DeepSeek’s Potential $10B+ Valuation, LinkedIn’s AI Agent Win and the Risks of Anthropic’s Mythos

0:13Welcome, everyone, to The Information's TI TV. My name is Akash Pasricha. It is Friday, April 17th. If you're walking around New York City this weekend, make sure to check out our new billboard up in Times Square thanks to our sponsorship with the New York Stock Exchange. I'm going to take a photo with it this weekend. If you take one as well, send it to me. I'm at akashattheinformation.com. Look forward to seeing it. I want to kick things off with a scoop that my colleagues Anissa Gardizi and Valida Pau published on Thursday. OpenAI has agreed to spend more than$20 billion to use chips from the chip company Cerebris over the next three years.

0:51That figure is double what was previously reported about the deal. In return, OpenAI is getting warrants in Cerebrus with the potential to own up to 10 % as its usage scales. OpenAI is also putting up about$1 billion to help finance the data centers that will run those chips. We've got that story up on our website. I encourage you to check it out. On today's show, we are unpacking the information's latest scoop from our Asia Bureau about DeepSeek. We'll then discuss Netflix's quarterly results. We'll also dig into LinkedIn's booming AI product, which has become a bit of a bright light for parent company Microsoft.

1:29And we'll close out the show with our weekly edition of The Editor's Cut. Today, we're talking about the potential risks that Anthropics' mythos may pose to financial systems and to society. It's going to be a fun show, so let's get right on into it. The Information's Asia Bureau, Channer Liu, Zheng Yang, and Juro Osawa, published a big scoop about DeepSeek. I'm joined now by Juro to share more with us about what he learned. What have we learned, Juro, about DeepSeek? Hi, Kash. So we just published an exclusive story, and DeepSeek is in talks to raise money for the first time. And, you know, DeepSeek has actually never raised outside capital before, so this will be a very big shift for the company.

2:15And we heard it's talking to multiple investors, and the talks are about raising at least$300 million and the potential valuation more than$10 billion. It could be significantly higher than that as well. So yes, those are what we reported just now. So why hasn't DeepSeek raised money in the past? So it hasn't raised money in the past because it didn't have to. And so DeepSeek's parent company is a very prominent Chinese hedge fund, quantitative fund. And the money from that fund has been able to provide the capital that DeepSeek needed. And also the DeepSeek founder has been, his preference has been to really operate without the outside money, to really focus on the research.

3:13And this company was really built around a culture that's much more like an academic lab than an enterprise. So, you know, over the years that has worked really well. And, you know, but so this is like that's why, you know, this really comes as a pretty big shift for the company. And so what has changed? Have DeepSeek's models gotten less competitive? Does it need the capital to invest in compute and hiring researchers? I mean, what's going on? So, yes, multiple factors, I think. So one is definitely, you know, computing cost to keep competing at the highest level and really, you know, try to develop, keep developing the frontier models.

4:03So the cost of, you know, doing that is one thing. And then also, DeepSeek has had some departures of key researchers. And quite recently as well, there's one researcher who left and went to ByteDance, the other Chinese competitor. And in those cases, the competitors are paying a lot more money to hire those people. So raising more money, outside money, can also help DeepSeek retain talent or just hold on to that top research talent. And remind us, what is the latest model that DeepSeek has on the market right now? How good is it? Just remind us there. So, yes, DeepSeek, obviously the big hit that they had was more than a year ago when they released R1, the reasoning model that really changed the world.

5:04And since then, they have had some updates. So they've released several updates to another model called V3. And then the next generation model is we've reported previously, but they have been working on it. And it's, you know, they're preparing to release it. So it should be coming soon. But that really will be the, you know, completely new generation model. So basically, over the past year, DeepSeek has some updates and also has published some influential papers. but the real complete update and the new generation model hasn't really come. And obviously, a lot of people have been waiting for this moment.

5:59And so we're waiting for V4. V4 is what we're waiting for. Yes. Okay. You know, one question I had for you, Juro, is as this company seeks to raise this capital, I mean, I think we all know the story of it. It was a hedge fund before. And I mean, I guess if you, I was wondering if you have any sense for like how the structure of the company has changed or could change. I mean, are they still making most of their money from the hedge fund side of the business? Is that still very much alive and running? Or has the shape of the company changed entirely to now become maybe closer to what an open AI is or an anthropic is?

6:41so um uh i think the uh the shape the company hasn't really changed but uh i think the the environment has changed a lot over the past year and uh especially the competitive environment so uh when deep seek had the big hit r1 uh at the beginning of you know 2025 the market looked very different and you know they really became this like a new open source model that was really competitive but then since then a lot of other companies especially other chinese companies uh really came up with new you know open source models of their own that you know also gradually became more competitive as well so uh you know the uh the market now has a lot more kind of options and uh so you know that is something that you know deep seek itself as a as a company and a team may not have changed fundamentally, but the environment that it's operating in has really changed.

7:46And last question for you, Jiro. I imagine DeepSeek is a pretty hot commodity for investors given all of the traction it's showed it's been able to get, at least at certain points. Do we have any sense for how competitive this process is in terms of venture capitalists getting into this round? Or even do we know who is bidding to get into the round? I think a lot of the details will be, we will try to find out more for the next story. But we have heard that they are talking to multiple investors and obviously there is a lot of interest. And China has had some very successful IPOs of AI model companies in Hong Kong, over the past, you know, earlier this year, especially like Minimax, Jupu, those are the two model companies that already went public.

8:46So for investors, you know, if there's any chance to get into this market, especially, you know, a very prominent player like DeepSeek. So if there is a chance for outside investors to get in, you know, there will be a lot of interest. Great. Well, Jira, I want to thank you for coming on. That is Jira Osawa, our Asia reporter. here at The Information. Netflix reported 16 % revenue growth in its first quarter, slightly above projections, but the company is projecting a slightly slower full-year growth this year compared to last year. Shares are falling in the aftermath. I want to break down the results with Barton Crockett, Managing Director and Senior Research Analyst at Rosenblatt Securities.

9:29Barton, welcome to the show. It's great to have you here. Great, thanks. Really happy to be here. Love the information and really excited to be on the show with you today. What stood out to you from the quarterly results yesterday? Well, look, I mean, we have a neutral rating on Netflix. And, you know, I think that this report was a report that it shouldn't prompt anyone to dramatically rethink their view of the company. A little bit of upside in the first quarter, a little bit of weakness versus expectations for the second quarter, maintaining a full year guide. Right. The setup with Netflix, we think, is this is a great company that has a stock that already reflects that.

10:08I mean, the company is trading at about 24 times enterprise value, the EBITDA. We think they're growing their EBITDA at about a 24 % CAGR last year to next year with margin expansion on mid-teens revenue growth. And that's not really a crazy price for the stock to be trading at. So look, if it got a lot cheaper, again, let's say they decided to make a run at some other media conglomerate and then back off, look for entry points. But I don't think there's any reason that you need to rush into it today right here. So I want to talk to you about the valuation, about the multiple, because the quarterly results showed that growth seems to be slowing gradually.

10:50We're not sure what's going to happen past this year. But, I mean, the company, if you look at the multiple, it's trading well above what other slow-growing TV companies, you know, older media companies are trading at. And it sort of prompted the question, in our newsroom at least, you know, if Netflix starts to slow and it starts to look a little bit closer to what an older TV media company looks like, I mean, shouldn't the multiple follow suit? What do you think? Well, it would. I mean, to be clear, they're going a heck of a lot faster than traditional media companies are right now. I mean, there's Disney, you know, whatever the Paramount Skydance combination, it looks like.

11:33They're not growing EBITDA at 24 % CAGR, nothing close to that. You know, I think they're happy with, you know, maybe low, high single digit, low double digit growth in the case at Disney, mostly driven by theme parks and Paramount, Warner Brothers, whatever that's going to be called. I mean, who knows what happens there. But so they're not there yet. And so the question is, is that where they're going? And look, if you continue to see the revenue growth decelerate tick after tick after tick, yeah. And, you know, and that's a little bit of our concern is that this is a company that looks like it might be on kind of a maturing arc.

12:09And what we've seen is, you know, maturing media companies, perhaps with a cloud of some secular threat around them, you know, can be stocks where you play kind of valuation limbo, which is how low can you go? And, you know, for a decade plus, people were, you know, really just besides themselves with how cheap Paramount was and how cheap some of the other media companies were. But then they ended up, you know, being right. The market was predicting kind of secular demise. Netflix is making the argument that that's, you know, nothing could be further from the truth for them. They talk about a$670 billion total addressable market between what people pay for subscription television and streaming and what you see in terms of advertising on those platforms and spending on video games by the consumer outside of Russia and China, and that they're only thinly penetrated at 7 % and they can grow that penetration.

13:05And I think that they're right to look at the addressable market. I think they're right to think that they can play in those spaces over time. But, you know, mid-teens to low double-digit growth does not argue that you're really, you know, taking your total addressable market by storm. So, you know, there's a little bit of concern around that. And, you know, at this point, we're sticking with the neutral because, you know, we think it's slightly better than, you know, it's much better than the media companies. And, you know, for now, we're going to sit here okay so so now you talked about the total addressable market let's talk about the the the advertising market specifically i mean how big a business do you think advertising could be for netflix not just given the dynamics of advertising altogether but all the competition that netflix sees from other streaming platforms youtube you know and and you've certainly got newer technologies that they could spend on too?

14:05Talk about it. Yeah. So, I mean, they're doubling their ad revenue this year to$3 billion off a relatively kind of modest portion of our overall sales, 6 % this year. They talk about an ambition of 10%, which certainly they're on a path to kind of get there within a year or two. And then we'll see after that. Once you get above 10%, thing you start to really move the needle on total revenue growth. Look, they've got a billion audience if you count not just subscribers, but those co-viewing. And the opportunity to monetize a billion viewers, watching your service for nearly two hours a day or so is really great.

14:54But this stuff doesn't happen instantly. I mean, you have to invest in the tools, you have to invest in the relationships you have to invest um in um time basically in building a relationship with marketers um in this type of area so yeah that would be kind of the upside argument that would you know separate them from the rest of traditional media that doesn't have this type of scale um but look there you know the ad market is um going through a lot of flux i mean open ai is talking about going to you know 100 billion dollars i think you guys were reporting in uh 2.4 this year 2.4 this year yeah but 100 billion by like 2030 is like right guys you know that was the reporting 2.4 billion this year off of you know basically nothing last year and um you know those goals sound kind of like a stretch i mean i'm not really banking on open ai kind of hitting that stuff but uh but if they're successful i mean there could be a sucking sound of advertising from other categories that could be a headwind for many players maybe including netflix So just to go back to it, what is your projection for how big the ad business could be for Netflix two to three years out?

16:02Have you put a number on it? Yeah, we've got numbers. I mean, I have to check in my model to get into the specific numbers. But, you know, leaving that aside, the direction is really, you know, doubling to, you know, high. It's a curve of kind of general kind of deceleration that we're modeling. And, you know, law of large numbers gets bigger, you know, bigger base, slower, but, you know, crosses 10 % of sales within a couple of years and continues to grow from there. So, you know, once you're above 10 % of sales, that moves the needle and maybe you can see revenue growth re-accelerate. And that might be kind of the longer term story.

16:37We're a couple of years away from that really being visible in the consolidated numbers. Now, very quickly, you talked about the user base. Netflix insists that pricing changes won't have any impact on membership numbers. Do you believe that? Yeah, I mean, we kind of have to believe it because they don't give us any numbers. So, you know, I have what they tell us, but I don't have the numbers. They're not reporting that anymore. In the past, they've had a lot of pricing leverage, generally. I mean, there have been quarters when they've, you know, early on when they were, you know, whipsawed by price changes, but they've learned how to navigate that much, much more subtly over time.

17:20And the more recent trend when they were giving us numbers is that they could raise prices and grow subs pretty successfully, that they had pretty good pricing leverage. and so if they're telling us that the US price hike of you know 10 % plus effective that people are absorbing that in an environment where everything's going up in price and it's not going up as much as your gas bill you know maybe I can believe that right because I mean you know the I guess the candid concern you could say investors might have is okay revenue growth is decelerating prices are coming up to what extent is there revenue growth coming from the price increase as opposed to getting more people on the platform once you get into that game, how high can you raise the price?

18:04I mean, you know, there is one way of thinking about it that, hey, this is sort of, it's a tough narrative to get around. On the other hand, like you said, there is leverage. Yeah, there's, you know, there's pricing leverage. Clearly, you know, the long-term exit from that doom loop you just laid out is advertising. You know, if they're able to grow advertising at an elevated pace for a long time, bam, there's your answer. Right. Let me ask you one more question here, looking at the newer bets that Netflix is making with sports, with gaming, and I guess video podcast is the most recent one. So maybe let's start with gaming because that's one they've leaned more into.

18:50If you were to give a letter grade uh to each of these pursuits and there's a number of ways we could rank them but i'm sort of asking you based on like execution and yeah how well you think their netflix is doing you know in each of these categories so let's start with gaming what letter grade would you give in there you know i give them a um a b um you know i think that gaming is just a very long-term play. It's, I think, meaningfully a retention play. You know, they launched this new kind of playground kind of setup for targeted at kids on their iPads and iPhones. And I, you know, that makes a ton of sense.

19:34And I think is, you know, one other reason that parents are not going to cancel Netflix. And because that's your babysitter right there. But the, you know, I think that they have a CFO who comes from the gaming industry. And I think there's a lot of kind of hopes that over time, this can be a content form factor that is a really kind of big part of consumption right now. I think it's kind of on the periphery. But the nice thing is, if you get a hit in games, that can be really durable, right? And if they're able to get their hit, whatever that is, maybe leaning off playing off of you know their hit content games around that maybe something else that they come up with um you know there's real option value there and it could you know quickly go to an a plus if uh um you know they find that hit okay i think that's a great that's a great point by the way what about what about sports sports i give them an a plus i mean they've just broken all the rules and sports and one um which is amazing i mean they you know they went up to the NFL and who tells the NFL anything who pushes the NFL around Netflix you know they they said to the NFL we don't want to have all your regular season games we just want big spectacles and carve that out of all of the oodles of cash that your media partners are paying you and give them to us and Netflix said you know sure dad so yeah I mean it was it's incredible what they've done with that, what they did with, you know, these spectacle events like Mike Tyson, which, you know, was hard to watch, but a lot of people watched it, set records.

21:18And, you know, they're doing this differently. Nobody makes money in sports except Netflix, because they're just taking what works and leaving aside, you know, the weight for others. And, you know, I applaud them. It's really amazing what they've been able to do. Okay. And last, well, how about the video podcast play? the video podcast play is still you know i think it's too early to give them a grade um and uh you know i'd say a midterm you know and i'd say the midterm grade we're kind of you know early into it is um um an a minus maybe um you know i think that they are uh uh you know they're talking about 10 of their user base kind of sampling it um they've uh um are using uh the video podcast to get into uh a more mobile kind of cohort they haven't really been much of a player on mobile phones, you sit in front of your TV, but on the podcast, people are listening to it when they're out and about.

22:13And, you know, that's important, right? If you think about this TAM, you want to be able to go off of the big screen that's, you know, on the couch in the mobile world where so much media consumption is happening now, and this really starts to get them there. Well, you know, I was reading a column that my former colleague who now writes a column at Scalable put out about the Netflix podcasting pursuits. And the point that she made is that, you know, less of an impact on Netflix, more of an impact on the shows themselves. I mean, some of these shows that have given the rights or access to Netflix to air their content.

22:57I mean, it's pretty staggering how many, how they've seen traffic in other channels drop. And I'm thinking of a very popular hockey podcast, Spittin' Chicklets. I mean, the stats are quite staggering insofar as how they've seen traffic just go all to Netflix. and you know it's it's so it's um it is quite staggering i think to think about the power that netflix has when they go all in on these these pushups yeah you know and look i mean there's a there's a check that covers that traffic loss and that folks will have to write that check because people will understand um and netflix understands i mean if you know if there's uh you know advantage to access on netflix maybe get the full video there and not elsewhere um that the audience will follow you there.

23:42So yeah, but you know, they can write those checks because they've got the business to support it and the content, you know, line that that's just going to be an insignificant kind of rounding error. They can move from one bucket to the other for that podcast. So podcasters, I think, you know, like the idea of certainty, right? You get that big check. You don't have to worry too much. Just kick back, enjoy it and talk on your podcast. Great. Well, Barton, I want to thank you for coming on. That is Barton Crockett from Rosenblatt Securities here on TIT. As Microsoft tries to get enterprises to adopt its core AI products, LinkedIn, which is owned by Microsoft, has itself surprisingly seen extraordinary uptake for its AI products.

24:26My colleagues Aaron Holmes and Laura Bratton wrote about that in our Applied AI newsletter. I want to bring on Aaron to share with us more about what he knows. Aaron, welcome back to the show. It's great to have you here. Happy to be here. What is going on at LinkedIn? So LinkedIn has this AI agent product that only came out a few months ago, but has kind of become a surprise hit for Microsoft when it comes to generating revenue by selling AI tools to businesses. And specifically, this is a sort of recruiter AI agent that they call hiring assistant that basically you can just tell the AI, you know, in natural language, I'm looking for a job candidate for X, Y, and Z role, and it will automatically, you know, search through LinkedIn system for you and find a small number of candidates.

25:15And it's not cheap. It costs, you know, several hundred dollars, in some cases over a thousand dollars per seat per month. But I am told that, you know, businesses have been spending pretty significantly on this tool and other departments at Microsoft have taken note of the success of this tool specifically. So look, I think the product itself, I haven't used it, at least not knowingly. Maybe it's more meant for recruiters. But I think it was pretty obvious that the data that LinkedIn is sitting on and the convenience that AI can offer, it seemed like an obvious product. Let's get to the business impact.

25:53Do we know anything about how much revenue this is generating for LinkedIn? And also, you spoke with the chief business officer. What did you glean from that conversation? Yeah, so I don't know the total revenue. I do know that it is seen as significant, and I'm told that it's possible Microsoft will even give shareholders some sense of sales during their upcoming earnings call, which is in a couple of weeks. But what I do know is that, yeah, the LinkedIn chief business officer told me that the number of customers using this tool has grown by around 30 % every single week since it launched in September.

26:30and they also are tracking the fact that people who have paid for the tool are using it more and more, which is notable because in the past, Microsoft has internally worried a little bit about how much people are paying for Copilot and then actually using the tool after they pay for it. So that retention is an important metric for them. Now, this could be a stretch, but can we glean anything from Microsoft's AI strategy from how LinkedIn is using AI? And I ask that because one of the things you pointed out is that the CEO of LinkedIn got a promotion and is now overseeing a lot more at Microsoft, right?

27:10Yeah, that's right. Ryan Roslansky, who has been the CEO of LinkedIn for several years, last year got a promotion to also serve as an executive vice president of Office 365 products. And, you know, in that role, he is now overseeing some amount of the development of 365 Copilot, which is Microsoft's sort of flagship AI product that it's trying to convince businesses to spend money on. And yeah, I mean, actually, the reason that I, you know, originally started looking into this LinkedIn tool is because I was hearing from other executives at Microsoft, as were some of my colleagues who were saying this specific AI agent from LinkedIn is seen as a success internally, and other teams working on Copilot were sort of studying what worked with this LinkedIn product and trying to take away lessons to improve Copilot.

28:02Great. Well, Aaron, I want to thank you for coming on. That is Aaron Holmes, our Microsoft reporter here at The Information. This week on The Editor's Cut, we want to talk a little bit about anthropics dangerously good model claude mythos the long-term repercussions of that technology have come up lately in discussions that our editors have been having and so i want to bring on martin pierce and meredith mazzelli to walk us through their thoughts on all of this martin and meredith welcome back to the show it's great to have you both martin i want to start with you what do you think of me what what what's been percolating in the mind of Martin Piers are on Mythos lately?

28:37Aaron Holmes:I am rather concerned that I'm going to wake up one day and my bank account will have zeros in it, not because I've spent too much, but because some hacker has used Mythos. They've gotten access to it and they have used it and they have taken the money out of everybody's accounts. And that is making me wonder, should I be converting some of the money I've got the bank into gold or should i be going to the atm and withdrawing large amounts of uh currency just in case this happens it's just weighing on me this morning i was thinking should i alert my family and friends to this do people really are we really paying attention to what could happen okay i wish I'm concerned that we're so used to the big AI doing scenarios that we're not actually taking the notice of something that could really happen.

29:39So, Martin, you are like the most skeptical person that I know. Why are you so scared about this? I feel like we're really getting ahead of us here.

29:47Aaron Holmes:Because I don't want to be left bankrupt. I don't want to wake up one day and not have any money to buy food or whatever. That just freaks me out. And because I'm skeptical, I'm skeptical of the people who are skeptical. Okay. So when I brought this up, Amir was like, yeah, don't worry about it. I'm like, man, I'm skeptical. I think this is a real concern. Yeah. I mean, look, at a high level, I do think that AI helps bad people do bad things more efficiently, just better. But I'm not really clear what the fear on mythos specifically is. I know, you know, Treasury and Fed officials met with bank CEOs to talk about it.

30:30That sounds super scary, but I mean, that's kind of the Fed and the Treasury's job is to get ahead of potential emerging risks. We're not talking about a live hack here. We're not talking about dollars coming out of bank accounts. And I think that's actually evidence that the system is working as it should. If this model even is that powerful, the banks are getting it early. The regulators are aware and in contact with the banks. And yeah, I was not really clear why we should be so scared. So Martin, is your concern, Martin, that this is going to get out there essentially beyond what Anthropic has already given?

31:08Aaron Holmes:Absolutely. I think the idea that they can make this model available to 40 different organizations, and they're going to be able to ensure that that doesn't get out to some bad person is zero. So therefore, when that bad person gets this, they could really destroy the country. And I'm just worried that we're all assuming everything will be fine. I mean, you know, the risk of ignoring this is so enormous that I don't really understand why I'm the only person who's, you know, freaking out about it. Right. But the banks aren't ignoring it and the regulators aren't ignoring it. I don't know that they're not a deal.