OpenAI’s Raising At $100B Valuation, Telegram’s Pavel Durov Arrested, NVIDIA's Just Okay Earnings

0:00OpenAI is fundraising at a$100 billion valuation. Telegram CEO Pavel Durev has been arrested and charged in France. And NVIDIA earnings have come in. They're impressive, but are they impressive enough to sustain the AI boom? All that and more coming up on a special Big Technology Podcast Friday edition right after this.

0:23You're used to hearing my voice on the world bringing you interviews from around the globe. And you hear me reporting environment and climate news. I'm Carolyn Buehler. And I'm Marco Werman. we're now with you hosting The World Together. More global journalism with a fresh new sound. Listen to The World on your local public radio station and wherever you find your podcasts.

0:50Welcome to Big Technology Podcast Friday edition where we break down the news in our traditional cool-headed and nuanced format. We have a big week of news, including massive open AI fundraising news. We said it was coming and it is on the way, along with NVIDIA earnings, which we'll get to at the end of the show. And very big news in France, where the CEO of Telegram, Pavel Dourav, has been arrested in connection with facilitating or having some terrible stuff happen on the platform, including child exploitative materials. So it's a hot button issue in Silicon Valley. And we'll discuss whether it was merited or not.

1:27Joining me as always on Fridays is Ranjan Roy of Margins. Ranjan, great to see you. Bright and early. Welcome to the show. Last weekend of summer. That's right. We're up bright and early. We're making a road trip edition. So if you're on the way to somewhere, on the way back, or you're just sitting at home and saying, where have all my friends gone to? This is the right show for you. Put it on. Play it in full. One company that's going to have a very busy Labor Day or busy summer is OpenAI. Now, we've been saying on the show for a while that some of the things they've been doing, like SurgeGPT, have sort of been winking at the fact that they might be fundraising and they might be struggling to raise money.

2:05Well, this week we actually got the news that that fundraising is indeed underway. It looks like it's going to be about$100 billion in valuation. I've seen numbers up to 125. But what I find interesting here is a few things. First of all, it's taken them this long, right? Right around the time when Sam Altman was ousted. This was also the number they were talking about. There was also discussion that they were fundraising. So they've waited up until now to actually like have it leak out, which is quite interesting. And then it's also not your traditional fundraise because OpenAI remains a nonprofit.

2:40And if you invest in OpenAI, you're really just entitled to a percentage of the profits they make, not a percentage of the company. It's very untraditional. So Ranjan, I want to turn it over to you. I'm curious what you make of this, the timing, the number, and whether it's actually a good investment to put money into OpenAI. I mean, this funding round has it all. It has Apple potentially investing, NVIDIA looking at it. The numbers are just astounding at$100 billion for a private market round. And as you called it, Search GPT, they started positioning themselves for this another gigantic funding round.

3:18And also, this is at a time where we saw articles from Ed Zitron and others starting to question, could this ever be a profitable company? You know, what is that actual line to profitability? So I think this is a very, very important funding ground. And as you said, not from a purely financial perspective, but more from a big technology strategic perspective. If Apple is actually involved and Apple is now building a tighter connection with OpenAI, I think this could be very interesting. And obviously, when we see, can Siri actually work in the upcoming iOS 18 results? And, you know, I think this is going to be the relationship between OpenAI and other technology firms.

4:06This is potentially going to outline what it looks like for the next few years. And so you mentioned it, that we do have Apple involved in this round, potentially. We also have NVIDIA potentially involved in this round, and of course, Microsoft. So there's like four big players here. There's Apple, NVIDIA, Microsoft, and Thrive, which is Josh Kushner's venture capital firm. And Thrive is expected to lead this round. Do you read anything into the fact that Microsoft has sort of gone from the big funder to one that might take not necessarily a backseat, but at least the co-pilot seat on this one to use one of their favorite phrases.

4:48Cheeky there, Alex. I think, well, first of all, I do thrive. Leading the round was very interesting to me because, you know, I haven't seen them making headlines necessarily in the past few years. And only from this announcement did I go back and look that they went from, it was only a billion under management a few years ago to now 14 billion under management that they just raised a gigantic$5 billion fund last year. So clearly they're ready to deploy it. And seeing the Kushners back in the news in some kind of capacity is always an interesting thing. But yeah, I think this is going to be the Apple connection is the most important one because it's clear that Microsoft is no longer the investor and partner of an open AI, but still does have more specific economic interest than any other company will ever be able to have with them.

5:43And all that was outlined in that early$10 billion round back in January 2023. So this company, it just becomes messier and more complicated and more unclear of what is the financial structure? This isn't even a traditional cap table. who is investing and why. I don't know. I think, yeah, it's less clear with all of the news that came out than I think even before the headlines started to leak. And here's my hot take on Microsoft's position in the funding, and I know nothing about this. I expect Microsoft to be, of those four, the company that puts the least amount of money in, if all four go in.

6:22I think this is something that Microsoft understands that it needs to continue to invest in, to show adjacency to open AI, but is really working hard to build its own AI ecosystem and research house and product organization. And it sees that a lot of this has become commoditized and therefore it's not going to continue to put the amount of money that it did into open AI as it has in the past. Now there's one other interesting thing that's happened here, which is that Thrive, right? So Thrive Capital, they led the sort of, well, it's not really a funding, it's the buyout of employee shares at an$86 billion valuation last year.

7:02And now they're going to be, you know, they're said to be leading this one. And there's an item in the information that talks about how, this is from Corey Weinberg. These inside rounds usually signal that smart investors are not banging down companies' doors to invest at that price. There are exceptions, of course. Thrive's doubling or tripling down on OpenAI has been raising eyebrows among competitors. The prospect of OpenAI raising billions or trillions more directly from sovereign wealth funds in the Middle East may raise eyebrows even higher with an important constituency of the U.S. government.

7:37So basically what Corey's saying is like, if you have something like a Thrive going in, basically against the grain, buying shares, and then leading the next round, the indication is that the rest of Silicon Valley isn't so eager to put as much money into OpenAI as perhaps OpenAI is seeking. Is this a red flag here? I think it's definitely a red flag, especially the idea of at a certain point, where do you get more money? When you get into numbers this astronomical, there's still only a somewhat finite capital base that could actually, you know, feed this continued need. And as he said, Middle Eastern sovereign wealth funds are one of those commonly associated funding sources.

8:22And with something sensitive like generative AI, that could be a problem. So I think, but the question around, you know, again, buying out employee shares rather than investing more in a company that is losing$5 billion is insane to me. Like rather than just continued investment into products that can grow and will grow and will kind of be that source of revenue that will actually justify these insane valuations. So yeah, I think it definitely should raise eyebrows. And where else do you go from here? I mean, again,$100 billion valuation as a private company. I think there's ByteDance is somewhere around there or was around there.

9:06There's only a very, Stripe is one of the only other ones that, and these are, and ByteDance and Stripe, again, are companies with very, very mature revenue products and they're mature businesses, whereas OpenAI is still this crazy bet that does not have any clear path to actually becoming a great business. So yeah, I think it should raise red flags. Right. So here's a question for you then. Is this a good investment? Like, all right, you could say that Microsoft made a great investment in OpenAI back in the day when OpenAI was pioneering all these new techniques and release chat GPT started the generative AI wave.

9:44I think we can agree, great money spent by Microsoft. But then you look at some of the numbers, and you mentioned one of them, that are coming out around this funding round. So first of all, the information has that opening, I could lose up to$5 billion this year. This is based off of an information analysis. We also know that ChatGPT, they say the ChatGPT usage has doubled. So now this is the first official new number that we've gotten. ChatGPT is doing 200 million weekly active users. But we also know that OpenAI spent$100 million. Actually, this might be good numbers. OpenAI spent more than$100 million to build GPT-4.

10:25And this year, OpenAI said its revenue was$3.4 billion on an annualized basis. So put that all together. I mean, it does seem like... Wait, hold on a second. How are they losing$5 billion if GPT-4 only cost$100 million to build? Do you think that's all inference costs? Like, where are they losing this money? Or is it just training the next model? Yeah, I mean, combination of the actual ongoing compute for every single ChatGPT query, especially for unpaid members of people, like non-subscribers using ChatGPT, training the next model. I mean, I don't think that's an unreasonable thing, again, around just the overall compute required to operate a service like this.

11:12I mean, every dolly image you're creating, every API call. So I think that's definitely not unreasonable. It is interesting to me, though, that, I mean, in terms of the way these funding announcements leak, I think is worth bringing up. because one thing that happened in the last few years, especially during COVID, especially with these kind of high octane funding rounds is this idea that the leak happens during the fundraising itself in order to set that target valuation. That once the leak happens and they say$100 billion in everyone's mind, it's$100 billion. Whereas in the past, the news was the funding is done and here's what it was done at.

12:00And a lot of the times people protected the valuation of that fundraising round with all of their heart. Like every banker involved, every VC involved would never give up the actual valuation. So I think it's clear that this game of trying to create signaling around where these numbers are in order to drive interest at those numbers is still happening and I think is never a good sign. Right. And also, one of the things that I've heard about OpenAI is that they're just not trying to sell anything at a profit now. So basically, they're trying to show that the engagement numbers are going up, that they're entrenched as far as every AI business.

12:38And if you're building, they want you building an OpenAI and they're not really putting profit at the forefront. But again, I think the main issue here is that you're only entitled to a certain percentage of OpenAI's profits. You don't even own part of the company. so like let's say they convert and they become you know some sort of traditional structure and you've invested in this non-profit and then they IPO you don't even have a stake in that like that to me is like why would you ever want to invest in that unless you're trying to make this strategic bridge like Apple obviously is like Microsoft has in the past maybe Nvidia has some interest there but the Thrive putting all this money in it and leading the round I think is really, it's an interesting business play.

13:23I don't see how it aligns with the traditional AI, sorry, the traditional VC investing, unless what it's trying to do is say, hey, we have open AI in our portfolio company and it believes that AI is going to be big enough, that that will be sort of like this flagship investment that will attract others. But it's still a lot of money to pay for marketing. Well, no, you're right. This is the weirdest company in existence. The more we're digging into this, and I'm actually surprised that most of the reporting treats it like a traditional funding round. But remember, this is a company where the fundraising in the past, a lot of was in cloud credits.

14:01And then now even you see a lot of it is strategic considerations for an Apple. I mean, Apple can't give them cloud credits. Maybe they'll give them, I don't know, free iPhones or MacBooks or something else. I mean, at every level, this is a company where nothing is normal, nothing is standard in terms of its structure. And we still know so little about exactly what are the terms of these different deals, what are the terms of whose financial interest lies where. Because, again, even that structure of, what is it? It's like up until a certain percentage of profits that it still gets capped at a certain point.

14:42Yeah, exactly. We know that from, I think, two years ago. So from any subsequent funding round, were there different stipulations? Were there some cloud credits? How much is cash? Yeah, there remains a lot unknown about OpenAI and its business structure. Do you think meta is a threat to ChatGPT? I mean, we talked about in the past about how Mark Zuckerberg saw ChatGPT take off, massive consumer product. He needs to build his own. Of course, he's built his own. And now they've released this Meta AI, which is built on top of their LLAMA model. And this week they put out some numbers with the information.

15:19They say it has at least 400 million monthly active users, which is double OpenAI's ChatGPT weekly active, which again is comparing apples to oranges. But there's just the comparison. And 40 million daily active users. this is according to two current employees but then meta had an executive comment so interesting media strategy there but um 40 million daily active users of meta ai which is uh baked into like almost all of meta's messaging products is that good i would tend to think it's not i think that's actually terrible and it's clear that that number lee because i actually i got a i don't know do you do you follow mark zuckerberg on instagram yes because yeah he gets i get i got a broadcast channel message i don't know if you know the broadcast channels is something meta is really trying to push the idea that you can have like one-to-many message push from uh for influencers and creators and zuckerberg is one of the more active ones but basically he was saying he sent a message saying that llama is growing even faster than i expected almost 350 million downloads and a 10x jump in monthly usage since the start of the year.

16:32So he's pushing this message. And again, we need to separate out Llama from Meta AI. And Meta AI, the only time I've used it is when I've accidentally ended up in it with an Instagram search. And it returns results that are worse because I want to actually search on Instagram. And instead, it gives me a more half-baked generative AI response. So I'm guessing that the meta AI usage. And you can only imagine on Facebook and the average Facebook user ending up in some weird rabbit hole of generative AI accidentally. So I think the meta AI side is probably being oversold a little bit. It's still certainly very well positioned when you can start properly banking it into consumer products.

17:23But I think Llama is the bigger threat to open AI. And especially, I mean, in the context of this giant funding, potential funding round, LAMA is such a brilliant strategic move, and we've talked about it, to go open source and say, basically, we're not going to charge you for every API call. Here's the model hosted on your own compute and servers. I think that to the open AI business is a much, much bigger threat than meta AI is to ChatGPT as a consumer product. Yep, that totally makes sense. But it also sort of goes to show that this stuff just still has like limited appeal, I think, to consumers.

18:03I think mostly because they're just overwhelmed. Like regular people are just overwhelmed in terms of what do I do with this? Like, okay, you can sort of speak with these agents and have like interesting conversations, but people just don't know where to go. And I think in most of them, including perplexity and chat GPT, you're starting to see some of these like prompts that sort of push you towards different activities that you should try to sort of engage with within them. But I do love the one line from the information story. After it gives these new updated numbers and has like proud quotes from meta executives, it says, it's unclear how many people are using meta AI unintentionally.

18:41I'm like, well, yeah. That's a pretty important caveat, wouldn't you say? As we said, I mean, and that's the only way I've used it, but I think we got it. We have to distinguish perplexity from meta AI because perplexity is going after search and search is a function that we all know how to use and use in our daily lives all the time versus the meta AI products, again, have been jammed into where your people regularly do search and search on social platforms has become huge. Like TikTok, there's been endless coverage that it could be cutting into Google search itself and it's become the search engine for Gen Z.

19:21So I actually was more surprised because, again, when someone searches on Instagram, especially when they're searching more for reels and other content like that, they want an output of a reel or a photo. They don't want a generative AI answer. So it's almost the most backwards way to try to release this kind of product to your consumer base. And I think it's actually bad for the users. Yeah, it's really weird. Do you think that meta AI and threads have kind of a similar feel to them, which is that you sort of get it having seen it work elsewhere, but it still just doesn't really work in the setting that they exist in?

20:01Well, no, the thing I wonder, and I think Zuckerberg is peaking right now, and certainly like the days of just trying to, you know, everyone going after him are a few years past. But I do wonder these kind of products at a company like that, do they believe they're successful? Are they still sitting there like high fiving each other saying Threads is a massive success? because these numbers, 400 million, whatever threads is it now, 100, or I think they said 200 million monthly active, when you just jam it into a product, the three products that 3.2 billion people use every month, you can get those numbers easily, even with a bad product that is not actually working and will not grow.

20:49Even again, on Instagram, every fifth scroll, I get a row of threads that is trying to encourage me to link out to, which if I didn't know what it was, I could very accidentally click on those. So I do think that, yeah, I think it feels similar to threads in that it's very easy to hit those numbers that seem really big without a product that is necessarily doing great. So I wonder how realistically they actually do evaluate these internally oh i think i'm sure they view it with like real science and they have like so many metrics that we've never heard of before looking at like the momentum the acceleration the retention all these things but i do think that they also like everybody in silicon valley know that you need to hype something a little bit before it's real to make it real because this is just like network effects on both of these with threads it's like you need people to be posting and then interesting stuff circulating in the timeline and with meta ai you need people to have good experiences with it so that it spreads with word of mouth and like you're like calling meta ai into a group chat which you can do no no i just i disagree i think it's actually the opposite because i think meta ai does not need network effects it just needs good output because again the word of mouth this was cool this was interesting is almost not needed it was needed for ChatGPT, but for meta, just make it good, make it actually make sense in the product.

22:20And when you have that committed user base, that's all you need to do. So I think actually versus threads, yeah, that pure network effect type product. But I think, again, meta AI has become only more confusing for me as it's been developed. But the llama strategy, I think, is vicious in terms of the whole landscape, the way Zuckerberg and the company have unleashed that on the whole ecosystem. Right. So look, I think we could both agree that there's still a lot of time before we can both believe that AI for consumers is going to work. But it does seem like every week there's a new headline about how it's making inroads in enterprise.

23:04And one of the stories that you picked out was how within Amazon, there's a belief now that you're really not going to need to code anymore pretty soon. So you want to talk a little bit more about that? Yeah. So Amazon's website, the AWS CEO, Matt Garman, there's a recording that I believe was leaked in terms of like an internal chat. But he said, if you go forward 24 months from now or some amount of time, I can't exactly predict where it is. It's possible that most developers are not coding. Coding is just kind of the language that we talk to computers, and it's not necessarily the skill in and of itself.

23:42I think that is exactly correct. A couple of months ago, we talked about my cloud coding journey, where I literally went and built a React app and deployed it to GitHub pages. These are things that I don't do, and I don't know how to do, but just talking and in conversation using natural language was able to build a small app experience for the first time in my life. And that is exactly it, is that coding was always just a way to translate communication to computers. So now if you can actually just speak in natural language to computers, the actual coding side of it, you don't need to be able to do.

24:24And I don't think this is a, it's interesting because there's a few different ways people approach this. It's either coding is now, like developers are going to lose their jobs, or then you have the opposite, more optimistic take, Satya Nadella saying, now we're going to have 1 billion developers. And I actually think that's the more correct take. The idea that building software to actually solve problems, there's going to be, this is going to now be accessible to so many more people. But that role of a traditional coder that just creates code is going to significantly change. Did you use the artifacts function on Claude to build the app that you're talking about?

25:08Yeah, exactly. I mean, it's insane. Like you literally, I'm like, wait, actually, you're doing this wrong. Can you just change this? And then it shows the code being generated. And then it shows the output in the window to the right screen. It's a split screen i mean it was incredible to be able to see it see the code being generated seeing the app being built just by writing sentences as though you're talking to a developer friend who's doing this for you yeah it's interesting it's like we got the iphone because we had the blackberry first right like people i mean maybe not a direct causality thing but like you get what i'm saying Like smartphones, the first use case was enterprise, then we went to consumer.

25:53And maybe we're going to have a similar thing with AI, where it's just going to take a little bit longer to figure out the perfect consumer use case first. Because with business stuff and with enterprise stuff, you can find a lot of broken things to fix before the obvious example comes through in consumer. I still I disagree with that a bit because I think businesses, the hype is there, but the actual opera like getting things operational is still incredibly difficult. Whereas it's funny, everyone I know now is using ChatGPT or some other equivalent and like not the most like like hardcore tech forward friends.

26:35It's getting more and more incorporated just into people's daily lives, especially non-native English speakers for any kind of writing output are using it, everyone I know. So it's become, I do think it's become incredibly commonplace. And it's just not one of those things that it's almost that people are embarrassed when their output is actually from a chatbot. But I think, I don't know, I've seen it much, much more ingrained into just everyday, I don't want even to say workflows, because it's just life. But in terms of people just doing stuff day to day. So I think the consumer adoption and use case side of it, aside from meta AI jamming it into Instagram, is actually there.

27:21But don't you think that you're then talking about a skewed sample because we only have 200 million weekly active users of ChatGPT? Or is that number bigger than I'm giving you credit for? I thought you were going to be talking about weekly listeners to the Big Technology Podcast. Well, we know that we're mainstream for sure. Of course. Well, no, that's the thing. But yes, it's skewed, I'm sure, in my network a bit towards people who are in that 200 million monthly active user or weekly active user cohort. I recognize this. I'm sure for big technology podcast listeners, everyone is probably skewing into that cohort as well.

28:02But it's one of those things that it's gone far past only early adopters. We're way, way past the early adopter phase. And actually, I think one of the most interesting things to me is in the latest Chrome update, you can now search from you can now interact with Gemini directly from the input bar. And I remember Chrome, one of the most powerful things Google did, and which probably is relevant in the whole antitrust conversation, is by pushing Chrome to the whole world and then creating the behavior that you no longer had to go to google.com, but you could just type your search query directly into the input bar instead of just only typing in a web address in there, I think was one of the most powerful ways to get people searching even more.