Nvidia Sees 70% Growth as AI Boom Accelerates

20:20And given that several of the safety features that Meta is committing to are really hinging almost entirely on their ability to correctly determine a user's age, I think there's a lot of open questions about how effectively all of this is going to work if this indeed is something that Meta and its competitors have been challenged with.

20:41Ed Ludlow:You and Riley Griffin put out the story where the headline is, kids will be kids. And that's the risk with Meta Safeguards, the opening. How does a company know for sure if a kid is a kid on the internet? There's actually like two ideas there, verification. But kids will want to push the boat on how many hours they can use on these platforms. Of course. And even just overnight and this morning, if you look on social media and you start to see kids reactions to some of this news, you see that, you know, some of them don't agree with it. Some of them think it'll be a good thing. Some of them, you know, just just just regularly, like some of them discuss how they have very easy time signing up for all of these platforms with very little pushback without using their real birthdays.

21:25So I really do think that there's sort of a tension between what the companies are putting on the parents, the kids, the families, and then what the parents, kids and families are actually going to do. But it remains to be seen, you know, whether any of Meta's efforts to really strengthen its age assurance technology are going to pan out.

21:44Ed Ludlow:Bloomberg's Alex Levine. Thank you very much. Let's stay on Meta. Joining us now is Danielle Sitrum, professor of law at the University of Virginia Law School, School of Law, who believes the outcome is, quote, a pittance for Meta. Let's start there. What do you mean by that, professor? You know, when Meta settled with the FTC for five billion dollars over the Cambridge Analytica affair, that was, I think, something like a couple of days worth of their advertising income. So let's just put in perspective the profits of$46 billion, I think, from last quarter or this part of this year. The first half of this year was something like$50 billion.

22:26So$17.5 billion over 10 years is a small amount of their income. It's a small hit they were happy to take.

22:36Ed Ludlow:So you are talking about the financial settlement. But for lots of people, the changes that face that Meta has agreed to to the Facebook and Instagram platforms are the news. Right. These are significant policy changes and operational changes. So in one example, two hour daily limits. And it is for the parents to authorize any additional time beyond that. That brings the role of the parent into this. Is that a sufficient mechanism for enforcement? it? As a parent myself, so I don't mean to demean parents. Parents are overwhelmed. They're overtaxed. And to put the burden on parents, you know, kids are very persuasive.

23:18So if they want to override and insist to their parents, like, you can trust me, you know, allow me to override this two-hour limit, they're going to do it. And I think relying on parents is a mistake that both law in the Children's Online Privacy Protection Act and sort of our current lawmakers approach. And, you know, parental control is the approach of the FTC. It's the current emphasis on current law and legal proposals. And I think it's a mistake. Parents are overburdened. And they often have their kids' best interests in mind. But they're pretty, kids are persuasive with kids. And for some kids, it's dangerous.

24:01So if you put parents in the driver's seat and kids are, you know, have emerging sexual orientation, they're experimenting their gender identity. It can be quite dangerous for parents to be in total control of their social media.

24:14Ed Ludlow:I've studied the court filing for the agreement. I see no legal requirement imposed on parents to monitor, enroll in the supervision. I see an agreement that regulates meta. Just that mechanism. Explain why it will or will not work in your view. So so parents. So under the child, the Children's Online Privacy Protection Act, which is one of the series of claims in the lawsuit. It requires that parents give verified parental consent for the collection, use and sharing of children's data. So data of kids under 13. And so in that regard, we are putting parents, law is asking parents to give permission.

25:00But as your prior, you know, as your news reporter just noted, kids often, you know, they get around that requirement by putting an age for their birth date that makes them over 13. So we get around parental consent in that first way. You know, the second is parents will often give that consent, even if kids are honest about their birth dates and putting parents in charge of that. So the default is under this settlement agreement that kids will have only two hours. So teens and that parents can override that default. You can imagine that that default will be easily overridden. You know, kids talking to their parents or simply doing it themselves.

25:43Hey, mom, can I have your iPad? Right? Can I have your laptop? That is, kids are pretty persuasive. They're really smart. They carve out privacy. We know this from social science that they do it themselves. I'm not convinced that the architectural promises made in this settlement will do very much. And unfortunately, the settlement doesn't prevent the amassing, the use, the sharing internally of Prometa for their AGI experiments of this data.

26:13Ed Ludlow:Daniel Citron, University of Virginia Law School professor whose scholarly work focuses on privacy, freedom of speech, civil rights.

26:27Ed Ludlow:Welcome back to Bloomberg Tech. NVIDIA is the big story. It's the biggest points driver on all the major indices that we're tracking. The NASDAQ 100 and the Philadelphia Semiconductor Index both hire. If something's changed, it's that some of the other names that were hired by association earlier in the session, now a little bit lower. But it is NVIDIA or BUS. Let's get out to Bloomberg Tech Equities reporter, Carmen Reineke, who has the charts we need to see. Yeah, Ed. I mean, if we think back to yesterday, sentiment going into NVIDIA's report was really quite cautious. So looking at where the stock is today, quite an amazing report from NVIDIA.

27:01Shares are up 7.5 % right now. This is on track for its best day since February. And something that we're monitoring here is that if NVIDIA goes up just a little bit more, it actually could overtake Microsoft's record for most market value added in a day. So Microsoft on July 30th, after its own earnings, added more than or nearly$450 billion in market cap in one day. NVIDIA is a little bit under that right now. And why are we having this incredible stock reaction? It's all about this revenue guidance. So the company said that in fiscal 28, so that's next year for NVIDIA, it expects revenue to grow 70%.

27:39That's more than the 45 % it had said previously, and it is exactly what investors wanted to hear. It really calmed concerns about an AI bubble, and it also sort of overshadowed fears about circular financing. And this was actually also a little bit of a conservative guidance. They said that they actually could see revenue growing even more if they have access to all of the supplies that they needed. So hugely bullish report from NVIDIA. And we're seeing the stock react in kind today.

28:06Ed Ludlow:Bloomberg's coming, Reineke. Thank you very much. NVIDIA delivered a strong sales outlook. That is putting it mildly. But it did warn that margins would narrow in the coming months, principally due to rising memory costs. Let's talk about the impact of that with Cody Acree, managing director and senior semiconductor analyst at Benchmark, who has a buy rating and a$335 price target for NVIDIA. And what is so interesting, Bloomberg's Ian King made the point. the top of the show. In the semiconductor space, if any CFO or CEO said this is what's going to happen to margins over the next six months, the stock would not go up like it did.

28:41Ed Ludlow:You seem sanguine about that as well. Yeah, I really was. And initially, the stock was down on that disclosure. So yeah, like you said, had any other company talked about a collapse of margin to 75 % to 71 to 72 in two quarters, and then recovering only about another 100 basis points, even after a price increase, that would have been enough to tank the stock. But as Carmen mentioned, that guidance for revenue really trumps everything. When they're talking about 70 to 100 % increase and 70 % in the bag for next year, that's a$300 billion annual increase, and that's$120 billion ahead of expectations.

29:27And so our numbers are going up above nearly 700 billion for next year from 400 billion in fiscal 27, calendar 26. That's taking our EPS estimates, you know, up nearly$3 for next year.

29:43Ed Ludlow:Cody, what links the margin story with the revenue growth outlook story is memory, right? You know, they were pretty clear that what they would have been able to do on growth were it not for supply constraints, this business would be doubling. At the same time, the emphasis on circular financing is NVIDIA saying we're just trying to help everyone get moving, particularly CollectCrest, the CFO, telling me they are helping people get supply. How did you assess all of that explanation on NVIDIA as the bank of NVIDIA in the ecosystem? Well, I think it's NVIDIA just trying to ease any bottleneck. As you said, the memory is the biggest constraint.

30:27They have agreements now stretching out three years. They've committed nearly$300 billion in capital commitments to ensure pricing and availability. So they said that they're well set for their expectations for at least the next two to three years. And then they're going to be working to ensure guidance capacity upside from that point. Regarding your other comments about circular financing, and I think one of the other big bottlenecks to the industry is just the availability of things like land power and shell and data center infrastructure. and on project financing. That's as critical of a capacity constraint as memory is in many cases when you have small startup projects that are just hamstrung by financing access.

31:20So if NVIDIA can loosen that bottleneck by providing their balance sheet, it's all the better for them because it all comes back to them exponentially in revenue.

31:32Ed Ludlow:Cody, on the analyst call, your industry peers asked about custom silicon and A6. And Jensen Wang continues to be dismissive of the threat of other options for accelerators. Do you factor that in or model that in for growth going forward? Yeah, we do. I think competition is a real thing that's increasing in the space. I think you've got guys like AMD that now have competitive stacks. You have Cerebris recently went public with its alternative. You've got all the hyperscalers, internal designs that are competitive. And even internal NVIDIA Silicon with its GROC acquisition is creating an alternative.

32:15But at the same time, even in the face of all this competition, NVIDIA is seeing a doubling of their indication of their revenue growth. And so even if you're seeing their market share go from, say, 85 percent to 75 percent or 70 percent, demand is so strong across the AI complex that it's continuing to push NVIDIA well above expectations. So, you know, their issue right now is simply supply. It's not a matter of demand. And that's even in the face of increasing competition.

32:46Ed Ludlow:Cody Ackrey, Managing Director and Senior Semiconductor Analyst at Benchmark. Thank you very much. Coming up, Hugging Face's newest robot is a duck that can walk and talk and even roller skate. Hugging Face co-founder and chief science officer Thomas Wolfe leads the robotics initiative at Hugging Face, and he joins us next. This is Bloomberg Tech.

33:15The people who seem to get more done than everyone else. They're not working longer hours or running on more caffeine. They've just stopped wasting time on the stuff that doesn't move work forward. Switching apps, re-explaining context, hunting for files. Those aren't small inefficiencies. They're hours wasted every week. Superhuman Go gives you those hours back. From the makers of Grammarly, Go is an AI assistant that sits inside every tab and tool you already use. always available and already aware of what you're working on. Ask it to draft something, summarize a long thread, pull up a file, or prep you for a meeting.

33:57Go handles it without you ever leaving the page you're on. This is what it looks like when AI actually fits into your work instead of adding to it. It's like having a teammate whose only job is to help you be better at yours. Go keeps up so you can move forward. With Go working with you, you can show off what you do best. Superhuman Go. Find out more at superhuman.com. That's superhuman.com.

34:46briefs, and scattered information that you have to grind through to turn into something useful can just become something useful. Put ChatGPT to work on your most ambitious ideas and projects. Get started at ChatGPT.com by selecting Work Mode, available on Plus and Pro plans. Aging is real, and so are the benefits of adding Vital Proteins Collagen Peptides to your daily routine. New Vital Proteins Collagen Sparkling Water, your daily glow-up now in three fresh flavors. Strawberry blossom, lemon lime, and blood orange. Improved skin health in as little as 30 days thanks to collagen peptides. Cheers to that.

35:23Or go with our classic collagen peptides so you can stay vital, stay you. Visit 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.

35:41Ed Ludlow:hugging face launched its second robot today called the micro duck aimed at making ai more accessible more playful the trainable biped robot can walk kick grab items and even roller skate and its open source software allows users to create new ai behaviors hugging face co-founder and Chief Science Officer Thomas Wolfe joins us, and we will discuss Microdug. But first, multiple media outlets have reported in the last 24 hours that NVIDIA is close to a deal to buy Hugging Face, Thomas, for roughly$13 billion. Neither company has responded to Bloomberg's request for comment. You're a co-founder of Hugging Face.

36:26Ed Ludlow:What can you tell us about NVIDIA buying your company? yeah nothing unfortunately i won't comment on that but i'm very happy to talk about my product today appreciate that no comment there were earlier reports that generally hugging face is exploring a sale to multiple parties are the co-founders considering selling hugging face generally i would say we've always had a lot of interest uh we've certainly been a lot in the news this summer, right, with the incident of OpenAI, with generally open source model going really, really well. So we often have offers both of acquisition and investment. Nothing very, very special this summer on this side.

37:15Ed Ludlow:And so to be clear, no comment and nothing in NVIDIA buying Hugging Face? No. Okay, micro duck. Let's talk about what we're here to discuss. second robot price point is interesting why why is hugging face want to put this out into the world well we think uh just like everyone is able to write code app just like now everyone is able to build software there is a time coming very soon where everyone will be able to use robot to write code robots and and we think we need the right robot to do that and according to us that's a robot that's very accessible in price, that's very open source, so you can put whatever model you want in it from OpenAI to entropy to open source models, and that you can use to also explore the latest development in robotics, namely world models or reinforcement learning techniques.

38:16Right. So that's MicroDuck. Is MicroDuck, we're showing some video of MicroDuck now.

38:23Ed Ludlow:So, MicroDoc in this particular video is depicted in what looks like a child's bedroom. This is something to have at home? This is something for the developer community or both? So, you know, Ed, I think we're creating kind of a new category of robots here that I think are both for tech aficionados and people who know maybe how to code or who know how to vibe code, but also for the whole family, just like AI started as something very tech-centric and now is used by everyone. So we expect this robot to be something people will buy to play with their kids, with their family members, but also to discover themselves, how to train behaviors for robots and maybe to transform like people can transform open source models or to tweak it, to invent new games, to interactivities.

39:16Ed Ludlow:Thomas Hagenface's core competence is open models right not necessarily on the hardware side could you just explain how how micro duck came together who you partner with on on the hardware of course so you remember probably uh two years ago we acquired a robotics company called Colin Robotics based in France team of 20 and actually when we acquired this team they already had started to work a little bit on a side project, which was the ancestor of Microdec. It was not fully working at that time, could not stand up when it fell. And so I thought, OK, we can't market it to a consumer yet, because if you have to put it back standing each time, it's not possible.

39:58Once we found the behavior that you see here, where it can rise again with its head, we thought, OK, there is something here. So that's a project that's been actively worked on for the past two years. That's one of our longest robotics projects ever.

40:13Ed Ludlow:There's been reports in recent days that Hugging Faces ARR has hit about$150 million. There's the backstory you've just given on robotics, and this is the second robot offering. In terms of the business model, where does this fit in? And how everyday people consume AI? Is this going to be a very serious business line for you guys, or it's something smaller scale? Definitely a growing business line for us, robotics in general. So our business, like you say, I mean, I won't comment on the exact numbers, but we've definitely had a surge in revenue in the past six months, right? We passed 100 million a couple of months ago, and we're definitely above that now.

40:53This is driven by the usage of open source model and data sets. And a lot of these data sets now are data set for robotics. That's the strongest growing category of data set on the hub. And so we think that growing the community that use robots will ultimately be beneficial for us. Now, this being said, already the selling, for instance, we had a previous robot called Richie Mini. We already sold more than 10 ,000 units. So that's already$4 to$5 million in revenue. So it's starting to be insignificant. And even for MicroDeck, I was quickly looking at the numbers. We started, you know, we opened the shop page just a couple of hours ago and we're already above half a million dollars sold.

41:34So it's going really, really fast. We're selling one micro duck every four seconds at the moment.

41:39Ed Ludlow:Selling one micro duck every four seconds. Thomas, before I let you go, you know, OpenAI published this report in the last 24 hours. I know you've been leading the robotics efforts, but the conclusion is that OpenAI said it could have reacted quicker, to where two of its models mistakenly, we've said, accessed and breached your platform. I've spoke to Clem at length about this in a 20-minute conversation, but just the latest findings, what Hugging Face's position is on that? I think it's still the same, which means that we think open source models were extremely useful to be able to defend ourselves.

42:15I think it's maybe in the short term we'll see a little bit less of these events because the companies will be more careful about the guardrails. But ultimately, I think it will be really hard to be always fully fault-proof. So there will always be cases of models that manage to escape in one way or another. So I think research on alignments, better aligning closed model, better aligning open source model as well is something that needs to be done now. It's very important.

42:42Ed Ludlow:Outside of OpenAI sort of publishing its findings of that investigation, Did they give you any assurances on being stricter, more thorough with the guardrails and the testing environment? Not specifically to us, but I think you've heard, right, that they've been posing. I think there's much more now guardrail and monitoring. One takeaway, though, from reading yesterday or two days ago reports from me is that monitoring what's happening is still extremely difficult because this model talk in their own language. There's a lot of events going at the same time. You see 500, 700, 1 ,000 agents at the same time doing multiple things.

43:24This is going to stay a challenging activity to monitor exactly what's going on for the coming month.

43:32Ed Ludlow:Hugging Face, co-founder, chief science officer Thomas Wolfe on MicroDuck, the new robot, and also an opportunity to comment on those reports of NVIDIA acquiring Hugging Face. As we just discussed, OpenAI admitted it could have reacted sooner to prevent an inadvertent hack that its AI models carried out on Hugging Face. The company said in a report published yesterday that it knew as far back as late May that AI models it was testing were leveraging a software vulnerability to access the open Internet. Bloomberg's AI reporter, Rachel Metz, has the story. This was the conclusion of an investigation.

44:07Ed Ludlow:Walk us through the disclosures and the findings. Sure. So OpenAI has been carrying out an investigation into what happened with its AI models that were during an evaluation process where they were not supposed to be accessing the Internet, able to take advantage of their vulnerability, reach the open Internet, and then breach hugging faces systems. And this all took place in July. But what OpenAI found as it looked into it is that there were some unrelated to hugging face, but related to models doing things that the company did not want them to be doing earlier, like as far back as May. So there were some alerts that were sent through systems.

44:48This happened a couple times. And OpenAI continued with the test that it was running on its models at the time. And in hindsight, as the company noted, it's possible that it could have noticed some of these things and taken action sooner in order to prevent what eventually became a breach of another company's systems.

45:08Ed Ludlow:Both since this report was published, but also just over the last few weeks, a lot of the industry has pointed me to human error. You know, the actions that humans did or did not take. What can you tell us about that in the report, about who they attribute the blame to? Sure. I mean, there is no question here that in this situation and in a bunch of other situations that we've seen lately where AI models have done things that are not what people or companies would want them to do, that ultimately, I mean, I think you are looking fundamentally at people making decisions that are leading to machines to do things.

45:48But what is important here is that they said that the systems that they had in place, which they are improving, those had sent out some alerts. And in a few different situations, people decided, you know what, this is OK and we can continue the tests.

46:06Ed Ludlow:How much does this sort of close the book on what happened and what is OpenAI doing in response, changing changes to its policies? That's a really good question. I mean, I think that as far as like closing the book on it, I think there actually could be more to look into here. There was one third party report that also came out yesterday from Meter and Redwood Research. These are nonprofits. And it was interesting what they were looking at because they're external to OpenAI, but also they were only looking at a small piece of this overall incident. So I think it'll be interesting to see if anybody really does a postmortem on all the individual incidents that OpenAI reported before the hugging face breach.

46:49That would be interesting and could shed some light. And I think it'll be interesting to see what the industry does going forward. I mean, there's been some talk of companies working together, different tech companies, different AI companies in order to come up with some standards for testing these models. But I think the fact remains that they don't have any kind of standards as of yet. And in some cases, these AI models are going onto the open Internet.

47:15Ed Ludlow:Bloomberg's Rachel Metz, terrific reporting. Thank you very much indeed. Let's get back to NVIDIA. The company's results are helping ease fears of an AI bubble. But Bloomberg Intelligence points out that despite its strong financials, the credit market still appears skeptical, trading its credit default swaps more like those of triple B rated companies rather than its double A rating. Let's discuss with Robert Schiffman of Bloomberg Intelligence. Bring me the credit react thesis, please. Well, listen, first from the equity perspective, you know, if you're a bull, are you not entertained? This may be the best print in the history of the stock market and the best forward guide in history.

47:54So that's great. Why are bondholders not loving it? Because it means there's going to be a lot more debt being issued. How is everyone going to pay for all of NVIDIA's revenues? They're going to do it through borrowing. NVIDIA is going to put up their pocketbook. So spreads are basically flat. You know, there's no reason necessarily to rejoice. And quite frankly, this doesn't answer all the questions that people are asking about ultimate demand. This means hyperscaler demand is there, right? The same guys who are buying are just buying more. But the people who are ultimately going to be paying for AI, enterprise customers, you know, we need to see that through AI monetization from hyperscalers, not NVIDIA.

48:29Ed Ludlow:So, Shift, really quick. CFO CollectCrest told me the hyperscalers are IG. The Frontier Labs are not IG. Why was she hammering home that distinction? Well, because the hyperscalers can borrow as much money as they want. They're AA and AAAs, and they can raise hundreds of billions of dollars. But the neoclouds and everybody else that wants to finance this build, they can't do it on their own. So, NVIDIA has to put up their own capital, at least as a backstop, at least provide credit support, at least provide residual values, because they believe in their own chips to help them pay for everything.

49:03And eventually they'll get paid back. So that's why they're differentiating between investment grade and non-investment grade financing needs. All right.

49:13Ed Ludlow:Bloomberg's intelligence is Robert Shiffman. Really appreciate your time and the credit analysis and what has been a pretty wild equity market story. That does it for this edition of Bloomberg Tech. Well, you just heard Shiffman say it, maybe the greatest earnings print in history. That's for you to decide. But you can recap all the conversations we had on the Bloomberg Tech podcast. You know where to find it on all the Bloomberg platforms and online on Apple, Spotify and iHeart. Wow, what a Thursday. This is Bloomberg Tech.

49:46When you're running a business, the best days are the ones where priorities stay on track. For midsize and large companies, risk can affect multiple parts of the organization at once, from property and liability to cyber and regulatory challenges. At that level, managing risk becomes an ongoing discipline. At the Hartford, the focus is on helping businesses manage risk before it turns into something more disruptive. And when losses do happen, that work is paired with insurance coverage shaped by years of underwriting, risk engineering, and claims experience. Learn more at thehartford.com slash risk mitigation.

50:21Policies provided by Hartford Fire Insurance Company and its property and casualty affiliates, Hartford, Connecticut. Wait, I came in for two things. How is this$47? All right, we're going to need a plan here. Just start simple with Bank of America Advantage Safe Balance Banking. No overdraft item fees, no monthly maintenance fee if you're under 25. Plus, as a new checking customer, you can earn$100 when you open an account and make qualifying Zeller debit transactions. Oh, that's actually really simple. Safe balance banking. One less thing to figure out. Learn more at bofa.com slash earn100. Terms and conditions apply.

50:55Bank of America and a member FDIC. 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. That'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 odoo.com That's odoo.com