We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode Instant Reaction: Nvidia Gives Solid Forecast, Despite China Concerns

Instant Reaction: Nvidia Gives Solid Forecast, Despite China Concerns

2025/5/28
logo of podcast Bloomberg Businessweek

Bloomberg Businessweek

AI Deep Dive AI Chapters Transcript
People
J
Joe Kaiser
M
Mandeep Singh
Topics
Mandeep Singh: 我认为,尽管英伟达在中国市场面临80亿美元的收入损失,但其整体业绩展望依然稳健。如果没有中国市场的损失,英伟达的收入本可以更高。然而,英伟达似乎能够通过其他客户的需求来弥补这一损失,例如来自阿联酋和其他企业客户的订单。长期来看,失去中国市场意味着失去潜在客户,因为许多新兴的LLM公司都在中国。不过,我认为英伟达的芯片在性能上仍然具有优势,只有大型云服务商才会寻求替代方案以降低依赖性。如果英伟达失去中国市场,投资者可能会担心其对大型云服务商的依赖。此外,我认为主权AI是一个巨大的机会,例如来自阿联酋和沙特阿拉伯的收入。许多实体希望建立自己的大型语言模型基础设施,这对于英伟达的增长来说是一个真正的利好因素。最后,我认为英伟达应该效仿苹果,通过股票回购等方式向股东返还现金。 Joe Kaiser: 我认为,英伟达的增长和盈利能力是前所未有的,远超微软和苹果在相似发展阶段的表现。我认为英伟达还有很大的上升空间,因为AI在企业应用中的部署还处于早期阶段,英伟达有很长的发展道路。Blackwell芯片在降低功耗方面取得了重大突破。美国有足够的资源和技术来解决电力问题,建设电厂的主要障碍是审批流程。

Deep Dive

Chapters
Nvidia's Q2 forecast met analyst expectations, despite an $8 billion revenue loss from China due to export controls. The company's ability to offset this loss with other customers highlights its strong market position and diversified clientele. This positive forecast is impressive considering the significant revenue hit from China.
  • $44 billion revenue (would have been 5% higher without China losses)
  • Q2 forecast in line with analyst estimates
  • Significant revenue loss from China ($8 billion)
  • Stock rose 4% in extended trading

Shownotes Transcript

Translations:
中文

This is an iHeart Podcast. Your best bottling plant employs 3,300 people. How do you get 3,300 people working at peak efficiency? Your best store has reduced waste, water, and energy usage. How do you make every store like your best store? Your best property has every guest raving. How do you make every property like your best property? The answer is Ecolab.

Better performance, better outcomes, better impact. Ecolab. Now every location is your best location.

Hiscox Small Business Insurance knows there is no business like your business. Across America, over 600,000 small businesses, from accountants and architects to photographers and yoga instructors, look to Hiscox Insurance for protection. Find flexible coverage that adapts to the needs of your small business with a fast, easy online quote at Hiscox.com. That's H-I-S-C-O-X dot com.

There's no business like small business. Hiscox Small Business Insurance. Bloomberg 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.

NVIDIA is out, folks. Let's get to it. Right now we've got in studio our Mandeep Singh, Bloomberg Intelligence Senior Tech Industry Analyst. Mandeep, tell us what you like, what's interesting in this report. I mean, had it not been for the China write-down, this company would have posted revenue which was 5% higher than the $44 billion number. Right. And the guide would have been almost 15%, 20% higher. So we're talking about $8 billion in lost revenue. So clearly...

Whatever they're doing, they are able to make up that revenue lost in China with some other customers because it's in line with consensus. Where are they making up for it? I mean, there are other customers. We saw the deals in UAE. I think there is demand from other enterprise customers. And the one anecdote they provided was token generation was up 10 times. Explain what that is. Yeah.

The way these AI models work is if you type in a query on ChatGPT, it's going to generate, you know, hundreds of tokens from that. It's going to be passed through the model and the output is also, you know, hundreds of tokens. So Google said,

at their IO event, their token generation was up almost 50x. Same thing from Microsoft. 50? Yes. And so we are getting those type of numbers from pretty much anyone who is exposed to demand for large language models. And in this case, NVIDIA is exposed through ChatGPT because all the weekly active users that are going on ChatGPT

and using the product for, let's say, 20 minutes. Right. That's what's resulting in the 10x increase in tokens. And I'm not surprised that everyone needs additional compute because if ChatGPT needs to serve more users, they need additional compute. The traditional CPU compute is just not good enough. And that's where I think they are finding that additional demand.

Our Ian King on our live blog says, how many companies could front load their earnings release with details of what they're missing out on and still get a positive reaction from investors? So it's pretty impressive. Okay.

We just had a guest on, Jay Goldberg, over at Seaport. He's got the only sell on the company. And he's got some questions about maybe all the good stuff is already factored in. I think questions about the AI and the build going to take longer. What do you say to some of that?

I mean, look, everyone acknowledges that, you know, semiconductors is cyclical. You will see an air pocket or some sort of digestion. And these kind of these kind of growth rates are too hard to sustain, you know. So you are not going to see 60, 70 percent growth when your revenue base is 200 billion dollar run rate annually.

I mean, from that perspective, there will be a tapering of growth. But look at all the hyperscalers. NVIDIA is still growing top line way above everyone else, including Microsoft and, you know, Meta that are growing at a very healthy high teens rate. So from that perspective, there is still a lot of runway when it comes to NVIDIA's growth. China and the China specific story here, the age 20, the restrictions, everything.

What do you see as the long-term effect of what the Trump administration has recently said, but also export restrictions on some of NVIDIA's products, specifically with regard to China? I mean, that is the most plausible bear case against NVIDIA. Because look, right now, China...

to be about 20% of NVIDIA's revenue. It's very hard to fill that hole with all the additional demand, even though chat GPT and other LLMs need more compute and they can take whatever NVIDIA is making. But in

But in the long term, if you're looking five years out without China, I mean, that's where all the secondary LLM companies are besides, you know, ChatGPT, Anthropic and Google. And so you are missing out on a lot of potential customers that would have used your chips. What did we say about DeepSeek before that they had an update? And like we're trying to understand. So.

If this is all real with what they upgraded, the R1 AI model that helped propel it to global prominence earlier this year. Now, we don't still know exactly how they're doing all of this, but if they are doing it with less advanced chips, I mean, does that make the NVIDIA story not as exciting?

No, because NVIDIA's... And again, this is all speculation. Yeah, yeah. And look, there could be some pull forward in China, you know, with DeepSeek using NVIDIA's chips because they anticipated these kind of restrictions. Right, right. But in other regions, everyone realizes NVIDIA has the best chip. The only reason why you are going to look for an alternative is a hyperscaler looking to really...

kind of reduce their dependency. So Google doesn't use NVIDIA chips because they have their in-house chip and they serve traffic that's pretty much the seven most frequently used apps on the internet. So if they were to rely on NVIDIA, I mean, NVIDIA's revenue could actually grow 50% just by serving Google. Just...

But does Google then send those chips? Would Google sell those chips outside? They don't. So they don't use the chips outside of their family of apps. They just use it for Google search, YouTube, and their family of apps. Why wouldn't they sell it? Because it's not...

a chip that is generic in nature, that it can serve everyone else's traffic. It's customized for the data. Do we see Amazon then doing that too? I think that's the bare case that everyone is thinking, but so far it hasn't happened and everyone is trying. Would you disagree with Jay Goldberg who just told us that he thinks that Broadcom is the closest competitor right now or not necessarily closest competitor, excuse me.

And I know you can't make calls, Andy, but he's the only one who has a sell rating on NVIDIA. And he argues right now that investors should be putting their money toward Broadcom.

I mean, look, we are in an AI infrastructure super cycle. And this is a tide that will lift all the boats. Broadcom will benefit. Other chip makers will benefit. So it's very hard to say only Broadcom will benefit with AI infrastructure demand. And I think that's where you're too early to put a sell rating on a stock like NVIDIA. All right. So stock is up 4%. And certainly investors buying in the aftermarket. There is a headline that crossed...

about 10 minutes ago, that NVIDIA is saying maybe unable to create a competitive product for China would have to foreclose from competing in China market. China market foreclosure would materially hit the business. Export controls applicable to China are complex. Help me make some sense out of these headlines. I mean, look, if you want to see multiple expansion in a stock like NVIDIA, you've got to convince investors that you can keep growing 20%, 25%.

And the way you do that is by showing pockets of demand, you know, geographic diversification. If you lose out China revenue, suddenly everything is like, OK, you're too dependent on hyperscalers. What if Amazon actually develops a chip and reduces their dependency altogether on NVIDIA? And so that is the big bear case, the concentration of customers. And that's where losing China revenue is a risk. So, Mandeep, is this just saying, here, we're laying it out for you just to understand?

Yeah. In terms of China. Is that pressure on the Trump administration? Not really, because they keep eating their numbers like this. I mean, the administration, you see the numbers and you're like, these guys can manage without China revenue. But if they're warning of that impact in terms of China, is that kind of a little bit of a subtle, considering he just traveled with him to the Middle East, you know, a little bit of a like subtle message reminder? That's a fair characterization. I would agree with that. Okay. Yeah.

I want to bring in Joe Kaiser, CEO and Managing Director at Mercado Partners. It's a growth stage investment firm. It's got over a billion dollars in assets under management. Mercado Partners, he joins us from Salt Lake City. Joe, you've had a chance to sit back and look at some of these results over the last 30 minutes or so. First quarter data revenue coming in close to estimates. Second quarter forecast solid despite an $8 billion in expected lost China sales. The company's stock up about 4% in the after hours. What's your view?

I think this growth and the profitable growth that we're seeing is really unprecedented. In the venture community, we still use a term called rule of 40. So it's combining the top line growth rate coupled with EBITDA margin. And I haven't seen the EBITDA number, but if we use the prior quarter's EBITDA number, like they're approaching 200 from a rule of 40 standpoint. And to give you some historic perspective,

when Microsoft and Apple were at the same stages from a revenue size standpoint, those companies were at 90 and 60 respectively. So this is truly unprecedented performance. And at the end of the day, what we're seeing is

probably the greatest invention in the history of mankind and the company that singularly powers that. And that's the performance that we're seeing before us now. Are you bullish that there is more upside, Joe, to this company? Jay Goldberg is not.

Well, I think it pays to be provocative at some occasions. But yes, I think there is a lot of upside. I think, you know, when we think about the deployment of AI, particularly in enterprise applications, both in the United States and abroad, we are very early innings in that deployment.

They've talked about the data center deployment around the world, across the GCC countries, as well as domestically. Lots and lots of runway in that regard. So there is a long road ahead for NVIDIA. Joe, I know you're well-versed on the power side of AI, but I want to bring Mandeep in on that. Mandeep, is there a point...

You know, I think about, you know, doing like basic charts and we're like the ramp up in AI and the power demands. We talk about it all the time. Is there a point where AI deployment and usage might be limited because we just don't have the power? Yeah. And that's where actually NVIDIA's architecture excels because they give you performance per power watt.

that your likes of AMD and Intel aren't able to give. It's not as if AMD doesn't have an offering GPU that can compete with NVIDIA, but NVIDIA's GPUs just give you better performance per power watt. And that is the constraint right now. Everyone acknowledges the power constraint. So from that perspective, also NVIDIA does have an advantage.

Joe, I see you nodding. This is certainly your wheelhouse, given the history of what you've done at the firm and outside of the firm, too. How is NVIDIA doing this, and how solid and how big is their moat? Well, that is one of the big developments with the Blackwell chip, is the reduction in power consumption from Hopper to Blackwell. And that is one of the key breakthroughs.

A lot of people, including NVIDIA, including the hyperscalers, talk about the lack of power as the headwind here. The reality is we, as a nation, we have the resources. We have the technologies. I invested in a company called Taurus, which has solved the storage challenge.

We have, from a long-term standpoint, I think a guest earlier today talked about small modular reactors and the technology of a small footprint nuclear. That technology exists today. But SMRs are taking a while. Yeah, we're still talking...

quite a, quite a runaway. Everybody talks about it and SMRs are the way to go, but it's going to be several years before we really see that build up. Very true, Carol. But I think the challenge is we have plenty of natural gas as a country as well. And, and the talk is, well, it takes a decade to build a power plant, but the reality is the long pole in the tent from building a power plant is the permitting process and the process of getting a

the right to turn on and use the power out of that power plant. So if we streamline that process, and I know both the federal government as well as states like Utah are in the early innings of streamlining, but that is the long pole in the tent.

All right. We're talking with two guests. We've got Joe Kaiser. He's CEO and Managing Director at Mercado Partners out there in Utah. Mandeep Singh, Senior Tech Industry Analyst, Bloomberg Intelligence, right here in our Bloomberg Interactive Broker Studio. NVIDIA out. Tim, stock is now up about 4.5% here in the aftermarket. Yeah. First quarter data center revenue coming in close to estimate. Second quarter forecast is solid despite that $8 billion in expected loss to China's sales.

Nvidia shares are rising in extended trading. It's providing a boost to other chip stocks as well. We're seeing shares of Micron also higher in the after hours by about 1.4%. Mandeep, you've been glued to the terminal looking through these numbers. You got to go in a minute. What's the last thought you want to leave us with ahead of this call? I mean, sovereign AI is a big opportunity. Sovereign AI. Which is where you would put revenue from the likes of UAE, Saudi Arabia, the deals they just signed.

I just want to get a sense of how big this could get, because so far this was like low double-digit revenue in terms of the revenue exposure that NVIDIA had. And look, they are losing China revenue. So what makes up? How many...

types of entities, sovereign entities, want to build their own large-angle model infrastructure like the UAE and the Saudi Arabia. And that would determine, you know, the pace of growth when it comes to the non-hyperscale piece of NVIDIA's growth. The Middle East trip, like, makes so much more sense, right? Maybe they should be doing more trips with the administration. Well, where else can they go? Anywhere, right? Yeah, anywhere. I mean, look, you've got so many pockets where they...

There is so much sensitivity around AI in terms of developing your own AI, serving it with your own infrastructure, that that's the real bull case when it comes to sovereign AI. Can I ask one last question? Cash, cash equivalents, marketable securities, $53.7 billion up from $31.4 billion a year ago and $43.2 billion a quarter ago. What do they need to do? They tried to buy ARM, couldn't do it, right?

Do they need to do bigger dividends? Do they need to do buybacks? Or what do you think? I mean, stock buybacks. This company is generating as much free cash flow as Apple now. We're close to $100 billion in free cash flow. So Apple does have a playbook when it comes to returning cash to shareholders. I think Nvidia should probably follow them. Thank you.

Colgate Total may make your favorite toothpaste, but it's also a science innovator committed to oral health. For instance, the Colgate Total Active Prevention System with

With a cutting-edge toothbrush, refreshing antibacterial mouthwash, and a reformulated toothpaste, with a technology so innovative it won the 2024 Edison Patent Award, the Colgate Total Active Prevention System is 15 times more effective at reducing bacteria buildup to fight the root cause of oral health problems in six weeks, starting from week one, compared to a non-antibacterial fluoride toothpaste and flat-trimmed toothbrush. Talk about science.

Get the Colgate Total Active Prevention System today so you can be dentist ready. Shop now by visiting shop.colgate.com slash total. Hiscox Small Business Insurance knows there is no business like your business. Across America, over 600,000 small businesses, from accountants and architects to photographers and yoga instructors, look to Hiscox Insurance for help.

This is an iHeart Podcast.