Nvidia’s China Shift, Meta AI Drama, Paramount's Pitch | Diet TBPN

0:01NVIDIA can sell H200s to China now. It's a big reversal from the chip controls a few years ago, which actually started during the Biden admin. We were really pushing to not allow big, you know, the best and most amazing chips to go to China. Well, all of that's changed with this H200 news. Now, Nvidia can sell to China, but they got to pay 25 % to Uncle Sam. The big man. The big man, Uncle Sam. The estimate for you, the taxpayer, assuming you're an American taxpayer, is that you will be$5 billion richer because NVIDIA is estimating that they will be able to sell$5 billion of chips to China every quarter.

0:43That's right. 25 % of that, of course, is$5 billion per year. But it's unclear where this will actually pan out. I'm not sure where it will land. It is an interesting question because there's been a lot of diversion. There's been a lot of engineering that's happened in China to work around the frontier. but then also they might have gotten their hand on some Blackwell. There's a whole bunch of crazy headlines going back and forth. So there's a bunch of good arguments on both sides, and there's a bunch of terrible arguments as well. The best argument that I have heard for allowing an American company, NVIDIA, to sell H200s to Chinese companies is basically just free trade.

1:22Keep the government out of the boardroom. NVIDIA is over 30 years old. They make computer parts. If you don't believe in nuclear-level AI capabilities, then why would you be so worried about AI going to China? It's just auto complete. It's just better SaaS. You know, yes, it might increase their GDP, but there are tons of things that we sell to China that does increase their GDP. We bring the business of manufacturing iPhones there, right? We bring them quite a lot of business. We sell soybeans. We sell all sorts of food. They sell us stuff. Like free trade has happened for decades. Like, yes, there is a little bit of a decoupling, But for the most part, the modus operandi of the American geopolitical system has not been try and reduce GDP growth in all of our geopolitical rivals at a full stop.

2:10Right. No, it's never been that way. And at the end of the day, if you're an American company, like I am sympathetic to this idea that Jensen is just like, look, I'm just trying to build a big company. I want to sell my stuff everywhere. I think my stuff's good. I don't think it's dangerous. I don't think I'm making nuclear weapons. I didn't get into the nuclear weapon business, and now I'm being treated like I'm selling nuclear weapons. It doesn't quite make sense. I understand that there are other - It's hard for him to argue, go over there, go to Washington and argue. It's not, it's really not that important.

2:41It's just kind of like an extension of the existing software paradigm. It's just SAS. Well, yes, yes, yes, yes. You're right. It's hard to sell this to. Because you can't do that and then also, you know, continue to sell this. Yes. Did we just lose power? What's going on, boys? Yeah, what's going on with the lights? Probably a nation state, as usual. Who knows? Trying to take us offline. One sort of gigabrain take is, you know, if you believe that AI slop will one-shot people, it will reduce the birth rate because people will be so obsessed with AI avatars, romantic companions. And if you are anti-China, maybe you should export as much as possible.

3:20I don't see anyone making those arguments today, but I do wonder if there's someone, if you are doom-pilled in that way where you think that AI is bad, but you also think China is bad, do you have to argue in favor of the export of the H200? Maybe. Let them have. I don't know. We also don't know exactly how big of a deal this H200 unbanning is because last time there was news on this front, the CCP basically said like, hey, we don't want our Chinese companies buying these, even if they are available. We don't want them. We don't want them here. And the CCP, you know, kind of doesn't want NVIDIA chips filling up Chinese data centers.

4:01It feels like China, at the very least, has been somewhat receptive to this idea that they don't want to become dependent on the American tech stack. But clearly, Chinese companies do want NVIDIA. We saw this with the Deep Seek team figuring out how to get some Blackwell apparently. And it's actually going further than that. There's some Chinese nationals that have been recently risking legal consequences from the American justice system. This is very interesting. On Monday, the US Attorney's Office in Houston revealed something called Operation Gatekeeper. Very exciting name. The government is weirdly great at naming and coining some of these operations.

4:40Interestingly, when you Google Operation Gatekeeper, you get something about building a border wall, but from the presidency of Bill Clinton. So Operation Gatekeeper originally was a measure implemented during the presidency of Bill Clinton by the United States Border Patrol, then the INS, aimed at halting illegal immigration to the United States at the US-Mexico border near San Diego. They constructed a wall on the beach in the 1990s. he literally built a wall, which is crazy to me because I thought that was like a Trump era thing. Anyway, the new Operation Gatekeeper... The operation for that Canadian snowboarder is called Operation Giant Slalom.

5:20Giant Slalom, that's a good name too. So the new Operation Gatekeeper has nothing to do with the domestic immigration across the U.S.-Mexico border. It instead is focused on AI chip smuggling, and this particular operation has just uncovered$160 million worth of illegal AI chip smuggling. A 43-year-old Chinese citizen named Fan Yui Gong was living in— Yeah, his name's Gong, which, I mean, we love Gongs here, but I don't love this because— Before we figure out the bad stuff that he's doing, let's give a quick wrap. That's a fantastic name. So this guy's living in Brooklyn, 43. And also separately a Canadian citizen from China.

6:06They independently conspired with employees of a Hong Kong-based logistics company and a China-based AI technology company to circumvent U.S. export controls. Basically, they bought a ton of NVIDIA GPUs, then they would send them to a co-packer, like a fulfillment center, remove the NVIDIA labels, put new labels on the chips that would say Sand Kian. Once the chips were packaged back up, they would ship them. Oh, I said chip them. I need to change that. Probably already went out. They would ship them to China or Hong Kong under a generic classification of generic computer parts. And so everyone involved faces significant legal consequences.

6:50Now, I couldn't figure out exactly what the consequences are. I'm not sure if it's like years in prison, but it doesn't seem good. The question then becomes, how did it get in their hands initially? Oh, because they were just like, hey, I'm a Brooklyn guy. I want to buy some NVIDIA. I know, but$160 million, it's a big company. They do a lot of revenue, but you would think, oh, what's this address that we just got this order from. Oh, is that a... It'd be like, yeah, I'm shipping it to a company in Brooklyn. Oh, it's probably like a coffee shop or something like that. Yeah, maybe it's like, you know, I want to have an accelerated design studio in Brooklyn or something.

7:26I don't know. I know, I know, but the... Seems like a KYC failure. I don't know. There was one of the guys involved, it was technically a Canadian citizen who was from China. And so you stack these all these up and you have like, I'm shipping it to Brooklyn and the guy's a Canadian national. And you know, at a certain point you're like, and it could have been through some sort of reseller. I don't know. I'm sure we'll find out more. His last name's Gong. His last name's Gong. GDPN hits a gong a lot. It's probably - They probably got too excited. And they were just like, wait, we sold, we sold 160 million bucks today.

8:01Let's ring the gong. Mr. Gong will unfortunately not be hitting any gongs soon because he might be going to jail. So we initially were selling the nerfed one. Then now with this new Trump rule, you no longer have to sell the nerfed version. You can just sell the real American version, which of course is great for Nvidia. But that's because Blackwells are actually coming out. Blackwells are shipping now. And so we're still, like the US policy is still keeping China one iteration behind. And maybe that will be the ongoing decision here, the ongoing strategy. I'm not exactly sure. It doesn't like the the administration hasn't really laid out a philosophy of this at all.

8:41I mean, originally, what was it? Lutnik said like we're not giving them the best we have. We're not giving them the second best we have. We're not even giving them the third best we have. We're giving them the fourth best. But now it's like we really are giving them the second best we have. So I don't know. That seems like a shift in policy, but we shall see. What's interesting is that now that Nvidia is approved for export, I'm interested to understand where will sales to China land? They've said if there were no geopolitical considerations, we could sell five billion a quarter, 20 billion a year, which is not a drop in the bucket, but it's not exactly half of their revenue.

9:16I think they did 60 billion last quarter or something like that. You're looking at like a$200 billion company. This is like 10 % of their revenue is going to be in China next year if they really take the handcuffs off. The big question is like how much smuggling was actually happening. like we know 160 million was happening because we caught them. But was there a billion happening? Because you don't catch 100 % of what's going on, right? So in theory, there could be like a couple billion, maybe 10 billion that was happening. What is demand actually for NVIDIA GPUs in China? I would imagine it's higher, but then of course you have the dynamic of the CCP.

9:49Yeah, it seems like netting at around 10 % of global sales makes sense given they're projecting 210 billion, or that sort of consensus for next year and nearly 300 billion for 2027. I would expect something more like 30, 40, 50 % of China's really trying to like keep up and win the race and this is like the best supplier. There's a lot of debate over this particular deal, the decision to sell H200s to China for 25 % tax. There's some debate over like whether this should be some sort of like, you know, big picture trade deal where we sort out the rare earth element picture in Taiwan. And we also work through the soybean issue.

10:33And we work through a whole bunch of different export things all at once instead of just doing one off, one deal after another in isolation. I'm not that nervous about that. I don't really feel like I really need it all to be resolved once. I feel like it's fine to have this happen and then watch the consequences of that and see how the chips stack up and how valuable is this. I don't know that it has to all be done within one deal. But regardless of whether America wins from this or China wins from this, it's clearly that NVIDIA wins and also that smugglers lose because the smugglers are going to jail.

11:11The Warner deal will take a while. M &A, AI data, trillion dollar IPOs, and good TV is bad for stocks. We're really doing Bloomberg debt. Warner wars. It's perhaps worth saying that Paramount's offer for Warner Brothers Discovery, Inc. is not exactly a classic hostile tender offer. Warner has signed a merger agreement with Netflix in which Netflix would buy most of Warner for about$27 per share in cash and stock, leaving Warner shareholders with a bit of the company worth somewhere between$1 and$5 per share. That merger will take a long time to complete. For one thing, Warner shareholders have to vote on it, which means that Netflix and Warner need to put together a proxy statement and prospectus for the deal.

11:49file it with the SEC and hold a shareholder vote that could take months. For another bigger thing, the U.S. Department of Justice will need to review the deal for antitrust concerns. Those concerns are significant, and Netflix and Warner have budgeted at least a year for that review. That deal was announced last Friday, and on Monday, Paramount jumped in with an all-cash,$30 per share offer to buy all of Warner, which we discussed on Monday. It took the offer directly to Warner shareholders. It launched a tender offer scheduled to expire on Jan 8th to buy those shares the two deals operate on different timelines warner plans to ask its shareholders to vote on the netflix deal but that vote will happen long after jan 8 and if paramount buys all the shares before the vote then the question is moot because they'll own all the shares they'll vote no if the shareholders all sell their shares to paramount in january they can't vote on the netflix deal in march and if 51 of the shareholders sell to paramount paramount will control warner vote down the netflix deal and acquire the company itself that is the classic benefit of the hostile tender offer.

12:44It's fast. If Paramount's tender is more appealing than Netflix's merger, then Warner's shareholders will sell their shares to Paramount before the Netflix vote and Paramount will win. And if the deals are roughly equally appealing, if shareholders are more or less indifferent between the two bids, then the speed of the tender offer is a real advantage. The headline here is that this deal is going to take a long time. And here he's saying it's fast. So let's see where he gets it. This is interesting. I'm hooked. Then the speed of the tender offer is a real advantage. Shareholders think, I don't care too much, but if I tender to Paramount, I'll get this done faster, so I might as well tender, and Paramount wins.

13:17That is a classic theory, but it is hard to achieve in practice, and it's not really true here. Paramount's offer says at the top that it expires on Jan 8, but it's not like it will buy the shares on Jan 9. Even if 51 % or, for that matter, 100 % of Warner shareholders tender into Paramount's offer, the offer will not close until two other conditions are met. One, Paramount's deal also requires antitrust clearance. It is not legally allowed to buy the Warner shares until it gets that clearance. For a combination of fundamental and Trumpy reasons, Paramount thinks it will have a much easier time getting antitrust clearance than Netflix would.

13:53Yeah, Paramount's proposal expects to receive regulatory approvals likely within 12 months. Faster than Netflix's expected timeline, but a lot slower than Jan 8. Two, Paramount's deal is by its terms conditional on becoming a friendly deal. The condition of its offer is that Warner Brothers shall have entered a definitive merger agreement with Paramount and the purchaser substantially in the form of the merger agreement submitted by Paramount to Warner Bros. on December 4th, 2025. That is, Paramount does not just want to buy a majority of Warner stock and block the Netflix deal that way. It wants Warner's board to abandon the Netflix deal, pay that$2.8 billion reverse termination fee, and sign a deal with Paramount instead.

14:34Interesting. The deal is too big and the antitrust approvals and financing are too complicated to do as a purely hostile deal. Paramount will need Warner's help. In some sense, then the Paramount deal is not a real tender offer, one that depends only on the shareholders, one that shareholders can accept even over the board's objections. The Paramount deal is a pressure tactic, a way to get shareholders thinking, hey, I would rather get$30 in a year than$27.75 in 18 months and telling the board that. If the shareholders all would tender into the Paramount deal, then it's hard for the board to stick with the Netflix deal as a matter of fiduciary duties and shareholder pressure.

15:11But it's not like if those shareholders all do tender into the Paramount deal, it will just close. It's a jumping off point for negotiations. Paramount and Netflix are girding for a battle they predict will stretch well into 2026. Lucas Shaw says Warner Brothers was given 10 business days to respond to Paramount's hostile$30 a share bid for the company on Monday since that offer was already rejected once. The Warner Brothers board isn't planning to cancel the merger agreement signed last week with Netflix, according to people familiar with the company's thinking. Doing so would require Warner Brothers to pay Netflix a$2.8 billion termination fee.

15:41That puts the onus on Paramount to make the next move in what everyone expects to be a drawn-out affair lasting months. Paramount can follow through on its tender offer to buy Warner Brothers shares from investors at$30 each on Jan 8. It can also extend the bid, sue to stop the Netflix deal or increase the terms. Shareholders of Warner Brothers, one of Hollywood's biggest film and TV companies, are hoping for a bidding war that further boosts the price of the deal. Both companies have communicated that they have the ability to increase their offers. Paramount is privately weighing an increase or whether to instead add sweeteners intended to give WBD's board greater confidence in its regulatory prospects versus Netflix.

16:17This position that the Ellisons are in where it feels like every other week they have to offer a higher price, it's kind of a vicious cycle. It's pretty crazy to have them going back and forth here. I feel like it's going to land with Netflix. It just feels like the breakup fee is real. The fact that there's a massive multi-billion breakup fee on both sides, that's material to their market cap. Material to their... You have to pay essentially 1%, 2 % of your market cap in the event of a breakup. up, that feels like Warner Brothers and Netflix were like, let's actually make this happen. Let's be really sure this is going to happen.

17:02I don't know. Then again, like the Adobe Figma thing, I was shocked when that broke up because that felt like a crazy breakup fee. But$1 billion was still lower than what Adobe was trading at at the point. It was less than 1%. This is higher on a percentage of market cap. Meta's new AI superstars are chafing against the rest of the company. It's Chafegate. We've heard about this. We've established this. It's an us versus them mentality has emerged between Meta's top artificial intelligence team and longtime lieutenants to Mark Zuckerberg. Yes. Eli writes, when Mark Zuckerberg revamped Meta's AI operations this year, he recruited a new leader.

17:37Former guest of the show, Alexander Wang. Let's go. A 28-year-old entrepreneur to build a team of top researchers from rivals like OpenAI and Google. That team called TBD Lab for to be determined was placed in a siloed space next to Mr. Zuckerberg's office at the center of Meta's headquarters, surrounded by glass panels and sequoia trees. Mr. Zuckerberg wanted to separate the new AI group from the bureaucracy of the company, which owns Facebook, Instagram and WhatsApp. In meetings this fall, Mr. Wang has privately told people that he disagreed with some of Mr. Zuckerberg's longtime lieutenants.

18:10You're going to be interested in this, John. So in one case, Mr. Cox and Mr. Bosworth wanted Mr. Wang's team to concentrate on using Instagram and Facebook data to help train Meta's new foundation model to improve the company's social media feeds and advertising business. You've been wanting this this whole time. It's like do this whole thing, personal super intelligence, and then just make the algorithm better, make the ads better, grow the business in that way. So what did Mr. Wang say? He argued that the goal should be to catch up to rival AI models from OpenAI and Google before focusing on products, the people said.

18:45So this makes sense, right? This makes sense for Wang. I don't know that it makes sense for the company. From everything that we've seen so far, it feels like Meta could add a trillion dollars to its market cap by just focusing all this incredible talent on just making the core business better. but Wang and the rest of the team are going to be a lot less motivated by that than having the biggest data centers, doing the biggest training runs, having the best model, competing on the global stage. And so I can see why Wang is pushing the other direction. Okay. Let's actually unpack this because like, if, you know, Mr.

19:27Wang here, he says he's developing the model. He argues that the goal should be to catch up to rival AI models from OpenAI, like if they wind up having a model that is, you know, actually be in the same league as Claude and Gemini and OpenAI and XAI, and it's this like, you know, closed source, it's an API, it does well on MMLU, it does really well on the benchmarks. It's like, what value does that really bring? Like if you want to compete in search, knowledge retrieval, even agentic commerce, you can take models off the shelf in the same way that Apple's doing. So I'm just trying to think, you make the best model in the world.

20:08Are you going to go compete with, so we've seen this, we've run this with XAI, right? They are trying to build the best model in the world. And it's now becoming clear that just having a great model does not automatically give you a meaningful amount of market share. Yeah, no, no, you need to productize it. And OpenAI has productized very well with a viral like it is synonymous with AI consumer AI everyone has the app mostly on their home row they're using it a ton they have a billion DAUs it's growing you know doing great or MAUs then you have Google just so much surface area to actually stuff knowledge retrieval AI in knowledge retrieval products they are you're already trying to do that with Google search anthropic super focused on code super focused on on on B2B Facebook doesn't have a B2B team.

20:57Like they're not like a hyperscaler like that. Might make sense for Amazon to be like, hey, we want to get, you know, we want to take that approach. And then XAI is sort of the same thing. At least they have, you know, X to distribute through. But being the fifth, the fifth like hottest, even if the model is for this week, like the number one, I don't know, I would be very focused on what can be done within Instagram and Facebook and WhatsApp to a lesser extent. The reason to be excited about that as a Meta shareholder is that AI is such an obvious tailwind for Meta's business. Meta's been pushing this personal super intelligence narrative.

21:40The only problem with that is that I don't know what that means, right? It sounds awesome, and I'm happy to wait and see. but they've been having, you know, there's been a number of employees kind of churning out already and it's possible that they don't necessarily know exactly what that means themselves yet, right? They're running it like a startup internally. Yeah. What does personal super intelligence mean? I feel like there is room for product-led innovation in AI, new instantiations of the underlying, like new ways to interact with the fundamental, like, you know, we're having this AI moment, and then we get video models, we get image models, we get image editing, we get knowledge retrieval, we get agents, deep research, we get coding agents.

22:32Like, we've had three or four or five, like, really cool instantiations of it, and not every lab is frontier at every single instantiation. Anthropic doesn't have an image generator, right? Other labs, like, they're, you know, Obviously, OpenAI really thrives in the consumer and has created just a great app that is reliable and answers your questions reliably. Other companies have struggled to hit that instantiation. It would be very, very cool to watch the Facebook team figure out what is a way that they can bring AI to bear inside of Instagram in a cool way, in a new way. The problem is that Facebook and Instagram don't really have that DNA.

23:13They sort of tried it with the Vibes app, but also there's very low ROI on that because as soon as you... Let's say that they do come up with the next stories, and it's like, oh, wow. You take AI, and you stuff it in a social app, and you do this one special thing, and then everyone loves it. It's great. It's not just like a tap-off. Well, they didn't create that. I know. I know. What are you saying? But let's say that they did. It's going to get copied everywhere anyway, so it's not really that much... They really should just wait around for everyone else to do their R &D. They should wait. They should be...

23:42Evan Spiegel, what are you doing? Come on, invent something for us. And so I can actually see Alex Wang's pushback on like, hey, you want to use Instagram and Facebook? You want to do something more incremental? But like, how are we going to do something great in AI in those ecosystems? I don't know that there is that much. Part of what maybe we haven't discussed so far is how TBD will tie into Reality Lab, right? Is it possible that they need in order to fully realize the Reality Lab's vision of having a pair of glasses that see and process everything that you see and can provide you that personal super intelligence?

24:24So you're walking around, it's like, oh, I forget that person. You know, it just pops up like a name tag for somebody. Right. You see an item in the real world and you're like, you know, you do this and you buy it. Right. It's very possible that they need to eventually get that on device in order to be fast enough, in order to be super valuable. And so they do need this internal competency. There's, of course, stuff that we're missing here. But ultimately, I can see why Boz and Chris Cox were like, hey, why don't we just figure out how to make an extra$50 billion a year by taking some of the best researchers in the world and applying them to the core business?

25:03Gyms are coming to airports. This is big. This is big. Will you work out on your next plane? Consumers have been saying the other passengers aren't sweaty enough. Yeah. So this is from the - We got to have gyms in right next to the gate. This is hilarious because it seems to be a collab between Secretary Kennedy, who's the Health and Human Services Secretary. So there's HHS crossing over with DOT somehow. And they're like, let's make the airports healthier, which is like a funny thing. I feel like these collabs don't happen before, but they're happening now. The idea of working on an airport, it's kind of crazy because if you are all sweaty and then you have to.

25:47And they also said like you should dress up when you go to the airport. And if you're dressed up and then you are all sweaty, that seems pretty rough. But I don't know. I like the idea of doing something new in airports. I think it's cool. I think it's good that there's some opportunity for some sort of grant program. I guess the question is the$1 billion grant program with how much it costs to make, like, is this, like, 10 gyms? Well, so I don't know because, like, there's a world where you build, like, a proper gym in the 10 most premier airports that has, like, a sauna, showers, like, full laundry.