OpenAI’s & NVIDIA's $100 Billion Marriage, Meta’s Sloppy Vibes, TikTok Deal Arrives?

0:00NVIDIA plans to invest$100 billion in OpenAI, but will the full deal ever come to fruition? ChatGPT wants to give you morning updates on your interests. Meta has a new feed of AI slop. And the TikTok deal is here, seemingly. We'll cover it all on a Big Technology Podcast Friday edition right after this. Welcome to Big Technology Podcast Friday edition, where we break down the news in our traditional cool-headed and nuanced format. We have so much to talk about with you today. Great show coming up. We're going to cover everything from this massive tie-up between NVIDIA and OpenAI, where NVIDIA is going to invest$100 billion in OpenAI, but seemingly all that money is just going to come right back into NVIDIA's pocket.

0:42We'll also talk about this new ChatGPT feature called Pulse, this new feed that Meta has of AI-generated images and video called Vibes, and the fact that we are very close to a TikTok deal that will enable the app to continue to operate in the United States. We'll look at whether it's fair and what is going to be the result of this transaction. Joining us as always on Fridays is Ranjan Roy of Margins. Ranjan, great to see you. We got TikTok, Pulse, and Vibes. We're basically Gen Zers today, Alex. I'm excited for this. If one of us say the term no cap, then I think we will blend into our topics.

1:22So cringe. So cringe. I don't think we're going to pull this off, Ranjan. No, I don't think so either, but we tried. We did try. And speaking of trying, we have two companies really trying to pull something even more daring off, and that is this$100 billion investment that NVIDIA is planning to put into OpenAI. Here's from the Wall Street Journal. NVIDIA and OpenAI, two U.S. giants powering America's race for AI superintelligence, outlined an expansive partnership on Monday, including plans for an enormous data center build out and a$100 billion investment by the chip maker into the startup. I love how like the Wall Street Journal is already adopting the term super intelligence.

2:04The deal announced Monday will allow OpenAI to build and deploy at least 10 gigawatts of NVIDIA systems for its artificial intelligence data centers to train and run its next generation of models. The amount of electricity is roughly comparable to what is produced by more than four Hoover dams or the power consumed by 8 million homes. One key part here is that NVIDIA is going to make its investment in OpenAI progressively as each gigawatt is deployed to support data center and power capacity. That could allow it to hedge its risk should OpenAI not be able to continue growing at its current rate.

2:47Is this investment ever going to come to fruition? I mean, this sort of stepped process where we'll just keep investing our money in you as you grow, but it is contingent on you becoming or maintaining what Jensen called the fastest rate of growth in American software history. I think the number$100 billion is a very purposeful number. And I think it's exactly because$90 billion wouldn't have been exciting,$80 billion,$10 billion to start, and then an increasing stepping up of investment. $100 billion just sounds exciting, and I think that's why the number is there. And I don't know, a couple of – when was Oracle at$300 billion?

3:31Was that last week or was that two weeks ago? I believe so. Yeah. Yeah, time just is a flat circle when it comes to AI investment. What's$130 billion between friends, right? Well, yeah, and it is notable that Sam Altman and Jensen Huang apparently negotiated the deal with no banks involved. That's my favorite part because that's one of those that on one hand you can see people, especially on the tech community, saying the bankers aren't getting a cut of this. this is great. But then on the other hand, just trying to think through like, the complexity and the size of a deal like this, and it basically was just Jensen and Sam talking.

4:12So I think, like overall, I don't know, again, to me, when you're saying, will the money ever show up? I think so much of this feels right now, even the Oracle investment or the contract, I guess, which was it was still very unclear what happens if open ai cannot pay it in this with this one as well all of these it feels like they're just trying to secure their dominant position for the next five years and say we are already putting the money out there even though it doesn't exist now so we can remain dominant well let's talk about like you know your question of like whether OpenAI can afford these chips.

4:51It seems like NVIDIA is going to give OpenAI money and then OpenAI is going to buy chips from NVIDIA. Here's from the journal story. OpenAI will use the cash from NVIDIA's investments to help pay for new chips produced by NVIDIA. A circular arrangement that allows the chip company to turn its balance sheet cash into new revenue. Such circular arrangements are common in the AI world and have raised questions about the extent to which new sales reflect genuine market demand versus capital recycled within the industry? I mean, I guess NVIDIA would have to get that cash somewhere. Is it just from its stock that it sells some stock and then gives that money to OpenAI and OpenAI buys chips?

5:35And then NVIDIA has a revenue surge and then it goes back to Wall Street, which further inflates the valuation, which leads to further purchases while people may or may not use chat chpd it's a good deal is this a crazy way to think about it how is this going to work i think that's a very reasonable way to think about it and i think that's what it is um but but you look at it if we reach super intelligence if you there's also the mark zuckerberg quote this week around how i forget exactly what it was but it was something around the idea of like like the the insane decision would be to not spend when it's this once in a, not even generation, but this epochal shift in just technology overall.

6:17So if you have a chance to dominate, and only a few will, you have to spend whatever that money is. It feels like that's just the way people are approaching it, and the market is rewarding it right now. If the market were ever to not reward it, clearly things could unfold very quickly. But at least as of now, everyone's sitting fairly pretty. So OpenAI is expected to bring in around$13 billion in sales this year. And I'm looking at these numbers. I mean, and by the way, it's going to lose money, right? We think the data is or the expectations is that between now and 2029, it's supposed to lose more than$100 billion.

6:59To me, it's always been like, well, eventually these companies are going to need sales, right, to end users. whether that's companies building with AI or users using ChatGPT, or else none of this can actually happen. But I'm starting to second guess myself there. I mean, is there an end of the line here at some point, Rajan? What do you think is going to happen in terms of the financials here? I mean, I think that what you just said right there is such a perfect encapsulation of the moment where, as a very intelligent person, you just started to out loud question whether a business needs to ever make money to actually survive.

7:38That's my first problem, right? No, no, I mean, but given exactly what you said, making$13 billion this year, forecast to lose, I think it was$120 billion by 2029. And then it was interesting, too, I think it was Fiji SEMA, I was listening to a CNBC interview where someone from OpenAI was talking about how, like, you know, whatever the demand is right now, it's going to triple, quintuple every year. And we're hitting exponential growth around demand for compute and electricity. So you have to start planning for it. So overall, it's just fascinating to me that no one has tried to pretend that there's a clear revenue growth path to meet this.

8:25So it has to be from funding of some sort, and maybe people will keep plowing money into it. But it certainly none of it adds up in a simple way the stuff that will come out of this super brain will be remarkable in a way i think we don't really know how to think about yet any guess whose quote that's from of course it's from sam altman and i'm just gonna like i i'll say this i think i have a more positive view of sam altman uh than a lot of the ai critics out there uh i I mean, I wouldn't even put myself in the AI booster or critic camp. I feel like I'm just trying to think reasonably about this, which, again, is a sin in today's day and age.

9:06I think there are some people that find that Sam Altman is just a charlatan. I don't think that's the case. He's obviously led the development of OpenAI, which continues to surge after a lot of its top talent has left. So kudos to Sam for that. He ushered in ChatGPT. he's responsible in large part for popularizing this wave of technology. But that being said, I looked at that quote in the Wall Street Journal, and I'm not saying I believe this, but one of the thoughts that popped into my mind was, is the entire global economy being led by somebody selling this fever dream fantasy that will never be realized?

9:47And how is anybody believing these statements? Again, it's worth reading again, the stuff that will come out of this super brain will be remarkable in a way I think we don't really know how to think about yet. What does that even mean? Who could be swayed to put billions of dollars into something that there's a dream that things will happen and we have no idea about what they are yet? So give me all these crazy amounts of money. I really had to stop when I read that quote. I stopped because I kind of love Superbrain. And I don't think I've heard, you know, you hear super intelligence, you hear an ASI is all the rage now, but now we got super brain.

10:32That's going to be the next official kind of benchmark in the overall AI race. But yeah, I agree. It's ridiculous. But then on the actual kind of like product side of things, and we're going to get into this idea of Pulse. But already, to me, one of the interesting things was we've debated this a lot. GPT-5 does a lot more. And this whole move towards doing things rather than just thinking and being a thought partner. But in doing things, there's like a number of articles that came out around how by defaulting to GPT-5, overthinking for a lot of simple queries is something that ChatGPT as a product is doing more.

11:16Or OpenAI even recognized this and commented that they would try to work on it or fix it. But that's compute. Like they are creating ways. I was thinking about this. I have never seen a chat GPT query that just answered my question and stopped there. There's always a follow-up. Would you like me to do X or Y? And it's funny because it's built. Obviously, that's like any growth hacker knows that there's going to be some kind of tricks to actually driving more utilization. But in reality, this demand for compute, they are pushing themselves within the product. And maybe that maybe they keep losing more money because, you know, they're not raising prices significantly, if at all.

11:59But they're going to make us all just suck up more compute to just plan our hike in Nepal. Right. So this is going to be kind of building on that an unanswerable question. but I feel like it's worth teasing out the question at least, and we'll try to figure out where we land on it. This$100 billion, is this going to be money that's being used primarily to scale up larger training runs? Or do you think it's going to be primarily used for inference and sort of getting things done on the models that we have today? Because you know whether you're a model or a product person doesn't really matter the the thing we've seen in recent months is that you know maybe there's diminishing returns from scale maybe there's not but certainly to get the same oomph or the same impact from scaling up these models and training uh is not as easy as it was it's it's almost like you have to work much harder to get that you know uh you know exponential improvement and we we really haven't seen it come out of of these labs uh yet so um or at least not without difficulty and expense so where do you land on this well yeah i mean where that expense will be i am assuming both both on the kind of like model training as well as the inference i think because like they have to they have to drive compute to make any of this make sense.

13:34If they start to show that the demand for actual compute is decreasing, even if utilization is increasing, but if aggregate demand starts to increase, that actually hurts everybody. And then I read something where it is that AI-related stocks have accounted for 75 % of S &P returns, 80 % of earnings growth, and 90 % of capital spending growth since ChatGPT launched in November 2022. And it's become kind of like a standard talking point that the AI economy is driving the entire U.S. stock market, is driving overall aggregate performance while non-AI companies are falling behind. But it has to, the thirst for computer, it has to just keep going.

14:21Otherwise, none of this works. Right. It's getting to the point where it's bigger than just like, if this doesn't work, then, okay, so NVIDIA's stock will decrease a little bit and everyone will be maybe endure a down year on the stock market, but it won't be systemic. I want to float this out to you, right? So not only is AI accounting for that large of a percentage of earnings on the stock market, it's also now responsible for some very real economic activity, whether it's the data center builders, the power plants, like if this all slows companies that may be borrowing, there's a lot of debt here, especially those that are trying to provide the power to these data centers and build these data centers that, you know, they could go under.

15:07I don't think this is a systemic risk, but you also like see it spreading. Now NVIDIA is investing, what,$5 billion in Intel and making, you know, investments in OpenAI and elsewhere. And they're a big shareholder in CoreWeave. And so many companies, you know, both in tech and outside of tech are now integrating an AI and, you know, in tech's case selling it. You lived through a financial crisis. Is this something where we have risk for an AI-inspired, let's say if things slow down, an AI-inspired financial crisis? I don't think so, mainly because exactly as you said a moment ago, it's still a concentrated group of companies that have benefited so far.

15:53So the actual unwind would still be within those companies. There's already been plenty of companies that have been decimated over the last few years because they were not the beneficiaries of this. So I don't think that kind of systemic risks exist in the same way. Obviously, market downturn and what that leads to and kind of unknown effects are always potential there. But I don't think I think this is more of an unwind rather than some kind of calamity. And and I I have like seen many after living through the final global financial crisis in 2008. I think there's like a decade where every single signal made me think the next one was coming.

16:39And maybe I'm going to eat my words significantly here. But I still see this more of a like, as you said, NVIDIA stock drops. there's some ancillary negative effects, but it's not some like economy-wide disaster. Well, I am also just like kind of questioning all the spending because it just seems to me that if you're going to spend so much, you're going to have to, like the thing that I think would really merit it is if these models were getting much better if you built these massive models and massive data centers to train them on. Maybe that will happen, But certainly with GPT-5, we saw some of the more specialized training, like coding, for instance, medical stuff, you know, actually start to prove out as opposed to the generalized scaling of the models overall.

17:30So I wonder, you know, when I was reading these stories, I thought to myself, it would be great. I don't even know if it would be great. It might be interesting to do this. Like, what would be the interesting counterfactual if you could do this a little bit slower? but you can't do it slower because you need the money to serve the models. I wish you could. I wish you could. I wish we all could take a breath and build the products. I think to me though, I'm curious your thoughts. We had talked about this the other week how a lot of what will show, a lot of the value created by AI will actually not be shown in like aggregate output in GDP.

18:09Like you have certain like efficiencies and ways like a lot of it will be benefiting bottom line, not necessarily top line. And obviously, like, you know, unleashing new waves of growth. In reality, if the whole thing is you're displacing large swaths of the white collar workforce who then no longer will have spending power, like, sure, there's going to be more compute used and more people spending money within a few companies, but as like an economy-wide force multiplier, I don't know, maybe that's too short-term bearish, but it's still like, will GDP as we measure it grow the way everyone's promising based on just the way AI works?

18:53I'm not sure if I believe that. Well, OpenAI is definitely trying to push that narrative because this week they have this new evaluation called GDPVal. This is a headline from their side, measuring the performance of our models on real world tasks. People often speculate about AI's broader impact on society, but the clearest way to understand its potential is by looking at what models are already capable of doing. History shows that major technologies from the internet to smartphones took more than a decade to go from invention to widespread adoption. Evaluations like GDPVal help ground conversations about future AI improvements and evidence rather than guesswork and can help us track model improvement over time.

19:37GDPVal will span 44 occupations from nine industries and they've been meticulously crafted and vetted by experienced professionals with over 14 years of experience on average in these fields. So at least they're trying to measure it. I got to say, I had somehow not heard of that at all. And I loved it. Exactly what I was trying to think about. One week after. Sam had something. Sam had it ready. GDP Val. I think, yeah, I mean. It is interesting to see the OpenAI blog posts trailing the big technology podcast conversations by a week. Last week, it was the three faces of Gen AI. Now it's GDP Val.

20:19What's next? What's next? What's next? Oh yeah, they are looking at real estate, rental and leasing, things like concierges, real estate brokers, government, recreation workers, compliance officers, manufacturing, mechanical engineers, and industrial engineers. Then there's also lawyers, software developers, accountants, and auditors, even RNs, registered nurses and nurse practitioners. It is interesting. It's sort of like an admission from OpenAI. Hey, we're going after all these occupations. Yeah. I mean, if you think, okay, think about like a real estate broker. Again, the very promise of the internet was supposed to be to actually displace kind of middlemen and real estate brokers.

20:58AI certainly helping you quickly evaluate should you buy something, trying to like streamline the document process between counterparties and buying real estate. Like that should completely remove real estate brokers from the economy. So you take that out. So maybe then do you have increased activity within the overall real estate sector? Maybe that starts to help boost GDP on the aggregate demand side. But then if people don't have money because they're no longer any real estate brokers, then they're not buying real estate. So I think, yeah, I would love to learn more about GDP, Val. I think I'm going to spend some time on that one.

21:39You know, it's a great question. and it's a reminder that there are all these secondary effects from doing things like replacing a job. It's like when you replace a job, you're not just like giving a company, you know, a certain amount of money and cost savings. There are ripple effects all throughout the economy. And in some way, sometimes that can be rough. Like if the company goes and fires that person and their spending is gone from the economy. But then if the company puts them on a higher value task and they're able to perform well, then they're actually adding to the economy. So the measurement of this stuff is just so, so difficult.

22:18I'll tell you one story. I once had a pricing conversation with a tech CEO and went to an analyst to sort of discuss this one thing that they were thinking about doing in their company. And I came back with some of the calculations and said, hey, this is kind of how it would impact thing. And they're like, oh, I didn't really think about the secondary effect. I mean, I think that's a good encapsulation of a lot of what we see in how this works. Yeah. All right. You know, we've been talking about how big the spend is and how exposed the economy is to the AI spend and what it's going to take, importantly, to return on all these billions and tens of billions and, I don't know, maybe hundreds of billions of dollars that has been invested.

23:01Great story in the Wall Street Journal this week, asking just that question. The headline is, Spending on AI is at epic levels. Will it ever pay off? Written by a friend of the show and multiple times former guest, Elliot Brown. Elliot gives some interesting comparisons here. He says, Telecom companies spent over$100 billion blanketing the country with fiber optic cables on the belief that the internet's growth would be so expensive. Most investment was justified. The result was a massive overbuilding. that made telecom the hardest hit sector in the dot-com bust. Industry giants toppled like dominoes, including Global Crossing, WorldCom, and 360 Network.

23:43So interesting. People always think about, oh, it was Pets.com that got hit the hardest, or the VCs. It's actually like the people building the infrastructure. They gave some really interesting information about what it would take to recoup the investment in AI that we're seeing today. Dave Kahn, a partner at venture capital firm Sequoia, estimates that the money invested in AI infrastructure in 2023 and 2024 alone requires consumers and companies to buy roughly$800 billion in AI products over the life of these chips and data centers to produce a good investment return. This week, consultants Bain & Co.

24:19estimated the wave of AI infrastructure spending will require$2 trillion in annual AI revenue by 2030. By comparison, that is more than the combined 2024 revenues of Amazon, Apple, Alphabet, Microsoft, Meta, and NVIDIA, and more than five times the size of the entire global subscription software market. You're not getting nervous by these numbers? I'm getting nervous reading these numbers. I mean, what had me nervous about this, and I basically entered the workforce as an intern at an internet startup, actually, in 2000 at the almost height of the dot-com bubble. And it's those names, Global Crossing and WorldCom.

25:03Because for reference, and I'm trying to remember the exact story, but basically capacity providers that were there was as the bubble started to be unwound, they were the ones that got caught. And then it was accounting fraud. Or actually, I believe Global Crossing also – actually, yeah, here it is. There's one controversial scheme that drew attention was a proposed revenue inflating swap with Enron where each company would pay the other for capacity they didn't need, thereby boosting both companies' reported revenues without actual economic substance. So some of this stuff sounds a little scary or based on our earlier conversation.

25:46It rhymes a little bit, doesn't it? Yeah, yeah. And again, this is where a bit of unwind and you start to see like, God help. We've talked about this for years now. What the accounting must look like for so many of these startups where money is going in and out. Forget now we're at the$10 and$100 billion level. A year ago or two years ago, we were at the kind of like laughing about and almost shocked by$1 billion of compute. Six billion. Or six billion. It was actually compute. It wasn't actual cash. And that compute went back to Microsoft or went back to Amazon. And is it counting as revenue?

26:27Where is it counting as revenue? What's an investment? Like these questions we've been asking, everyone has kind of been asking, but has just let it go. because no one's had to actually really dig into the numbers here or actually kind of have their feet held to the fire on them. We just went bearish. I think we're both bullish on the technology, but we question the economics. And it would be one thing if the AI enterprise build-out was going so well that this number was logical. But it's not. enterprises are super slow. There's this MIT study, which has holes in it, which I think we should talk about more in terms of like the rigor of that study.

27:14But that says that 95 % of businesses are not getting profit on their AI investment. And you know what? What's been interesting is less the study, but more the reaction where it's been just a lot of people nodding their heads saying, yeah, that feels right, as opposed to this is so wrong. We are getting that ROI. We haven't heard the chorus come back in opposition to that study. And so when you're in that moment where you're like, OK, you have to do two trillion in revenue by 2030, we're about to hit New Year's 2026. And we're not seeing we're not seeing the momentum, I think, that you would need to sort of justify this enthusiasm.

27:53That's where I think our job is to pause and be like, what the hell is happening here? Well, exactly. The growth is not coming from the end user spend. The growth is coming from the infrastructure spend. It's that simple. And until we see the growth coming from the end user spend, whether that's enterprises, whether that's all of us shelling out$200 a month for ChatGPT Pro instead of$20, until that happens, none of it makes sense. So, yeah, the MIT study, I definitely had issues with that, again, in terms of it was focused on, it's like looking at pilot programs. So within an enterprise, if individuals are getting value with individual tools, which every single person I know in the world, not in the world, in my social circles, I guess, is now, I don't know, have you noticed this?

28:49now every like call recap you get is like 70 bullet points in a perfectly structured way. Like, like sometimes it's crazy. The pitch emails are just like, you just read my website and just put it in chat CPT. Yeah. And it's sometimes like, it's like, just what can you tell me face to face? What are the key things that you actually listed out in that, in that email? So, but again, That is ROI. People are getting value from AI in these tools, but that was just more on actually explicit pilot programs that would be potentially cross-functional that would target specific use cases. that's where, and I think it's true, the way a lot was rolled out in 2023, 2024 was not the right way.