Yahoo Finance

Nvidia shares recovering today after closing down 5% on Wednesday. The chip giant has lost $900 billion since its all-time high back in June. This comes as earnings for some of Nvidia's biggest customers, those large cap tech names have been a bit mixed. Our next guest though seeing some opportunity saying that the volatility represents a reason to buy. Joining us now, we want to bring in Antoine Chakaibane. He is a Technology Infrastructure Analyst at New Street Research. Antoine, it's great to have you here. So looks like you recently upgraded Nvidia to a buy, you've got a $120 price target on the stock. And this comes at a time when it seems like many investors are still waiting on the sidelines when it comes to Nvidia. What makes you bullish at this point?

Thanks, Shona, and Matt for having me. So, yes, I would say that first of all, since the peak of June, Nvidia's stock has pulled back by nearly 30%. Uh so that's underperforming most other semiconductor stocks that are exposed to data center AI. And reality is that, you know, even despite the concerns, the recent concerns on, on the company, you know, the Blackwell delay, concerns around macro, we still expect Nvidia to dominate the data center XPU market. And um we see in-house XPUs doing well as well, you know, like companies like, um Amazon, uh Google are developing their own in-house XPUs. We also see alternatives to Nvidia GPUs like AMD GPUs also uh ramping, but we expect Nvidia to remain dominant in a market that is growing very, very fast. Um and so I would say that uh we see little downside to expectations for Nvidia's data center revenues. We see little downside to valuation. Even the stock is now trading below the the trough multiple of the 2018, 2019 down cycle. Um and so that makes it a very good investment at the moment.

I want to talk about what we've heard from some of Nvidia's biggest customers, as Shana was mentioning. You've got Microsoft, Meta, Alphabet, and Amazon contributing about 40% of Nvidia's overall revenue, and that's according to Bloomberg estimates. If Wall Street was frustrated by the AI ROI story that we heard from those four big tech names that I mentioned, couldn't that hurt Nvidia too?

Yes, so I think the right way to look at it, so yes, you mentioned 40% of total revenues. Nvidia actually also discloses, you know, that the large cloud service providers uh represent, you know, about half of their, of their data center sales. Um if you look at how much capex for the top four hyperscalers is set to grow this year and next year, um and you unpack that, you know, between investments in AI, investments in traditional infrastructure, well it's true that we still need to see a bit further revisions to what consensus is expecting for capex growth next year to make room for Nvidia to be able to meet expectations, but not that much revisions, maybe 10, 15% actually. Um and uh yes, of course, whether capex can keep growing at such a pace beyond 2025 is a question mark. It will depend on the pace of adoption of generative AI powered use cases. It will depend on the monetization of all the investments made over 2023, 2025. But reality is that even if we conservatively model a significant slowdown in investments in 2026, 2027 by these, these giants, that still makes room for good upside for all the names we cover exposed to data center AI spending. And so that's Nvidia of course, but also AMD, Broadcom, Arista, TSMC, Micron, and the list goes on.

Antoine, I'm curious well, something you just mentioned a moment ago, and that was the fact that you're not too worried about the reports of Blackwell delays. I guess I'm curious as to why, why is it that you see that as a limited setback at this point?

So yeah, just put a bit of context, you know. Blackwell is a chip that's designed with CoWoS-L. You know, it's a packaging technology by TSMC um that is being used in, in Blackwell. That's different from what was used in Hopper, which was CoWoS-S. And so that involves like high density local silicon interconnects and redistribution layers to improve the die-to-die interconnect density so that the two dies can speak to each other very, very quickly, very, very fast. Um and so the the integration of these components into the substrate has resulted in, in issues like thermal mismatch and warpage. But reality is that Blackwell is a cutting edge chip. It's pushing the boundaries of packaging. Um and it was brought to market very, very rapidly by Nvidia. You know, Nvidia is accelerating its product cadence. So inherently, there were risks of delay. Um but that being said, we, we still expect, you know, Nvidia and TSMC to bring Blackwell to market with acceptable yield and volume in 2025. And in the near term, Nvidia can mitigate the situation in many ways. They can extend the life of Hopper. Um they can release a single die Blackwell chip. Uh and so yes, maybe the shipments will be more weighted towards Hopper in the second half of 2024 and, and early 2025. Um but we still sense enormous and urgent demand across the board. And that mitigates the risk of opposing shipments as customers wait for the next generation of, of chips to be available in volumes.

In our final minute here, how big of a risk is the potential economic slowdown that we're hearing teased out? I mean JP Morgan this morning, not only raising their likelihood of a recession this year, but also raising their recession likelihood to a 45% chance in 2025. To what extent does that macro concern impact your view on Nvidia specifically?

So I think that for 2025, you know, you can, like things are fairly well set. You can look at things from multiple angles. We talked about the hyperscalers capex earlier. We know roughly how much they expect to grow capex. Plans are already set, you know, like they already have a plan to train, you know, the next generation of GPT models and the next generation of Gemini models. I think here there's probably limited room for much, much like changes in, in plans. Um you can also look at the at the supply front, you know, and we know that CoWoS capacity, that HBM with memory capacity is going to make room for Nvidia to meet expectations. So I think 2025 is fairly well set. Now yes, if we enter a recession, if the macro deteriorates, what happens beyond 2025 is still a big question mark, and we could end up in an AI winter, we could end up in a, in opposing investments. Um and that, of course, remains a risk.

An AI winter that could come with a recession, but sounds like you're definitely not pricing that in for now, as you still think Nvidia is a buy. Antoine, thank you so much for joining us and breaking that down for our audience. That was Antoine Chakaibane, a Technology Infrastructure Analyst with New Street Research, joining us on Nvidia.