sound bite there in the context of what you were saying earlier just in terms of the fact that you thought Jen Zong was able to alleviate some of those concerns around surrounding the delays and maybe the impact that that is ultimately going or initially had on Nvidia. How do you see how strong is Nvidia's pricing power still here in the quarters to come?
So I think the pricing power is going to remain very strong for the foreseeable future. You know, you don't really have um much competition out there uh that can match Nvidia. You know, you have AMD that's like more than an order of magnitude smaller today and who's who's you know struggling uh to to ramp more than like its two very large customers, you know, Microsoft and Meta.
Um, and then you have AS6, you have in-house AS6 and and so far you only have one hyperscalers who's successfully ramped, you know, like the TPU and I'm talking about Google, of course, uh, they've been working on that project for 10 years, you know, and it's no easy task. You have to, it's not only like designing the chip, it's also designing all the software on top of that to make a good use of that chip.
Um it's also making a chip that's flexible enough so that it can adapt to all this change you know in in the workload that's like all these AI models uh introduce new new new new new technologies from one generation to the next uh now we're talking you know about three scaling loads uh you have like reinforcement learning that comes into play all this variation is is easily handled by GPUs um it's much harder you know when you have like a chip that's more dedicated and so Bottom line is that there isn't that much competition today and so we expect Nvidia's pricing power to remain very strong and if the gross margins come down it's simply reflecting you know um uh first in the near term of course the black wall ramp and over the longer run as I mentioned uh like increasing uh content of memory uh inside these chips but nothing to worry about.
Key takeaways from CEO Jensen Huang's comments on Blackwell Ultra are that Blackwell Ultra is launching in the second half of 2025 with new networking, new memory, new processors, and a supply chain recovery after initial Blackwell delays. Nvidia still dominates because AMD is too small to be a real threat, mostly serving Microsoft and Meta.
Custom AI chips from hyperscalers are tough to scale. Google took 10 years for TPU, whereas Nvidia GPUs remain the most flexible, easily adapting to new AI models and workloads. Nvidia is still a buy if you believe in AI growth and Nvidia's dominance. Antoine, we only got about 15 seconds. Even with the move here that we're seeing pre-market, Nvidia will still be down year-to-ate scale here.
For investors trying to figure out before the opening bell and going into the start of today's trading activity, if it's an investable company at this level on that pullback, what would you tell them? I would just tell them that we're moving, you know, from so 2025 I think is fairly well well framed.
you know we have uh sequential growth that's going to be like in the mid- teens we expect 200 billion data center revenues for the year that's consistent with the supply chain on one side that's consistent with hypers scale capex on the other side you know all the hyperscalers have raised capex significantly and now we have like a consistent view of 25 I think the debate is shifting to 2026 and in 2026 you're going to have a handful of players who are going to move from clusters of hundreds of thousands of chips to maybe half a million you know on a trajectory to 1 million by 2027 right that's enough to drive all the growth that is going to you know be enough for Nvidia to reach to approach you know $6 of EPS and on today's stock price you know I think is pretty cheap uh when you when you look at this in that perspective Antoine thanks so much for taking the time here with us reacting to Nvidia earnings $200 billion in data center revenue is expected for 2025 hyperscalers like Amazon Google and Microsoft are boosting AI spending and moving from clusters of 100,000 GPUs to 500,000 to 1 million is expected by 2027.
The EPS estimate is $6 in 2026. At today's price, Nvidia looks undervalued. If AI demand stays strong, Nvidia could have another massive growth cycle. But if AI spending slows or competitors like AMD, Google, and Meta Catchup, margins could shrink. A big tailwind for your company was your relationship with Nvidia last year.
How is that now? Uh we are we had a great relationship with Nvidia. We've worked with them uh innovated together um for many years and uh we had a great uh demo at CES using Nvidia's uh hardware. That demo was very well received and we are going to be participating at the GTC conference in just a few weeks.
And then when you position yourself against like larger voice uh AI players like say um it sounds like this company has had a strong partnership with Nvidia benefiting from collaboration on AI innovations with a well-received CES demo and participation in Nvidia's GTC or GPU technology conference. They seem to be staying closely aligned with Nvidia's AI ecosystem.
Comments 0