AMC CEO Adam Aron & OpenAI Chairman Bret Taylor 7/20/26

31:30Would you ever sell? Do you know about this sort of thing? I'm aware of it. I would like to get it because it's supposed to be less carby, more protein. Olive oil, protein. It's everything that popcorn shouldn't have. You'd be a big seller, Adam. You should do your own. Or buy theirs. It's called Cloud or something because of Chloe. Do you know anything about it? I know everything about it. because I've talked to the head of her company. I've talked to some of her key investors. I tasted the popcorn before it went to market. We're happy for her success. But we, too, are really successful. You know, we sell tons and tons of popcorn every single day.

32:15No one on earth sells more popcorn than AMC. and we know for a fact that it's an indulgence that people greatly enjoy because they sure buy a lot of it and they have a good time as they did. I had a bucket and I was one person. And I didn't eat all of it but I ate almost all. So Adam, are you then thinking of a lighter popcorn?

32:43No. I think we're going to go with the tried and true. And as I said, look, we're happy for her success. We're very happy for her success. But, you know, we took our popcorn to the home as well two years back. And it's also a$30 million brand in over 10 ,000 stores, close to 20 ,000 stores around the United States. People do love AMC popcorn. No, Adam, I'm not suggesting that you have to do her popcorn. I'm just suggesting there might be two variants of popcorn. Now, I don't know if that's one more skew than you want, but I think there's a huge opportunity. Actually, on my watch as CEO of the company, we're up to four skews of that popcorn.

33:31We have our regular solid stuff. We have some often we have gourmet caramel corn. We usually have a cheese corn. We usually have some concoction that we make up throwing lots of stuff in. whether it's Tito's or orange. Joe and I are looking for the lighter version, Adam. We're looking for the lighter version. I'm sort of the... I want the lighter version that tastes like the heavy version. I have an idea for you. I just thought of it right now. How about deep fried popcorn? You think... Would that be your standard of... I mean, I think that would be a hot seller. I hate to say it. Probably not great for you, but, you know, if it comes with an angioplasty on the side, maybe.

34:16Look, we're we're we're joking and we're kidding. But, you know, food. I'm going to be serious for a second. Food and beverage is a very important part of what we do. And your your your branded popcorn, too. I've seen that before. Like the collectible buckets. That's going to be additive. Right. What, one hundred million dollar additive? eventually? We're on the way. And look, about half of AMC's profitability comes not only from the sale of movie tickets, but the sale of food and beverage in our theaters. It has been a priority at our company over the past several years to try to figure out what Americans want to buy when they go to our theaters in the U.S.

35:05and what Europeans want to buy when they go to our theaters across Europe. And we do it pretty well because we sell more food and beverage than any other mass operator. And we are the largest movie theater chain in the world. So I think we understand what our customers want, and we try very hard to give it to them. And this weekend with Chris Nolan's movie, If you haven't seen it yet, you've got to see it on the big screen. We served that food and beverage to 4.3 million people. What a great weekend, concluding just a great quarter. All right. Fantastic, Adam. It's good to have you on. It's been a while.

35:51You look fantastic. No worse for the wear. Hey, guys, I just want to say this. It's no secret that back in November on a business trip to London, I had a very mild stroke. The two of you were very complimentary and on air. I greatly appreciate your friendship and I'm happy to report I have made a spectacular recovery. I am in fighting shape and ready to continue to do battle. Fantastic. We're thrilled for you. Good to hear that. Thanks. And thanks for doing it here on Squawk Box. Appreciate it, Adam. You bet. See you soon. Coming up next on Squawk Pod, Brett Taylor, chairman of OpenAI and co-founder of Sierra.

36:41He talks about the evolution of artificial intelligence in the U.S. and in China, plus what the future of AI efficiency looks like and who will bring us there. At the end of the day, it's going to be, you know, how efficient is your inference? And I think the frontier labs are still far ahead here. And I think that's really where the value is going to accrue. because at the end of the day, you know, as you as a customer... Inference being every time I put in a prompt, I get an answer. That's right. It's essentially how efficient it is to use these models for a particular price point. I think this is an area where the Frontier Labs still have a significant advantage.

37:20Good morning, students. Zoccher teaches us teamwork, leadership, geometry, art, physics, and a lifetime of lessons we can take with us long after we leave the field. That's why Bank of America and U.S. Soccer are committed to helping bring soccer to every school. Soccer is officially in session. Raise your hand to help bring soccer to schools at bofa.com slash soccer at schools. Trading at Schwab is now powered by Ameritrade. Unlocking the power of Thinkorswim, the award-winning trading platforms loaded with features that let you dive deeper into the market. Visualize your trades in a new light on Thinkorswim Desktop with robust charting and analysis tools.

38:01All while you uncover new opportunities with up-to-the-minute market news and insights. Thinkorswim is available on desktop, web, and mobile to meet you where you are. It's built by the trading obsessed to help you trade brilliantly. Learn more at schwab.com slash trading. Need a lesson in savings? Who doesn't? That's where Burlington comes in for back-to-school style. Because those school lists get long, but your receipts shouldn't. Find backpacks and sneakers to sweatshirts and jeans. Burlington has them all from the brands your kids keep begging for. Why choose just a few looks? With our prices, just get them all.

38:41Burlington. Deals. Brands. Wow.

38:50Welcome back. This is Squawk Pod from CNBC. You're watching Squawk Box right here on CNBC. I'm Andrew Osorkin, along with Joe Kernan, as we said at the top of the six, just the boys today on this summer Monday, joining us to talk through everything AI this morning at the table. We're thrilled to have him. Brett Taylor is the co-founder of Sierra and the chairman of OpenAI. And there is so much debate going on right now in your world about open weight models, Chinese models, token maxing, the ROI on all of this and everything in between. I think the biggest thing that's happened even just the past couple of days is this new Chinese model that's open weight, potentially distilled or maybe stolen.

39:36I don't know why we use the word distilled, but from other models. I don't know if it's from yours or from Claude. There's some questions about that. But how do you think about that piece of it? Because literally the debate on X right now among, I think, a lot of your friends is about what this all means. Well, I like to take it back to first principles, which is why do people choose the model or the task at hand? And I think there's really two topics that I think every time I talk to probably 100 CEOs every single month. And so I'm pretty plugged into Top of Mind. And there's two things, which is AI tokenomics, which is how can I know that I'm getting the value from this spend on tokens?

40:13And then I'll use the word sovereignty, which is just making sure, like, what is my competitive moat in the age of AI as well? And I'll say one thing that I think is a little bit overblown about these open-weight models is they're not necessarily cheaper to run. Whether or not they're cheaper to train, you don't care. Because you're using just as many tokens. In fact, they may be less efficient. Yeah, there's this thing called token efficiency, and it turns out the frontier models are much, much more token efficient. And beyond that, just for the audience, because there's some people who are watching and maybe like token, token, what?

40:43Why do we call effectively energy expenditure token? Is that a fair description of what's happening here? I'll give an overly simplistic view that I'll tell you why it's wrong. You know, you can think of a token, a token is to intelligence like a watt is to electricity, like a unit of intelligence. And token efficient means how many tokens does it take to complete a task? To answer the question you've asked. To answer the question. But not every token is actually equal. And if you've used a really advanced model like 5.6 from OpenAI, you'll know, actually, wow, the code it writes is a lot better than the previous model.

41:18So it's not exactly the right analogy. And that's the thing is actually I think what we're seeing is for a lot of tasks, it turns out these frontier models from OpenAI and Anthropic are actually just better than these open-weight models. And I think it's important for the labs in the United States for any, if you just think about what do you want from a model, you might want it to be fast for some task. Let's say you're looking at every credit card transaction. You might want it to be really cheap because, you know, this task isn't a frontier task to do. You might actually want the highest level of intelligence.

41:46If you're an insurance company and you're doing actuary work, you want it to be great. I think the U.S. labs need for every price, performance, latency, cost, they need to have the best model available. And actually, I think that's more likely than not, just because it turns out that just having open weights isn't actually the main thing driving any of those costs. And, you know, token efficiency, compute efficiency is something that I think the Frontier Labs are acceptable. But let's also bring the audience that may not be on board with us into the other piece of the conversation about open weight models.

42:15In particular, this they're calling Kimi or K3. Three, this is a model out of China that some people believe is as good as, if not better than, OpenAI's 5.6 or Fable from Anthropoc in terms of what it can produce. It may be less efficient, but at the same time, ostensibly should be cheaper. And the other question is how you think they created the model. Some people think the model was effectively stolen. Other people think they actually built it themselves. What do you think? You know, this is my point, which is, you know, is it cheaper to use? is the most important part for anyone considering it.

42:50And I don't think that, I think the jury's out on that. Right. Cheaper to use for you, you and I, yes. Yeah. But there's also a question of how much it costs them to build it. And the reason why that matters is because a lot of the big frontier model makers in the U.S. are spending extraordinary amounts of money to build massive data centers, in part for us to use the models, but in part for you to build the models. And if you could build the models for a lot less, that's a different game, too. Yeah, I think at the end of the day, it's going to be, you know, how efficient is your inference? And I think the frontier labs are still far ahead here.

43:25And I think that's really where the value is going to accrue. Because at the end of the day, you know, as you as a customer. Inference being every time I put in a prompt, I get an answer. That's right. It's essentially how efficient it is to use these models for a particular price point. I think this is an area where the frontier labs still have a significant advantage. And you're seeing that from everyday use. I mean, the people, not everyone's used all these models yet, so I don't want to overstate confidence here. But if you're using things like Fable 5.6, K3, 5.6 is still actually, just if you think about what does it take to write this computer program with AI, 5.6 is leading by most of these metrics.

44:00And I think that's what the most important thing is. The other piece of this conversation is we had the CEO of Ramp in here just last week. He actually built a system so you can actually see how much money you're spending at any given moment and whether you're using the right models. Meaning sometimes people are asking the highest end model, you know, what the weather is. Yeah, like driving a Ferrari to the grocery store. Which is like a terrible efficiency situation. And we've had other people. Derek Khosrowshahi was on the broadcast earlier this week. He's argued that he's not seeing the ROI, the return on investment.

44:36The potential for AI to transform literally every single thing that we're doing as a company, to build out better services, deliver them at lower prices, and build incredible new experiences. We had Alex Karp in that famous now viral video where he says, nobody's getting their money's worth for all of this. These people are livid. They're like, I am paying for tokens that create no value. These people are stealing the weights and alpha of my business, and they're creating a wealth tax that does not help the poor. It just punishes, starts with the billionaires. Every single person at this table is going to be paying a wealth tax only to punish us.

45:17And the reason for it is because these models have been completely over irresponsibly overselled. And the sell is it's dangerous for everyone, which is why I can give it to all your adversaries, but I can't give it to the Department of War or I can't safely give it to an enterprise in this country without being certain that the alpha of that business could transfer to this model tomorrow. How do you think about that piece of it? Well, bring it to my day jobs here, because I actually think the market's in kind of a quirky spot where everyone's worrying about paying for tokens. And it would be like if you signed up for Gmail and you paid for CPU cycle or something.

45:52I think we're, I believe where the world is going is paying for outcomes. So at Sierra, we make AI agents that help companies drive revenue and drive customer service. But our customers don't pay per token. Right. And by the way, this is Sierra. Why don't you explain what Sierra does for those who don't know? Yeah. So at Sierra, we help power things. Like if you call DirecTV or SiriusXM on the phone, you'll talk to an AI agent now rather than a person you don't need to wait on hold. We just launched a new product called Horizon, which are long horizon agents that help people originate loans or do pre-authorization for medical procedures and truly drive revenue.

46:29But what's really interesting about it is two things. One is you don't pay per token. You pay per outcome. You pay per originated loan or per pre-authorization. And number two is we help you develop a competitive moat in AI because we essentially help you compound the data about your customers. So no matter as frontier intelligence improves, you're actually developing this compounding asset, which is the data you know about your customers and your relationships with your customers. And our thinking is it really addresses these two issues in AI, which is sovereignty and tokenomics. Let me ask you a question about your own tokenomics.

47:04So for every time that you're having to process one of these phone calls, I imagine you on your end are having to use an extraordinary amount of tokens. So does the math math for you, meaning when you have a successful outcome for your client, your customer, and they pay you per success, are they paying you more? I'm hoping you're going to say yes. I'm hoping they're paying you more than what it costs you to actually produce, but I'm not sure. Yes, they are. And I would argue it's easier for us as an applied AI company to essentially amortize the R &D cost of developing specialized models for low latency voice conversation.

47:47That is for every company in the world to do that individually. And I think the solution to a lot of the problems with AI tokenomics is we don't have a mature applied AI market. So rather than every company licensing tokens and building things from component parts, if we have more companies like Sierra providing these solutions out of the box that can pick up the phone for your toll-free number, and you pay for those outcomes, we can bear the burden of managing tokens. Tell me about the bearing the burden. The other piece of this, and it gets to maybe what the CEO of Ramp was talking about, is why are the model companies, maybe it's the incentives are just misaligned.

48:20And it doesn't seem, I mean, there was a period of time where OpenAI, for example, or ChatGPT would automatically send you to a certain model. But I don't think it does that anymore currently. And so most of the model makers are sort of forcing you to choose, do you want to use the Ferrari or do you want to drive in the Honda today? And most people are choosing the Ferrari. But what should be happening, I think, is when you say, I'm just going to the grocery store, I'll take the Honda or it should automatically say, well, we're going to use the Honda for that. But if you're going to go racing in the afternoon, we're going to put you on the Ferrari.

48:58And people aren't doing that. So people think, oh, they're just sort of forcing me to overspend. It's not forcing us, but that that the sort of human instinct is to choose the top top vehicle. I think the market's just immature. You know, we've all gotten these weights, these new models since, what, December of this past year. I think if you fast forward 12 months from now, IT departments will be really sophisticated about sort of, I'll say, the industrial applications of AI. For your marketing department, you might have one thing. For your software engineers, another thing. I think there'll be products and solutions developed in this space to help you like the RAM solution you mentioned.

49:33Applied AI companies like Sierra or Harvey in the legal space will be there. So you just don't need to think about the word token at all. So I think we're in this period of immaturity. It's a little bit like, you know, the mid 80s with the development of mainframes and PCs or the early days of the Internet in 1997, where if you wanted to get the value from the technology, you had to build it yourself. And if you go back to 97, people were spending tens of millions of dollars just to make a website where people could log in. And now it's easy. And we're just in this early stages of the technology.

50:04And I think it's a call for entrepreneurs to develop solutions so businesses don't need to deal with this. Well, let me ask you a separate question, and this maybe gets the open AI piece of it, but also your own business piece of it, because there were some fighting words on Twitter last week from Sam, I think effectively trying to go after Anthropik a little bit with the idea that, you know, there's been some claims more recently that there's a view that Anthropik is increasingly going into other people's business. The Figma situation is an example of that or, you know, Claude or Legal. I mean, they're trying to go into these different areas.

50:36And I think Sam was trying to say that OpenAI doesn't intend to do that. But I wonder how you think about that, because invariably, these models should be able to one day, if they're so good at what they do, be able to do what your company is doing and everybody else's company is doing. Well, I think at the end of the day, if you fast forward a few years, intelligence will be widely available. So you're not going to be able to differentiate on intelligence because you can just license it. You can rent it from all these frontier labs like OpenAI. So the big question for every CEO right now is, where is my differentiation?

51:10And as I said at CIRA, our opinion on this is that the relationship you have with your customers and the interactions you have with your customers can be a durable moat, if you will, in this age of AI. The way it works with Horizon is, let's say you make an AI agent that originates loans. Every time it makes a successful sales call, it learns from that, and it gets better over time. But that's your information. It doesn't make its way to your competitors or the frontier. But I think if companies are thinking they can differentiate on the models, I think that's probably a fool's errand. You know, at the end of the day, these are super intelligence agents that are getting better, especially as we get closer to recursive self-improvement.

51:49So I think at the end of the day, companies need to say, when super intelligence is democratized and widely available, what's my place on this chessboard of the competitive landscape? And that's a very nuanced question. But part of that question is, you know, should you be working or partnering with one of the big model makers? I mean, part of the Figma story was, oh, my goodness, they're now, you know, they partnered with Anthropik. Then all of a sudden Anthropik's in the design, you know, the cloud design business. Did they figure out stuff from the prompts that were going into it through the Figma relationship?

52:22How do you think CEOs should think about that? I think it's the right question. I mean, you need to understand, you know, what is your competitive dynamics over time? It's why at OpenAI, we've been really investing in our ecosystem. It's why that we I think it's very important that CEOs understand, you know, what's mine is mine and what is the the AI models is theirs. But is it is that's the question, is it? I think that's, by the way, what Alex Karp was getting at. When you put the prompt into the system is what mine is mine, really mine. I think it should be. And most large enterprises have contractual commitments that guarantee that.

52:56But it is the right question. And as I said, I think one of the misconceptions I often hear is, how can I make my intelligence better than my competitors? And what I'm saying is it's a little bit like the internet. Everyone's going to have a website. Having a website is not a competitive advantage. Having intelligence is actually going to be quite democratized. And you need to think about what are the things around it, like your customer relationships, that can be durable in the face of this super intelligence. You've been asked about it before, but I'm going to ask about it again, this Apple suit, suing you, suing OpenAI.

53:26This is claiming in a very sort of elaborate way that you were stealing their IP and orchestrating the stealing by effectively not only hiring people from that company, but educating them and instructing them how to take the IP out of the building, if you will. What do you know now? Well, I can't comment on it specifically because it's an ongoing investigation, but I can say unequivocally, we have no interest in any other companies trade secrets. And, you know, we're just trying to innovate, especially as we think about the future of intelligence. I think it will lead to, you know, perhaps devices that go beyond the smartphone.

54:03You know, fundamentally, the really interesting thing about AI is we could talk to computers now. You know, we don't need to type. We don't need to tap. But does it concern you, though, on a trust basis? Because part of the whole issue right now is should other companies trust working with open AI? Should they trust working with Anthropoc? Who do you trust? Who do not trust. And so when you see a lawsuit like that, invariably, I'm sure there are companies saying, yeah, that's happening. How much should I worry about whatever else is happening behind the scenes? Well, look, we take this very seriously.

54:31I don't want to misrepresent that at all. But as I said, we have no interest in other companies' trade secrets. And we're really trying to create the best models that every company in the world can depend on with trust. I got one more for you because you're the chairman of OpenAI. And you follow, I'm sure, Sam and others on X. And sometimes there's this back and forth that's kind of remarkable. And I want to know what you think as you're as you're listening to this. So at one point last week, Sam wrote to Elon, homeboy, you're the one selling public market investors on short term space data centers.

55:03This is after Musk went after him on the Apple front. And then Musk is responding by saying after stealing an open AI charity, then you stole Apple's technology and basically tells him that he can come watch the data centers going into space if your parole officer approves. And there's something sort of comical about all this. But I also wonder, as the chairman of the company, if you're saying to yourself, I wish my CEO was not on X right now. You know, what I care about is that OpenAI achieves its mission, that we have the best models available. And fundamentally, we achieve our mission of benefiting humanity with AGI.

55:44We live in a very authentic era. Sam leads with his heart and his head. He's authentic on social media. And I wouldn't change it. I think he's a remarkable leader. And, you know, I sometimes get these things sent to me and I laugh and occasionally wince. But, you know, he is who he is. And I'm very, very supportive. Right. It's great to see you this morning. If you go to a movie and nobody else wants any popcorn. Oh, now you're going to popcorn. Back to popcorn. Will you get the bucket or will, because it's more convenient or the bag that kind of falls over? Absolutely get the bucket. You'll get it for one person and you'll put that fake stuff on it.

56:23Absolutely. And you'll save 1 ,400 calories. Damn the torpedoes. And salt. It's the most important part of the film for me. And by the way, when they say they don't want any, they're going to share it. You know they're going to take it. Andrew, we talked before because he had seen the Adam Aaron. And you want this. What is this? You want this kale based popcorn with it's all like. No, but I, you know, I'm a feast or famine kind of guy. I go crazy. And no, sometimes. No, I don't like that. That's the running joke on the show. No, some days I eat like a crazy person. And then some days you're trying to be a little bit better.

57:00But at a movie. And so if you're at the movie, the idea that they're now selling popcorn that's like a marginally better for you. That seems like a good idea. You try to get out of there and get rid of the old popcorn. No, I'm not trying to get rid of the old popcorn. I'm trying to add one more skew to the variety. For variety. They want us to have nice things. Yeah. Yeah. I'm with Brett on this. Brett's going to. Are you going to buy the. What did he say he was going to try to sell? Adam said he was going to try to sell fried popcorn. Oh, wow. How do you feel about fried popcorn? I'm a traditionalist, I have to say.

57:30I want the... Fake butter. Yeah, exactly. And a lot of salt. On the bucket. Yeah. Okay. And we're aligned on this. I'm losing. I feel like I'm always losing. Brett, it's nice to see you. Thank you. And that is Squawk Pod for today. Thanks for listening. Thanks for starting your week with us. Squawk Box is hosted by Joe Kernan, Becky Quick, and Andrew Ross Sorkin. You can tune in weekday mornings on CNBC at 6 Eastern or get the smartest takes and analysis from our TV show right into your ears when you follow Squawk Pod wherever you get your podcasts. Have a great Monday. We'll meet you right back here tomorrow.

58:09We are clear. Thanks, guys.

58:41and subscribe at schwab.com slash market update podcast or find Schwab Market Update wherever you get your podcasts.