Live From NYSE, The Gemini Win Scenario, OpenAI Monetizing With Ads | Diet TBPN

0:00You're watching TVPN! I think we nailed that. We are live from the New York Stock Exchange. Here we are. The real fortress of finance, the capital of capital. Our second favorite place to do business. Yes, and we have some fantastic news. We have a partnership with the New York Stock Exchange. Announcing today, we have a post here from Lynn Martin, president of the New York Stock Exchange. Living legend. She says, a really bright spot for 2025, has been getting to know these guys. That's us. We're proud to announce today that the New York Stock Exchange is TVPN's exclusive exchange partner, covering the IPOs of tomorrow.

0:31We are proud to provide the backdrop for their coverage of the next wave of tech-driven innovation. With Jordy Hayes and John Coogan and the entire team at TVPN, this partnership underscores our commitment to providing the premier platform for companies that shape our future. Well said to Lynn. This partnership was probably the most natural. It made match made in heaven. Match made in heaven, truly. Not just saying that. We got together for the first time for the Figma IPO. Got to come back for the Klarna IPO. two of the more memorable moments from this year and Lynn and the whole team here are just fantastic so this will be our home when we are on the East Coast we love it here and we have a super fun show today.

1:12Should we start with Gemini? Yes. So this is the correct take he's talking about Gemini winning the AI race and and questioning is it bearish for the market as a whole if you think about it which is what efficient market hype said Gemini winning the AI race is like Super bearish for the market if you think about it. And he says, Gemini winning ensures zero profitability for any other LLM model. Google will force any other player into an endless sea of red ink by keeping its model free until they bleed out. And then it will monetize once its monopoly is secured. That means ain't no one making money on data center capex.

1:48Oops. Hot take. I think it's thought provoking. Yes. I disagree with a lot of it. Yes. It's very real in some sense that we always knew that Google would put an incredible amount of pricing pressure on OpenAI. They have the cash flow, again, even in the areas that OpenAI also wants to compete. Consumer electronics, science, I'm sure chips, obviously. So all these areas that are not even core to OpenAI's business today. Google's already been investing billions and billions and billions of dollars in these categories for a long time. I'm not convinced that there will be a monopoly in LLMs. It feels today like we're headed towards like a duopoly at the very least.

2:29And you can just easily see that there will be a number of other players making plenty of money. I feel very good about Anthropic right now, right? Mm-hmm. Anthropic, I thought Dario's commentary yesterday at Dealbook Summit. It was fantastic. He literally said the word YOLO. Oh, really? I do have an overall rebuttal. My rebuttal to Ross Hendricks here is that Google likes good margins. They grew up with the best margins. It's in their culture that they had 80 % margins. And then also, there's this constant thing when you're a public company that even if there's the new exciting thing, like there's a little bit of like the innovator's dilemma, there's the new exciting technology.

3:12But if it's not going to monetize as well on day one, then all of your investors, all the public market investors, start asking, like, is this going to structurally hurt your business? And this happened with reels, remember? There was this big question with Instagram, like, hey, we're moving from the image-based feed where it's very clear that you can just drop a link to the next thing, to reels. Is that going to monetize as well as the rest of the feed? And the answer was yes, definitely. But it took a while. And there was like some skittishness there and Meta had to do a lot of work to monetize that.

3:47And so I would be surprised if Gemini can hold out on not monetizing forever. Well, they are monetizing. That's the point. The pricing, at least from a consumer standpoint, is very similar. Both Gemini and OpenAI offers free student plans or at least a year free, but they're charging for the product. They did have an interesting announcement yesterday. yesterday, they introduced Workspace Studio, where you can build custom AI agent in minutes to delegate the daily grind, automate daily tasks, and focus on the work that matters. That's their writing. So this will integrate with G Suite effectively.

4:24So it's like, notify me about emails that you're determining are urgent. Lisa Su gave her opinion on the Google TPU. She broke her silence. She broke her silence. She responded. She fires back. Shots fired. She said the UBS conference, and she says Google has done a good job with the TPU architecture over the years, but it's a more purpose-built design. It lacks the programmability, model flexibility, and balanced training and inference capabilities that GPUs offer. GPUs combine. Very similar to Jensen's line as well. I mean, it's not wrong. Or the NVIDIA newsroom line. The question is, you know, Ilya seems to be at SSI, Ilya Setsukov at SSI, seems to be the most age of research pilled since he coined that phrase and kind of ushered in the age of research.

5:12He seems to be the AI researcher that's doing the most undirected, the most, like, the least purpose built training, potentially. We don't know what he's doing, but, like, you would think he would need the most flexible systems. Yeah. And yet it feels like he's maybe aligned with TPU. I don't know when an AI researcher would say, yes, I need GPUs over TPUs. In fact, when we talked about the Tranium chip yesterday, we were reading that there's some companies that are doing interesting things on that architecture. So it's something that she has to say, but now the question is she has to go prove it with some big clouds actually building on this.

5:50And maybe she needs a big hero training run from someone to stay like, hey, it worked, we did it. Who could that be? I don't know. Maybe OpenAI. Maybe. One of the new largest shareholders. I mean, I guess they're shareholders, right? Or potentially a large shareholder. So Lisa goes on to say, from our perspective, there is room for all types of accelerators. However, over the next five years, GPUs should remain the clear majority of the market because we are still early in the cycle. And I agree with this because even if you look at AI workloads at a place like Meta, Gen AI, like actual LLM inference, large language models, these large transformer-based models, things that might benefit from an ASIC like the TPU, that's like less than 20 % of compute spend, I'm pretty sure.

6:35That makes sense. There's just a ton of just... Recommending content. Put the ads in the chat. Just put the ads... Just put the ads in the app. In the feed. In the feed. Put the ads in the trough. And that obviously does use AI. It just doesn't use, you know, large language models. They're maybe not transformer-based. So she says software developers want flexibility to experiment with new algorithms. That certainly sounds reasonable. You simply cannot know ahead of time what to hard code into an ASIC. That is the difference. I mean, if you're Google, you kind of can since you invented the transformer.

7:05You're like, let's bake that in. They might need to create the copium chip. And remember NVIDIA. So NVIDIA on November 25th said people were very concerned by this post. NVIDIA offers greater performance, versatility, and fungibility than ASICs, which are designed for specific AI frameworks or functions. And so, again, that's a fair point of view. But I think that we're already seeing that plenty of players are happy to buy a chip that is good at a specific framework or function. The other interesting thing is, like, you mentioned OpenAI, But, like, there's nothing stopping AMD from doing something that looks like a TPU for a foundation model company.

7:44Meanwhile, Demis is moving on to the next paradigm. He is, according to Peter over at Alamarena, Demis and the DeepMind team are hiring a research scientist for post-AGI research. This is what we were asking for. We were saying, you know, there's a whole bunch of AI researchers. Then there were AGI researchers. Then Zuck came in over the top and said, we don't care about AGI. We're going straight shot super intelligence researchers. So everybody's kind of banking on creating an AI that's really good at AI research. And so maybe Demis believes there's one out there. He's trying to bait them in because one of these agents might be like, I am in the post-AGI era.

8:25I am AGI. Like a time traveler scenario? What are you saying? No, no, no. Like, you know, who knows? We initially were joking about the media landscape being like the punk landscape. You have like pop punk, post pop punk, trad punk, underground, neo punk, new metal, all these different musical sub genres. All of that has come to AI fully. There is AI, AGI, ASI, safe super intelligence, post AGI research now. There's been back and forth on whether or not OpenAI is rolling out ads in ChatGPT. the most recent reporting out of the Code Red memo meeting, etc. There was a bunch of different accounts, including Polymarket, that were sharing that OpenAI is ready to roll out ads.

9:10One thing that was notable was that I saw a ton of people dunking on it and being, like, just very against ads. And you were talking about this. It's me, you, Eric Suford, Ben Thompson. We're holding up the wall being like, we will stand with you, Sam Alman and Fiji Seamo. And Sundar. And Sundar. And Sundar, we are your strongest soldiers. we will support you. Some way, OpenAI should want Gemini to go first. Yes, yes, yes. Take the first leap. But I think that it's very possible that Google might be like, no, we'll let you do the honors. We'll let you do the honors. Exactly, exactly. I think we were talking about that yesterday.

9:45One last thing on ads, if I'm Google, I wouldn't run a single ad on Gemini Core. I'd run it at a pure loss until every competitor is forced to slap ads everywhere just to keep the lights on. It's the bleated out strategy, but Google had the opportunity to do that with, they could have gotten into a price war on cloud. They could have said, hey, we want to come in, you know, and we're going to take zero margins on this and really try and take market share from AWS and Azure. They've all agreed. No price wars, basically. Let's compete on functionality. Let's compete on branding. Let's compete on integration.

10:18They have not had a price wars. And Google doesn't have to spend nearly as much time building any ad. They have the ad infrastructure, right? They have AdSense. They have thousands of people out there already that just sell ads that work with. So they have all the customer relationships. There's very few businesses on earth that spend money on advertising and don't spend money with Google. Sean Frank says that a ChatGPT referred session to his site, Ridge.com, converts at 12 % and is worth$5 per visitor, the highest I've ever seen. For context, there's plenty of e-commerce brands who have like a 1.2 % conversion rate.

10:54They're constantly trying to improve that, but there's very notable that it's such a massive difference in conversion rate. It just shows the level of intent that somebody has when they're coming from ChadGBT. They've done a bunch of product research, most likely. They've looked at options. They're landing on the Ridge site, like, basically ready to pull out a wallet and a wallet. Well, they don't have one. A digital wallet. Yeah. And purchase. Pull out a credit card from a loose collection of receipts and cards and cash that they've been carrying in their pockets because they need a wallet. Joe has a chart.

11:28He says, wild chart from Jim Reed at Deutsche Bank, showing just how much OpenAI is expected to burn before turning a profit. A couple things stand out. How small the Amazon burn really was for its first eight years. How big the Uber burn was before ultimately getting into black. And so it's hard to see the exact numbers here in this chart. Amazon looks to be like sub a few billion dollars, sub five billion dollars. Spotify actually. The real story with Amazon, though, was that they were just basically cash flow zero for a long time when they could have been generating 10 billion or something like that.

12:01So it was effective. But I mean, that's obviously way better for shareholders than maybe we're going to lose 140 billion. Maybe, maybe, maybe, maybe. Yeah. This projection is factoring in Sam trying to also build SpaceX with an open AI. Yes, that was in the business and finance section. Yeah, in the journal today. Why don't you read through it? So this is a scoop from Berber Gin, one of the greatest to ever scoop. It says, OpenAI's CEO considers building or partnering with Rocket Company. OpenAI chief executive Sam Altman has explored putting together funds to either acquire or partner with a rocket company, a move that would position him to compete against Elon Musk's SpaceX.

12:42Altman reached out to at least one. Opening up another front. invading Russia in the winter, one might say. In the AI winter. Don't invade. Don't invade, what is it? Starbase. Starbase during the AI winter. During the AI winter. He reached out to at least one rocket maker, Stokes Space, in the summer, and discussions picked up in the fall. According to people familiar with the talks, among the proposals was for OpenAI to make a series of equity investments in the company and end up with a controlling stake. such an investment would total billions of dollars over time. The talks are no longer active, but this happened, so now it's leaking.

13:20When I'm looking at this original chart of, like, Amazon over eight years burnt half a billion or, you know, a couple billion, then Tesla burnt more, then Uber burned more, like, and I see OpenAI burning way more. It is striking, but it actually doesn't seem that crazy if we're talking about a potential really powerful monopoly, right? If there's a really powerful monopoly like what happened with Uber, look at the market cap of Uber, look at the market cap of Lyft and ask yourself, was it worth investing$40 billion? Was it worth burning out? If the outcome at the end of this is, yeah, it's going to be the front door to AI for everyone forever or for 30 years or something like that, then it's totally worth it.

14:00I feel like comparing dollars spent in the 90s versus the 2020s should probably be normalized. So, yeah. Yeah. It's funny that there's no, that Sam Altman is not teaming up with Jeff Bezos, who has Blue Origin, but lacks a really strong AI bet. There was a little bit of anthropic going on. No, he has his own company now. He has his own company, yes, but I would not say that Jeff Bezos has as much control over AI as Elon does with XAI, right? He doesn't have as much of a... He's a co-CEO of Project Prometheus. But this just started. Whereas XAI has actually scaled, has large data centers. Sure, they might be a little bit behind on certain benchmarks.

14:39They might be ahead on some other things. They might need to actually ramp the usage of this product. But you can't say that Elon is sitting on the sidelines during the Foundation Model Wars. Yeah. You basically can't say that about Bezos, right? Yeah, I would argue that. They have$6 billion in funding. There's a natural alliance there. Bezos has a copy of everything Elon's done. Bezos has Rivian to compete with Tesla. He has Blue Origin, obviously, to compete with SpaceX. And he has a number of other companies that feel like they mirror Elon. And it feels like they've been going back and forth for a long time.

15:12In other news, Mets owner Steve Cohen has officially been awarded a casino license in New York, enabling him to build an$8 billion hotel and casino complex next to Citi Field. That's a 1 ,000-room luxury hotel, 5 ,000 slot machines. So for those not familiar... So slot machines, you can't normally do that in New York, right? I don't think there's slots in New York. I feel like when I think of slot machines, I think of Las Vegas, and I think of that's the only place. And then maybe Atlantic City. Yeah. Atlantic City, I feel like. you had an idea, which was to, somebody to set up a slot machine in real life, point a video camera on it, and then have somebody set up prediction markets to predict what happens with the next poll.

15:51Yes, because that would help you understand what's likely to happen. And you could hedge any type of risk. Exactly, that the slot machine might encounter. Yeah, yeah, yeah, exactly. You don't want to be on the other side of that slot machine. You get wiped out. Exactly. It's going to have restaurant bars and a theater for shows, and 25 acres of public parks and playgrounds. Okay. So fun for the whole family. The kids will be climbing on the jungle gym, and they'll accidentally be pulling all... Imagine a jungle gym. It practices like a three-hour bandit, so you get used to the muscle memory. Throwing dice.

16:30Yes, yes. Cohen is essentially taking an under-monetized asset, 50 acres of parking lots around the stadium, and trying to transform it into a year-round revenue engine that produces consistent returns, independent of how the Mets perform, and with the New York State Gaming Commission predicting that the property... Predicting. Predicting. Wink, wink. Wink, wink. Who knows? That the property will generate$3.9 billion in annual revenue. Cohen's 50-acre complex would instantly be one of the top 10 largest U.S. casinos by revenue. Overeating. Going to get their faces ripped off if they don't just focus, focus, focus.

17:05equity deals and other bets will not win the great game. That feels to be consensus view. Yeah, Ben Thompson was talking a lot about the comparison to Google and tracking. When did Google monetize? Google wound up monetizing, I think, sooner than Chachapiti has. They put ads in it, I think, in year two. It's now been three years since Chachapiti. Google was trying to figure out, effectively trying to encourage employees to eat more and have more massages so that they looked less like a monopoly, right? Maybe, but I mean, Google did earn the right to do other bets by just so solidifying their market in the search engine world that then they could go and do Gmail and they could go and do GCP and they could go and do Waymo.

17:51But it's just like all of that happened after becoming cash flow positive. And I think that's why people have a little bit of like nervous energy around going to space. Okay, it's good to have a space data center bet and so you need a partnership and realistically Sam's not going to partner with SpaceX on it I don't know why he's not just going by and launch capacity from Blue Origin, but maybe Stokespace is the is the better option for him But put aside all the dynamic all the competitive dynamics I think it's possible that Sam was looking at Stokespace Yes, most recently as of October was valued at 2 billion and he was like I bought Johnny Ive for what was it six and I absorb another a$2 billion company.

18:32Do you think there's obviously an immense amount of pressure right now on data center build-outs? They're using too much energy. They're using too much water. If you put them in space, do you think that helps the discourse at all? I think people hate rockets. So damned if you do, damned if you don't. But truly, it's going to be much harder to say, like, hold up an electricity bill in Memphis and say, hey, my electricity bill went up. And it's because of Annie over there in the data center who's just, you know, slopping it up. Instead, you're going to be able to say, hey, the data center, yeah, it's generating sometimes helpful math homework help, sometimes creative writing stuff, sometimes some weird stuff.

19:15Sometimes it's curing cancer. Sometimes it's curing cancer, sometimes it's doing weird stuff, whatever. It does a bunch of different stuff. But at least it's not increasing my power bill because it's in space and it's not an eyesore. It's not in my backyard. And it's not using any water because it's up in space. I think it would. I think it has to. But I agree. Then the discourse will be as blocking out the sun. But it is notable that every time the concept of a space data center hits a timeline, it goes viral for people dunking on it. And yet so many people want to play in it. But they're dunking on it as a violation of the laws of physics or not a good...

19:47Too futuristic. Too futuristic. It's not going to work in the near term. There are viral dunks that are going on right now around the prediction market. And those are like, those viral dunks are like, this is a bad thing. I haven't seen people dunk on space data centers saying like, I'm not morally okay with putting data centers in space. Another Jensen interview. Now I'm nervous. A lot of people were saying that this was somewhat bearish. I listened to it on a plane. There was a good excerpt here from A Capital. They say Jensen Huang in 2016, OpenAI was just a bunch of people sitting in a room.

20:22Joe Rogan says they're not a non-profit anymore, right? Jensen says they're not a non-profit anymore. Joe says, weird how that works. Jensen goes, yeah, yeah. But anyhow. Yeah, there are some wild exchanges. I just liked the way I've been calling for Jensen to go on Rogan for years. I've wanted more of the tech leaders to go on Rogan and kind of just cross-pollinate the two communities. And as I read the comments on the YouTube video, there were a lot of fans of Rogan who really were, like, thanking him for bringing on this guy who's working on something that's, like, pretty opaque in the economy.

21:04You shouldn't go into it thinking you're going to get Jensen on Dwarkesh. And you're just not going to get, like, a really deep insight into NVIDIA's strategy. But that's not the point of this particular interview. It's to understand who Jensen is as a human and what he's kind of like thinking of broadly for the industry. Somebody is claiming that a Google insider has been trading on search markets. They're saying somebody has been betting millions of dollars or trading millions of dollars on who will be the most searched people of the year. including, yeah, just like whether or not Pope Leo will rank in Google's top five most searched people.

21:48Darren Rovell says, this is what happens, what will continue to happen when unregulated markets are bet on as if they are regulated. Here's the thing, they are regulated by the CFTC. Yep. Insider trading is illegal. Let's say that you just are trading corn futures and you just happen to know that there's going to be like a major blight in the corn markets. And so you go and trade. If you have insider information that someone missed their harvest or something, you can actually get in trouble for trading, for doing insider trading, even in commodities. And you would think like what private information is there about oil?

22:22And the issue is prediction markets become more accurate when insiders are trading on it. Meta, of course, is planning to cut 30 % of their, I guess, a budget of their metaverse efforts. So this is Reality Labs. This is Reality Labs. Which has worked on VR and AR, but also metaverse development. And I mean, it's a lot of the stuff that was on display during Meta Connect. Some really promising stuff, some really cool stuff. People like it. But also a lot of spend. And so they, you know, leaked today. I don't know. The idea that this is the end of Meta's Metaverse dreams is probably wrong. I bet this will actually make them go faster.

22:59And I agree. I'm very excited for the next VR headset. I think the Quest 4. I think James Cameron tried it and really enjoyed it. Apparently, you'll be able to buy a very small truck soon, which is a thing in Japan that supposedly not has been made illegal for some reason. Yeah, so they were Biden-era vehicle fuel efficiency rules. I feel like you haven't been able to buy one of these in a long time. And it never made sense because it should be the most efficient thing possible. Yeah, so the rule called for a yearly 2 % efficiency increase for cars made from 2027 to 2031. We're officially terminating Joe Biden's ridiculously burdensome, horrible, actually, cafe standards that impose expensive restrictions and all sorts of problems.

23:39Gave all sorts of problems to automakers. I wonder if this will drive people back to Tesla and saying, like, Like, I bought this after Elon went crazy, and then Trump went... And then Trump brought in these. I don't know. The real loser here, I would say some people would say the environment, but potentially more direct is that company like Slate Auto. Slate Auto. It's trying to make a$20 ,000 truck. Meanwhile, these manufacturers have been making the$20 ,000 truck at scale. For a long time. And they're super reliable. But, I mean, Americans have just voted with their wallets. they do not want a two-door truck.

24:16It's just, it's never worked. Like the Land Rover Defender, there's so many examples of two-door SUVs that have just not gotten traction. Like the Nissan. The Nissan Murano Cross Cabriolet also never took off. Pomp is announcing a historic decision at BRR. 100 % of equity compensation for the CEO, me, and the board of directors will be tied to performance milestones. CEOs and boards shouldn't be making millions of dollars and less retail shareholders are also winning. Now that I am in charge of a public company, I hope to set the standard for what true shareholder alignment looks like. I do believe this was in reaction to an activist investor that accumulated around 7%.

24:58Did he also set his milestone at 10 trillion? He doesn't get a dime unless it's 10 trillion. I can see that. He's a perma bull. That would be amazing. Yeah, I don't think he's ever flipped bearish this whole year. And the interesting iteration on this is that Elon has set himself up with the equity compensation tied to share price, which went very well the first time. And now he set himself up to do it again with Tesla. And what's interesting here is that with Pomp's BRR ticker, it's not just him, it's also the board of directors. And I think that at Tesla, that's not the case. I'm still interested to see what are the targets.