Instant Reaction: Nvidia Gives Lukewarm Forecast, Stoking Fears of AI Slowdown 

0:00What if data didn't sit still? What if intelligence moved with us? Not buried in reports, but activated in real time. Where lives are being shaped. Where decisions are being made. It all starts with a question. Where is the potential? Cotality turns data into clarity. Intelligence into insight. Insight into action. Because when intelligence moves, we all move forward. Cotality. Intelligence beyond bounds. This is Jacob Goldstein from What's Your Problem? Running a business is hard enough. Don't make it harder with a dozen apps that don't talk to each other. One for sales, another for inventory, a separate one for accounting.

0:42That's software overload. Odoo is the all-in-one platform that replaces them all. CRM, accounting, inventory, e-commerce, HR. Fully integrated, easy to use, and built to grow with your business. Thousands have already made the switch. Why not you? Try Odoo for free at O-D-O-O dot com. That's Odoo dot com.

1:29Visit VitalProteins.com to learn more and where to buy. These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease. Hi, I'm David Weston. Join me every week for the Wall Street Week podcast to hear stories of capitalism from around the world. From geopolitical tensions and central bank decisions to artificial intelligence, energy, and infrastructure, We sit down with the CEOs, economists, policymakers, and thought leaders whose decisions are shaping markets everywhere we find them. Subscribe to the Wall Street Week podcast on Apple, Spotify, or anywhere you listen.

2:09Bloomberg Audio Studios. Podcasts. Radio. News. This is a breaking news update from Bloomberg. instant reaction and analysis from our 3 000 journalists and analysts around the world okay here the numbers are crossing second quarter revenue did beat estimates second quarter adjusted earnings per share coming in at one dollar and five cents second quarter revenue coming in at 46.7 billion dollars the estimates for 46.23 billion dollars second quarter adjusted gross margin, 72.7%. Second quarter data center revenue, this is the big one,$41.1 billion. Estimates were for$41.29 billion. Second quarter gaming revenue,$4.3 billion.

2:56That did beat estimates. As far as the outlook goes, NVIDIA sees third quarter revenue coming in at$54 billion, plus or minus 2%. The estimate is for$53.46 billion. Another one worth repeating, meaning that data center revenue for the second quarter,$41.1 billion. The estimate was$41.29 billion. NVIDIA shares in the after hours, they're sinking. Mandeep, come on back in here. You're looking at this on the Bloomberg terminal on your phone. Why the downside reaction? I mean, just the guide, the 3Q expectations were$54 to$55 billion. And to my mind, this print is not something to be excited about.

3:40What about this? an additional$60 billion share buyback. That just crossing. A$60 billion share buyback when the stock is close to a record? What does that mean? Well, I mean, this company is printing almost$100 billion in free cash flow. So it doesn't surprise me that they are using some of that free cash flow for buybacks. But if the top line is decelerating, then the multiple will compress. When that headline crossed about the share buyback, we did see shares give back some of those losses. Some other headlines crossing the Bloomberg terminal with regard to NVIDIA. The company approves an additional$60 billion share buyback,$180 million release of previously reserved H20 inventory.

4:24NVIDIA says no H20 sales to China-based customers in the second quarter of the year. I do want to bring in Jay Goldberg of Seaport. He joins us from San Francisco. Certainly a disappointment when it came to that guide, as Mandeep said, but then there's that$60 billion share buyback. What's your instant reaction? I think this is what I was talking about, is just getting harder and harder for them to beat the way they have been. I think we've all gotten accustomed to them having these massive blowouts quarter after quarter, and it just couldn't continue. By any normal company, this would be an okay quarter, but it's NVIDIA.

5:01we all have heightened expectations. And it was just, I think, very difficult for them to live up to it. I feel like it's the outlook for sales of 54 billion plus or minus 2 percent that is fuzzy against estimates of around 53.46 billion. But in both cases, remember the wide range of analysts forecast due to the unknowns of China. And I feel like this is what we've talked about, Tim, how there's such a broad range and it's hard to predict now, especially with China. But the all-important forecast for revenue,$54 billion, looks good compared with consensus. I think it's OK. I think that the China factor is interesting because like a year ago, if somebody or some country or some even some company wasn't able to take allocation of their chips for whatever reason, there would be a line out the door of other customers waiting to take those chips.

5:46And it doesn't seem to be that the line is that long anymore. Right. There's still there's still demand there, but it's not this sort of triple oversubscribed demand that they enjoyed a year ago. So we're speaking right now with Jay Goldberg, he's senior analyst for semiconductors and electronics with Seaport Research Partners. Also with us is Mandeep Singh. He is global head of technology research at this point. He joins us here in the studio. Mandeep, you're glued to your Bloomberg terminal right now. I'm glued to the live blog, listening to what you and Jay are saying, looking at headlines here.

6:23Investors will want to know what's going on with Blackwell. The company said that Blackwell architecture revenue grew 17 % sequentially. What are you watching? Yeah. And look, I think that's where the gross margins are actually trending in the right direction. So they did say they expect to exit the year with mid-70 % gross margin. So that's a positive. The Blackwell ramp seems to be going well. It's just, you know, when it comes to the aggregate number, it just is not that big of a beat and raise. And that's why you're seeing a stock reaction. But there isn't anything related to China that I can see in the print, which is what I was hoping they gave some color around expectations for China for the remainder of the year.

7:09Also in the live vlog, we have Ed Ludlow. He's the co-host of Bloomberg Technology. He's supposed to be on vacation. He's probably just tuned in for this. He says, curiously, the company says it benefited from a$180 million release of previously reserved H20 inventory from approximately$650 million in unrestricted H20 sales to a customer outside of China. What does that mean, Mandeep? Well, I mean, they basically wrote off that inventory and now they are able to sell some of those H20 sales. Remember, H20 is a deprecated version of Blackwell. So they're able to find a customer outside of China.

7:49And really, I think this points to some sales related to that. But I mean, 180 million is not going to make a dent, you know, when we are talking a company that is expected to do 54 billion in revenue next quarter. Jake, come on in on this on the China question, the same thing that Isabel asked Mandeep with regard to what what China is looking like. And Mandeep mentioning that he's not seeing any outlook in the commentary related to China. Do you have more questions right now about China than answers? As per usual, yes, there's a lot of questions around this. I think I mean, I know from my conversations with people in the industry, there is a lot of demand for Blackwell in China.

8:34no one knows if they'll be able to get any or when they'll be able to get any. The company did say a quarter ago at Computex that they had a variant of Blackwell. It sounds like it's ready to go. They just need approval to send it. And it doesn't sound like they've gotten that yet because they explicitly say there's no China sales in their outlook. In regards to the H20 sales, that's interesting because the H20 is just a variant of the H100. It's a really simple operation to convert an H100 to an H20. It's, you know, 20 cents apart or something to make that change. The company had said that they couldn't ship any to anyone else.

9:10Only China wanted it. So it is a little bit encouraging to see them sell some of that, but not entirely surprising. I also want to point out that H20 margins are not quite as good as H100 margins. So there's a little bit of a margin headwind there. But to Mandeep's point,$180 million for sort of China-related, China tangential sales is just nothing really to write about. So our own Ian King, who covers semiconductors for Bloomberg News, is out with his write-up. The headline, NVIDIA gives lackluster forecasts stoking fears of AI slowdown. The company giving a tepid revenue forecast for the current period, it fuels concerns that a massive run-up in AI spending is slowing.

9:53Is that a concern of yours, Mandeep? I mean, I look at their last eight quarters. They increased their revenue sequentially by$4 to$5 billion. This is the first time when the data center sequential growth of$41.1 billion is up 5 % from the prior quarter, which is the slowest growth. So to my mind, you know, there's going to, anytime there is a deceleration, the fact that they were growing sequentially double digits, and now it's mid-single digit, even with the numbers that they put up, I mean, these are fantastic numbers, but there is a deceleration. And, you know, when they're talking about new architectures, Blackwell, and the margins, as I said, is heading in the right direction.

10:39But that sequential deceleration in data center growth, to me, that is why you are seeing such a stark reaction. So, Jay, I'll ask you the question now that the numbers are out. Were you proven right this quarter that you're the only analyst tracked by Bloomberg that has a sell rating on the terminal? Oh, I'll take I'll take the win. I'll take I'll take a lowercase w in this case. I mean, it's not trading at one hundred dollars yet. No, but like I said, my thesis has all along been that it's just getting harder for them to beat expectations. And I think this is exactly what happened here. To Mandeep's point, there's not much growth in data center.

11:17You read all the other headlines, it seems like AI is taking over the world. But in reality, I think we're going to need some time to digest the AI we already have. And I think it's natural that things slow down here for NVIDIA, who's been leading for so long. Is it a red flag to you in any way, Jay, that the company's authorizing this share buyback? Do growth companies do that, or is that more of a sign of mature companies? It's a little bit of a red sign. I want to see how they talk about it. I mean, it would seem to me that there's so much opportunity out there in AI. Why not spend that$60 billion in furthering their growth?

11:53I mean, as a shareholder, I would appreciate it. But as an outside observer, I have to wonder, couldn't they find other ways to deploy that? And the other big thing here is the CapEx increases from the hyperscalers. So we know Meta plans to increase their CapEx by over 30 % for 2026. But for other hyperscalers, we don't know if they're going to grow by 30 % or 20 % or 15%. And that will trickle down in terms of what it means for NVIDIA's data center growth. So Google clearly is the one that relies mostly on their own chips. And, you know, when you think about the incremental buyers for NVIDIA's chips among the hyperscalers, yes, there's Oracle and Corvive.

12:35But, you know, if you take out China, sovereigns aren't going to make up for that incremental buyer. And if another hyperscaler develops their own chips, then that growth rate will come down. Again, it's a question of whether they grow sequentially 5%, 10%, and that will have a bearing on the multiple. So, Nvide is – oh, go ahead, Jay. I'm just going to say, and sort of going further with what Mandeep said is, if you look at the other hyperscalers, especially Microsoft and Amazon, they have some pretty big plans next year for their internal silicon. These are going to be big ramp years for their internal silicon.

13:12And I have to wonder if that's sort of chipping away at the edges of demand for NVIDIA. That's Jay Goldberg. Jay, stick with us. Jay is Senior Analyst, Semiconductors and Electronics with Seaport Research Partners. We're also joined by Bloomberg Intelligence Global Head of Technology Research, Mandeep Singh, who's here in our studio. So you both make the point that it's just hard to grow from here on out. How much of diversification is hinged on that? We know that 40 % of the revenue is from Microsoft, Meta, Amazon, and I'm missing one. I can't, I'm blanking on that name. But NVIDIA now offers computers, networking gear, software services.

13:51Can we count on that? Mandeep, maybe you can go first. Yeah, I mean, look, everyone knows this is a multi-year investment cycle. And data center spend when it comes to AI would be around trillion dollars. It's just the growth rate, right? So when you are underwriting 40 times earnings on stock, there is expectations embedded in terms of the company compounding at 30%, 40%. So if that growth rate were to taper, then the multiple comes down. It's still a fantastic growth rate for NVIDIA. It's just the multiple compresses, and that's what I think is the risk here. Do you share the same view, Jay?

14:30Yeah, I think Mandip summed it up pretty well. Right. I think there's just – it's good company, good products, but this kind of exuberance seems to be getting ahead of reality of the market. And I mean I have to wonder, you have all the hyperscalers spending these immense amounts of money, half a trillion dollars among the six or seven of these companies. It's hard to see them getting a return on that anytime soon. You sort of look through it. I don't think they have a clear plan. They're building for something that's important and AI is coming. But in terms of sort of hard dollars and cents, the ROI on that massive investment, it's not clear how they generate that.

15:08And so I think we're starting to start asking more of those kinds of questions. What are we actually going to use all this AI for? Well, maybe Snowflake can answer the question. I want to bring Mandeep Singh back in here because he covers Snowflake, shares of Snowflake up now about 11 percent in the after hours. The company gave a sales outlook for the fiscal year that topped analyst estimates, overcoming anxiety that software vendors will be hurt as the economy slows and new AI companies take away business. The shares jumped in extended trading. The narrative has been that software companies have been beat up for a couple of reasons.

15:42One, AI firms are able to do this. And two, a lot of the spending has gone to hardware and to expanding AI capabilities. What is Snowflake earnings telling us? And actually, Snowflake earnings points to probably something that is incrementally negative for someone like NVIDIA because so far all the development in large Android models was I put all the open internet data into an LLM and LLM becomes smarter and smarter. But now what companies are doing with AI agents is they are training on their own enterprise data. That's where someone like Snowflake or MongoDB benefits because your training is getting a lot more specialized depending on the use case.

16:25And it's not just about the compute cluster anymore, but also about the data. And I think that points to why Snowflake is doing well. I want to add another name to our roundtable. I want to bring in Sarah Fryer. She's Bloomberg News big tech team leader. She joins us from San Francisco. This was the one we've all been waiting for. And finally, it has arrived, Sarah, the company giving a tepid revenue forecast for the current period. It fuels concerns that a massive run-up in AI spending is slowing. How are you thinking about the coverage of this report, especially in the context of the other big tech companies that we've heard from who have said, yeah, we are spending when it comes to AI infrastructure?

17:05Well, you know, I think we have to put it all in context because NVIDIA is in this space of having so much uncertainty around the China market, around the future of spend on data centers. the analysts who cover the company had like 15 billion dollars in variance on what they expected the revenue outlook to be whether or not they included China in their models so a little bit of a grain of salt on that revenue number forecast I think it's just it's just a company where people don't quite know what to make of their discussions with the Trump administration about what they are are not allowed to sell to China and whether to even include that in their models.

17:48And they did say that they didn't sell any of those H20 chips to China in the quarter. So I think there's that. So that doesn't necessarily indicate for us that the AI boom is over or anything like that. That said, you know, Mandeep is right. The training of models, like we've seen this with OpenAI, Like the latest and greatest model is not as much of a leap from the prior. And, you know, other companies are having similar issues. So the idea that AI, large language models are just going to accelerate in smartness until we reach super intelligence. I mean, that may not be, but there is real money going to building data centers.

18:35Meta's building one the size of Manhattan and Louisiana. Like that is not over. That is very much still a part of what companies are investing in and doing. I want to bring back in Jay Goldberg. He's senior analyst at Seaport Research Partners. Jay, aside from NVIDIA's struggles in China, the biggest impediment to growth seems to be also the availability of supply. Can you talk to us more about that, how NVIDIA doesn't own factories and relies on outsourced production? So, yeah, NVIDIA's manufacturing is done at TSMC. And for a long time, last couple of years, that's been the big constraint on NVIDIA's capacity is what they can get out of TSMC, in particular, the co-host packaging, the advanced packaging that TSMC does.

19:20That's starting to ease up a little bit. TSMC is a great company. They've added a lot of a lot of capacity. But now we're starting to realize that the other big constraint on data center growth is electricity. right there's just not enough electricity with everyone wanting to open up gigawatt scale data centers for ai the u.s grid just doesn't have enough of that in fact the only place that probably could even hope to cope with this is china where there's sort of different regulatory frameworks around building new electrical capacity so i think that's a big bottleneck here i think there's also issues around blackwell itself like you start to i've talked to people who are actually putting these systems in place it feels very much like a first generation system They've added a lot of new configuration to the racks and the servers that they've designed.

20:07And they're just not fully reliable yet. And there's a lot of complaints from people who have to go through all kinds of special tooling and get out special equipment in order to make basic configurations here. So I think all of those are sort of constraining Blackwell to some demand. Jay, you've been so generous with your time. I know you've got a run. You've got a pencil in your hand. I see you typing sometimes as well. before you go are you updating your price target tomorrow uh are you coming out with a note like what what's the conclusion that you have ahead of this call so i i i absolutely will come out a note i don't think i can comment on my price target but i i see this is a lot of things moving in the direction that i've been talking about for a while now jay goldberg we're gonna say goodbye to him he is a senior analyst for semiconductors and electronics with seaport research partners Reminder to everybody, he's the only analyst that's tracked by Bloomberg who's got a sell rating on NVIDIA.

21:02The price target that he has is$100, a far cry from the morning,$180 it was trading at before those numbers came out. Also with us is Sarah Fryer. She's Bloomberg News Big Tech Leader. And Mandeep Singh, Mandeep, is Bloomberg Intelligence Global Head of Technology Research.

21:24Hi, I'm Tom Keen, inviting you to join me for the Bloomberg Surveillance Podcast. It's about making you smarter every business day. I'm Paul Sweeney. We bring you complete coverage of stocks, bonds, commodities, even crypto. All the information you need to excel in the markets. And I'm Alexis Christophoris. Listen to us for essential conversations with the smartest names in economics, finance, investment, and international relations. That's the Bloomberg Surveillance Podcast. Subscribe today on Apple, Spotify, or anywhere you listen.