Pinterest CEO on AI Roadmap, China’s Emergency Nvidia Meeting, AWS’ New AI Model | Dec 10, 2025

25:32Right now, we're making great recommendations of things you'd want to buy, helping you go take action on those things. How much more helpful can that be? That is a moonshot in and of itself in terms of like can that get to a place where it's as good as the person you'd love going shopping with the most, whether that's a best friend, a sister, a brother, or if you're lucky enough to have a personal stylist. Can it be as helpful as that? We think that in and of itself is a moonshot and doesn't require a different form factor or things like that. And the last thing I'd say, so many moonshots end up either not panning out or come to fruition much later.

26:12And I've built five startups from zero, Venmo and Brains for being the two most recent. And I've been in Silicon Valley for a long time. And one of the things that, you know, two things you'll hear in Silicon Valley. One is like, well, I was right, but early. The other thing that any good VC would say in response is right, but early is the same as wrong. And so these moonshots, you've got to have a vision for where you want to go, but you've also got to be deeply tuned in to like what users are ready for today and what you can deliver today. I think we're balancing that really well, where we have a view of like, okay, shopping and human curation are at the core of our platform.

26:52That human curation is a real differentiator for us. We think over time that can let us do tremendous things to be as helpful as going shopping with your best friend or the really great sales associate at the boutique or whatever. We think that's where that can go over time. What are users ready for today? Well, today, you know, it's really great recommendations, really, you know, seamless ability to go purchase right inside of our platform. You can link an Amazon account and purchase right inside of our platform. Those kinds of things that really make their journeys easy today. Right. And then we'll work with our users.

27:23So we certainly have that long-term vision, but also really laser focus on what are users ready for now? And that's why we don't use words like agentic and things like that when we're talking to our users because as much as we preserve the human experience. Right. Yeah. Yeah, as much as we talk about agentic and Silicon Valley, the average consumer isn't thinking about agentic at all. They're just thinking about, did this thing help me do what I wanted to do? And on that, I think we're getting really high marks as evidenced by users flocking to the platform and engaging deeper and deeper over the last three years.

27:56Great. Well, Bill, it's great to have you on the show. I really appreciate the conversation. And there's a lot of exciting quarters to come, I'm sure, for the company. It's been a great run under you. And so I look forward to seeing how you lead the company from here. Thank you so much, Akash. Really appreciate it. That was Bill Reddy, the CEO of Pinterest here on TI TV. Okay. The Trump administration has given the green light on H200 sales to China. But now China is trying to figure out its stance. And the information reported China is trying to assess how much demand there is domestically for the chip.

28:32Joining me now to break it all down is Jing Yang, our Asia Bureau Chief. Jing, welcome back to the show. It's great to have you here. Always glad to be back, Akash. So tell us about these emergency meetings that have been happening in China. Yeah, so on Wednesday, well, today, Chinese government officials called a series of emergency meetings with Chinese tech giants, including Alibaba, ByteDance, and Tencent, to ask them how much of the NVIDIA H200 they actually need. And the reason that they caught these meetings is because that China, on one hand, wants to, you know, is keenly aware that China needs these powerful NVIDIA chips to develop the country's AI systems.

29:17But on the other hand, they also want to make China more self-reliant in semiconductors. So what Trump has made as some sort of a peace offering, seen as the Chinese side, also complicated somehow Beijing's policy goes. Now, I imagine there's a lot of demand for the H-200 chip, given how much more powerful it is than the H-20. Of course, we know it's not as good as the Blackwell. I wonder, though, what these meetings say about the direction in which the government is looking at AI. Is it different from the approach they took with H20? Yes and no. So basically, in today's meetings that we learned that the Chinese officials told the tech companies that they wanted the companies to go back and give them a very detailed assessment and justify their demand for the H200.

30:19Like, for example, you have to specify why you need this many of H200 that Chinese domestic chips just cannot replace. And then the companies were told that once officials have compiled and then looked through all these responses from the companies, then they will make a decision. By the way, once the Chinese government makes the decision, it's not going to tell the world about it. They are very likely going to use this so-called window guidance method. which is like a very common but very powerful method for Chinese regulators to privately tell companies their policy expectations instead of making it public.

31:00That, by the way, was exactly the same approach that Chinese government used a few months ago as we reported to ask Chinese companies not to purchase any of the NVIDIA H20 chips. Right. Well, I want to pivot to talking about DeepSeek. The Information published a story about how the company is pushing ahead with its own model development. Tell us about that story. Yeah, I mean, since DeepSeek rose to global stardom in January, I don't think we have learned much about what actually is going on in the company, right? And that just, you know, is such a mystery. And we spent all of this year trying to figure out what is the company's next big move.

31:43And finally, we were able to report today that it turns out that DeepSeq has been able to use black whale chips, by the way, which are banned from being exported to China by the U.S. government. But somehow they got their hands on these chips from, you know, the smuggling channel. And they needed these chips because they are racing to build their next flagship AI model so they can continue to stay competitive in the AI race in China and globally. Now, I wonder, connecting the two stories together, do you think that that story changes at all in the months to come in the face of the new H200 chip?

32:25Does the policies around the H200, could that change the different types of chips that DeepSeq tries to use? Yeah, I mean, not just DeepSeq, but all the Chinese companies that are trying to develop their own AI models to stay competitive in China and globally against U.S. companies as well. I mean, what I'll say right now is that it will definitely have an impact on the demand. And let's say if the Chinese government actually allows H200 to get into the country, it will definitely have an impact on the black market demand for black world trips. We just, it's just simply too early to tell to what extent it will be impacted.

33:10But it's, you know, it's a very, you know, fluid story is developing and all these Chinese tech companies and Chinese government are assessing. sort of the right combination, right? You know, let's say from Chinese domestic chips to H200 to the sort of contraband Blackwell chips. Let me ask you one last question before I let you go. We've been talking on the show about this issue, and Jessica Lesson, our editor-in-chief, was on a couple days ago talking about how, in the long term, her prediction is that the technology is kind of going to even out. And I know talking, and she was talking about the models specifically, And I know models are a slightly different battlefield than the chips.

33:53But one of the things that I've been thinking about is, look, whether it's the U.S. versus China, I do think that three years from now, I think the models are kind of going to even out. I mean, chips, I even think that the chips, the strength of them will even out. Is that your prediction as well? Do you agree with that? Or do you think that the story might go differently? I think this is still a little early to tell, to be honest, right? China, you know, with DeepSeq, Anibus, Kuan, and some other smaller startups have demonstrated, you know, awarding class capabilities in open source models. However, we know that open source approach usually is adopted by the second runner, right?

34:38The best players usually want to protect their technology, want to go prioritary, right? So the fact that China actually has not had a proprietary model that is being used by companies or developers outside of China shows that China still has a lot of catch-up to do. And there may be some geopolitical hindrance in the background, but by and large, the chip shortage can also not be ignored. I would say this, right? And if in the next few years, China can actually make the real breakthrough in chip design and chip manufacturing, then yes, then we can call that China will be on power even sometime, some years to overtake the U.S.

35:29But, you know, we just cannot discuss the AI race without looking at the chip element. Right. Well, that's why it's such an important story to cover. Thank you, Jing, for coming on. We really appreciate it. and I look forward to seeing the reporting coming out of your bureau in the weeks to come. That is Jing Yang, our Asia Bureau Chief here at The Information. When it comes to financing for big data centers, we tend to think about the big tech companies in North America that are spending heavily on capital expenditures. But my colleagues today published a deep dive on why MGX, a fund based in the United Arab Emirates, has become another major financier for these projects.

36:11Here to tell us more about the big role they've grown to play is Myles Krupa, our AI and finance reporter. Myles, welcome back to the show. It's great to have you here. Thanks, Akash. So who is MGX? Let's start there. Yeah, it's kind of the new kid on the block. It was started about a year ago by the UAE, and it's a joint venture basically between their sovereign wealth fund called Mubadala and this AI company called G42. And so they've both put in money to this fund and it's effectively pouring billions into data centers, chips, OpenAI, XAI. It's sort of investing all across the AI spectrum really rapidly.

36:58And so in terms of where this money comes from, you said the sovereign wealth fund is one source of the capital. G42, where does their money come from? Yeah, well, G42 was originally set up by the UAE, but it also has outside backers. It has Silverlake, which is a big tech-focused private equity firm here in the US. Ray Dalio, the former leader of the hedge fund Bridgewater, is also an investor. So you see this kind of cross-pollination happening. Okay, so they have a lot of money to deploy. You also wrote about how they've become a giant in the data center financing game. Tell us about some of the projects that MGX is backing.

37:40Yeah, they're doing sort of two things. One, they're investing big in data center companies. So what we revealed in this story is basically that they're going to be the largest shareholder in this company called Aligned Data Centers that struck a record$40 billion deal this year to be sold to a consortium that includes MGX, among other investors. So they're investing in companies like Aligned, Vantage Data Centers is another one. And then they're going to places like France and Italy and doing joint ventures with local companies to build data centers there, really huge ones of a gigawatt or more.

38:22So they're doing a few different things, but suffice to say, they're trying to build a really huge data center portfolio that's also global, not just in the UAE. And so, you know, one of the questions I wanted to ask you about is the U.S. government has been so sensitive right now to keeping AI domestic. I wonder as it relates to at least data center projects in the U.S., I mean, has there been any sort of increased scrutiny around where the money comes from for these data center projects and whether or not it's coming from outside of North America? Yeah, it's possible that this aligned acquisition will trigger what they call a CFIUS review to look at basically whether MGX would have control over aligned.

39:09That's a normal thing to do when any international investor comes in big into a U.S. company. but um the stance from the administration seems to basically be that they're welcoming money of any kind to build data centers um and ai infrastructure in the u.s um as long as it's green um you know uae has pledged 1.4 trillion dollars of spending in the u.s over the next decade um saudi arabia just pledged a trillion itself um i don't know that we would have seen this under the Biden administration, but the Trump administration certainly is welcoming that money. And I wonder, you know, as it relates to these data center projects, we obviously know that the big public tech companies are involved in those efforts.

39:57We also know that the fast-growing startups, in quotes, you know, the giants, OpenAI, and Anthropic, they are certainly becoming part of the story. And look, they have their own fundraising needs, which, you know, that's one way that companies like MGX could get involved. Could MGX also get involved in their data center projects a bit? You know, could the relationship at all get closer over the coming months? Yeah, these investments that they're making in Aligned and Vantage in particular are going to draw them a lot closer to both the big tech companies and open AI. So Aligned works very closely with some of the largest cloud companies, you know, think Amazon, Microsoft, Google, to build data centers for them.

40:45So that will sort of deepen the ties between the UAE and those companies. And then Vantage is actually really closely linked to open AI and Oracle through the Stargate project. Vantage is building at least three sites in Texas, Wisconsin, and Michigan for Oracle to provide open AI with cloud services. So, you know, it's really kind of getting closer to these important AI companies through data center investments. Let me ask you a bit of a more, you know, broader question, your own reflections on the data center space. Look, we have all these companies. We've talked about some companies like Crusoe, for example, on the show.

41:33We haven't talked too much about Aligned and certainly about Vantage as much. You know, you have studied these data center companies and the ways in which they're growing, some of the risks that obviously we've talked about the debt. I haven't even really paid attention to the founders, the people who are building the company on the ground. I wonder if you have any broader reflections on the people building these businesses. I mean, And do they have sort of an affinity for these really large-scale capital-expensive businesses? Are they sort of a different type of entrepreneur? Just talk to me a little bit about what you've learned about these companies.

42:10Yeah, well, fundamentally, data centers are real estate. So I think it's important to think of these people as really real estate investors and real estate builders. And what's sort of remarkable when you talk to them is how quickly they're having to adapt to the needs of these tech companies that suddenly want a gigawatt of power at their data centers, whereas before 100 megawatts would have been sufficient. You know, a lot of these companies are owned by private equity funds after being taken private because they weren't very attractive publicly traded companies. They had a lot of debt. You know, public investors couldn't quite understand them.

42:52And so what we're seeing now, I think, is these companies trying to reposition with this new growth coming from these huge, huge AI demands to try to find a way to either go public or like aligned, be acquired at a really nice multiple and sort of take advantage of this AI boom in the way that any sort of startup founder would. You know, they're having to sort of go into hyper growth after being in kind of a normal growth phase, if you will, for many years. It is kind of funny to me that as we talk about debt, I mean, look, and I'm not equating this bubble with the 2008 bubble by any means. But look, that was fundamentally at some points about real estate and about mortgages.

43:39You know, here we have a similar group of people that are getting involved in these data center efforts. And look, we're not calling the top of the bubble. We're not saying it's going to be as catastrophic, but real estate, I mean, gosh, it's kind of what everything somehow comes back to at the end of the day. Real estate is hard. You know, it may be in the end that it's supply chain issues, power procurement issues that put a cap on the bubble or pop the bubble, however you want to sort of view it. Yeah. Yeah, in the end, it's real estate. And as you said, it's the debt because this stuff is all very expensive and the funding mix is going more and more towards debt over time.

44:23Great. Well, Miles, thanks for coming on the show. It's not a company that we have talked about a lot yet, but I anticipate we're going to have more news to come. And so we'll bring you back on when there is. That is Miles Krupa, our AI and finance reporter here at The Information. Okay, our next segment is with our presenting partner, Amazon Web Services. Models have become a big battlefield for AI companies as they try to one-up each other with new releases every few months. Amazon has been working on its Nova model and made some new releases recently. And today, I want to look at how the company thinks about making its product suite competitive there.

44:59Joining me now is Sean Nandy, a director at AWS. Sean, welcome back to the show. It's great to have you here. Akash, it's so good to see you again. I am just back from a week in Vegas and adjusting to all this cold in New York City. Well, and it is cold. I'll tell you that, Sean. It's certainly taking us by surprise here. I want to talk about some of the announcements that Amazon and AWS made at reInvent. You know, I was really excited about the Nova models that the company talked about. Tell us about what's new in that family of products. Yeah, look, let me pull back for a second. Last week was our annual conference, reInvent.

45:39We had 60 ,000 plus attendees. It was really amazing. I think it was the 13th reInvent. I hope I have that right. And I've been to 10 of them. So it was my 10th. I went as a customer for half of them and now as an employee. And I'll tell you, there were a lot of announcements. And that's something we're proud of. We love to innovate fast. Now you asked about models specifically. And I'll tell you the announcements around models came in three sort of categories in my head, at least. This is Shown's analysis. First off, we announced the largest release of new models on Bedrock, which is our key platform.

46:15And the reason that's relevant is we had new model providers, like we had the open source models from Google, the Gemma models. We had Mistral's new family of models. We had just a broad set in different types and shapes, and we can talk about why that's actually important. Second, of course, we launched our new Nova 2 models, which you mentioned. You asked what was different. I will get to that. And third, we announced a great capability called Nova Forge around allowing customers to meld their data with our Nova models. Talk more about that. But just more specifically on what's new, we enhanced our speech-to-speech models.

46:52Super exciting. we launched the first reasoning model, Nova 2 Omni, that is multimodal. So it can take all kinds of input and have all kinds of output. And that's really important for advancement in the case of multimodal models. Now, one of the product announcements that I was interested in is the growth in the family of custom models and the tools that AWS offers. You know, why are custom models such a big focus for the company? Yeah, look, let me give a little history. I won't hopefully bore the audience too much with too much history, but it's all sort of recent, right? So in 2022 and 2023, I think a lot of customers I talked to, enterprises, core startups, were like, we got to build our own foundation model.

47:39Everyone's like, we got to be differentiated. We don't want to look like everyone else. And we all realized sort of the industry, building your own foundation model in 22, 23 took massive expertise, loss of cost. It was only viable for the largest players. That's why you saw these frontier model companies be so successful, including us, right? In 23 and 24, we heard all about RAG, retrieval augmented generation, other techniques to bring your data into models, but not really change them. And now in 25, we're seeing customers ask again, the frontier model advancement is slowing a bit. I don't mean slowing in terms of you see great stuff happening every couple months.

48:19But in terms of relevancy, if you're going a couple extra points of accuracy, does that really... And we've seen it too, is that the advances are starting to plateau a little bit. Yeah, I'm sure there'll be something that'll break that plateau. But for most use cases, customers like, has something I can do fundamentally changed my business? And so they're asking again about how am I different? what gives me an advantage? And it comes back to their data, Akash, right? Like data is a company's advantage and the companies with more data, they are in such a driver's seat for what they can do. So instead of saying what model we should use, they're starting to say, how do we unlock our data?

48:57Now, to answer your question that you started with, what we released at reInvent was a series of capabilities across our stack. And I'll focus on one for a second that let customers use your data more effectively. And Nova Forge and the rise of what we're calling open training models is a capability where customers can meld their data with a curated Amazon data set and effectively retrain the Nova model to be their own frontier model. That's sort of the Mexico model. And, you know, we've done it so that you don't have to have a bunch of data scientists and engineers and large numbers of GPUs, like starts at a much lower price point.

49:34And we'll see how customers react to this. I'm bullish that it'll become quite a big thing. Now, we've also reduced capabilities in Bedrock to allow easier fine-tuning. We've done a lot in space. But the net is we're making it cheaper and easier to customize. And customization, we think, is a massive competitive advantage. Let me ask you this. If you look at the process for developing these models, we've written about this recently at the Information. I mean, if you look at the phases, there's pre-training, there's evaluation, there's post-training, there's the launch, obviously. But if you look at these faces of developing models, I wonder where you think is the biggest competition right now.

50:16What is the biggest battlefield? What's the hardest thing to innovate in? And how are you seeing that? Yeah, look, I think that we have a lot of large companies. Maybe they're not large, but they're emerging companies that are really evolving how to build an extra-trior model, bringing the right research, the right capabilities, our own AGI team, AWS or Amazon included. But the thing I talked about with customization, there's more and more interest in pre-training. And I'll give you an example of like what you probably have to write about all the time, which is benchmarking. And every time a new model comes out, the company announces all the benchmarks.

50:52We do too, right? Because people want some sort of quantifiable, how different is this model? I personally think one thing you're going to see change is customers are going to start to say, I don't want to see your benchmarking model provider or hyperscaler. I want to run my own benchmarking with my own use cases, and I want to see how it performs. And that's part of where things like continuous pre-training, using your own data, are becoming more relevant. And we'd like to see that pre-training be the providence of customers themselves versus us, the hyperscaler, or versus a frontier model provider.

51:26And that's what we're trying to do with Noah Forge, put pre-training into the hands of customers. And we're going to see how that progresses, right? I mean, we had great early feedback from Reddit. They're one of the customers who helped us work on NovaForge as a customer. They were trying to improve content moderation and really bring in their data. And they were able to reduce a bunch of specialized ML workflows with just sort of one cohesive approach using NovaForge. But it's just one example. We're first with it. I'm sure others will innovate in this space and we'll see how it changes things.

51:54Right. Last question for you. we are coming up on the end of the year in 2026. It's now time for people to start making predictions about how the narratives around AI might change, what might be the hottest topics. I wonder, as it relates to the discussion around models, this year, if we reflect, I mean, there was a lot of talk about margins, there was a lot of talk about pre-training, a lot of talk about the benchmarks. How do you think the discussion might change in 2026? What sorts of predictions do you have? So you're asking for Shown's opinion, not AWS. I'll give you my opinion. And I'll pick a couple areas and hopefully they're not too boring.

52:34They're interesting. First off, you know, a couple of years ago, we talked a lot about model choice. And people, customers were like, great, but I just want to know the answer. Like which model's best, which is cheapest, which is fastest, which is safest. I think the rise of agent take over the next year or two will bring that model choice discussion to the forefront. customers are going to want and they're going to use ai to help them do it really curated model selection for their use cases so you're going to see many more models that are like targeted individually that's part a part b is this customization thing i do believe customization is going to be a big thing because it's becoming easier and when we talk about roi enterprises care about roi right we've got lots of questions on why a year ago not enough stuff went from proof of concept production, customers didn't necessarily feel the ROI, the investment's going down.

53:25You can do custom models cheaper. You can run models cheaper. You and I talked about this a few weeks ago where I said, I think the cost of inference will drop by 90%. As all these costs drop, as inference costs drop, as model customization costs radically drop, as you have more open-weight models that can be had very cheaply, you're going to see organization say, I want a tailor for me because the ROI doesn't have to be as, well, the ROI will be big, the investment is not as big. So the return doesn't have to justify that personalization. And that's what is every company becomes some type of AI company.

53:59By the way, I'm not saying everyone's going to become NVIDIA or us or meta, but as AI becomes endemic in every company, it's like internet became in every company. I think that personalization layer for that company will become so important using their data. That's the future. Great. Well, Sean, it's great to have you on. I appreciate you making time and we'll talk to you again very soon. Absolutely. I'm always happy to talk to you and have great holiday season. If I don't talk to you before, then it's coming up on us. We'll talk to you soon. Okay. Thanks, Chad. I appreciate it. Well, that does it for today's show.

54:30A reminder, we are on this stream Monday through Friday at 10 a.m. Pacific, 1 p.m. Eastern. I want to thank Amazon Web Services, who is our presenting sponsor for this production. And I want to thank you for tuning in. We really do appreciate your viewership. I'm already excited for our next show tomorrow. Have a great rest of your Wednesday. Bye-bye for now.