China Breaking Free from Nvidia, AWS re:Invent 2025, 2025 Gift Guide | Dec 1, 2025

0:14Welcome, everyone, to the information's TI TV. My name is Akash Pasricha. It is Monday, December 1st. We have got an all-star lineup of reporters and guests for you today. First up, we are digging into today's feature story on AWS with our enterprise software reporter. We're then turning to a conversation on the shifting landscape in the US-China chip war. And finally, because it is officially the holiday season, we are shifting gears from policy and infrastructure to the products, shaping our everyday lives with a spotlight on our 2025 holiday gift guide. Today, we're bringing you three founders behind the standout gifts on our list.

0:51They span board games, flight apps, and apparel. It's going to be a busy, busy show, and so let's get right on into things. It's a busy week for tech events, and Amazon Web Services' reInvent Conference is taking the stage in Las Vegas. Our very own Kevin McLaughlin, who covers enterprise software in his Sunday briefing, called it the Super Bowl of Cloud Computing. And he has a new story out today with details on what to expect from AWS. I want to bring him on to tell us more about what he is thinking about. Kevin, welcome to TITV. It's great to have you back on. Thanks for having me. So let's talk about reInvent.

1:25What can we expect this week from the conference? Well, it really is kind of the Super Bowl of cloud computing. I mean, there's no halftime show, so that's the only difference. but it is a every software vendor startup comes to pitch their wares on the expo hall floor in Las Vegas. It's also a chance for a lot of customer conversations to happen as well. So it's a big sales event. It's a customer conference. And it's also where Amazon announces their new, their big new stuff for the year, which these days is all about AI. And so what are your sources telling you that they are going to debut this week?

2:07We reported this morning there is a new Nova model potentially being talked about this week. The big difference, there was a new Nova model last year, and that was where they launched the Nova line. But the big difference this year is that this is an all-in-one model that can handle and process video, images, speech, and text. And what that means is that it's more of a direct competitor to models like Google Gemini and OpenAI. So this is definitely a big step up from what Amazon had before. It remains to be seen, though, whether how customers are using it or how many customers will actually adopt it.

2:48And I want to talk about the trajectory of AWS here in a second, but we've had AWS executives on the show before. They talked a lot about Bedrock. What's the difference here between Nova and Bedrock, just so we understand? Well, Nova is sold through Bedrock as part of a collection of other models from third parties, especially Anthropic, which is one of the main ones they lead with, but also Meta, Cohere, and several others. So Bedrock is the main AWS AI service. Got it. Got it. And Nova fits into that. Okay. Well, I want to talk a little bit about AWS's trajectory with AI overall. I think consensus is that they have lagged and fallen out of the conversation a bit.

3:29Can you walk us a little bit through the history of why that is the case and where they've faced challenges, really? Sure. I think a lot of this perception comes from the fact that AWS is still very much the king of the cloud in terms of market share and reach and basically every other metric. So AI has been the first time where Amazon has been sort of on its back foot. And what I mean by that is in the old world of machine learning, Amazon was, AWS was one of the first to actually offer a cloud service that companies could use to work and build AI applications. So it wasn't like they missed the boat or were late on that front.

4:08But when Chat GPT arrived three years ago, that was when things started to become apparent that, hey, maybe they are not the leaders here. AWS passed on early opportunities to forge close relationships with OpenAI and Anthropic, which could have changed the current state of play today. Of course, Amazon eventually did invest$8 billion in Anthropic and has a very close relationship with them. They do have a story to tell, and it's not like they're completely missing the boat, but the perception is still that they're behind OpenAI and Google and Anthropic. What else are you expecting from reInvent?

4:50You talked about Nova. What about the chips and some of its agents' products? Yeah, so we have, there was a product launched this summer called Bedrock Agent Core, and this was an attempt. Bedrock initially had some agent building tools with it when it launched two years ago, over two years ago, but they were not, my understanding is they were not widely used. So Amazon launched this service this summer called Bedrock Agent Core. And it's basically a way for enterprises to build agents and it's a little bit more sort of easier for them to do so. And the product has gotten a lot of positive buzz in the few months that it's been available.

5:29So we can expect to see customers' testimonials or reinvent, perhaps some new features and things like that. And then on the other hand, we have Tranium, which is Amazon's custom chips, were expected. They talked about Tranium 3, the third version last year at reInvent. So we should expect to see some customer testimonials about how they're using those chips. And I think it's really important for Amazon to talk about Tranium because that is their counterpart to the Google TPU AI chips. And on that note, we've talked a lot on this show last week about the traction that the TPUs are getting for Google.

6:09Do you anticipate that Tranium is going to get the same type of traction? Is there technical differences here that could still put it behind TPUs? Or how do you see that conversation unfolding? It's really too early to say. Amazon just launched Project Rainier, which is a giant supercomputer that they built in conjunction with Anthropic. That includes 500 ,000 Tranium chips when it launched. Right. And Amazon said that it expects Anthropic to scale that usage to more than a million by the end of the year, which is four weeks away. So surely Anthropic is going to be a very big customer of Tranium 3.

6:50But in terms of like the usage and how that's going to scale, I think it's sort of hard to say at this point. I wouldn't want to venture a guess. I do know that the first version of Tranium had some glitches that prevented many customers from using it. Tranium 2 fixed a lot of those issues and no one iterates better than AWS in terms of like working out the kinks with products. So we can expect to see Tranium 3 probably having more customers and perhaps even some surprising new customers. Right. Now you talked about the$8 billion that AWS at Amazon has invested in Anthropic and obviously they are leaning on them heavily for a lot of their experimentation and development with these AI products.

7:32How is that relationship faring overall? I think it's still fine as far as I could tell. Like any relationship, there are bumps in the road, and certainly Amazon and Anthropic have gone through those, just like Microsoft and OpenAI have gone through. I think that my impression is that Amazon is signing deals with Microsoft, and they're expanding their agreement recently with Google Cloud. I think Anthropic, just like OpenAI, has an enormous appetite for compute. And so those relationships and those expansions are probably being driven by that as well. I also, I wonder, though, if Anthropic wanted AWS to feature it more, the way that Microsoft features OpenAI and Azure.

8:17I don't have any reporting to back that up, but it was just kind of a gut feeling that I had in some of the conversations where Amazon is really intent on promoting its own models. the Nova line, which is fine. They invested in those and they want to have a first-party model to show that they're actually in the game in AI. But at the same time, if you're anthropic, you might, maybe you want it to be featured a little bit more prominently. Right. Well, Kevin, thanks for coming on. We really appreciate it. And we'll talk to you again very soon. Thank you. Okay. For years, the story of NVIDIA in China has been defined by a cat and mouse game with Washington, with U.S.

8:55export controls tightening the leash on what American chip technology could and couldn't be sold in the country. But an exclusive new report from the Information reveals that dynamic has shifted. And now China regulators are now the center of focus. Joining me now is Channer Liu, reporter at The Information. Channer, welcome to TITV. It's great to have you here. Thank you for having me. So let's talk about these pair of reports that you published over the weekend and last week. Talk to us a little bit about how China's latest move is faring in terms of how they aim to grow adoption of their chips domestically.

9:31Sure. We have seen the major escalation in China's effort to support local AI chip markets. In August, we reported that Chinese regulator asked tech companies to stop ordering NVIDIA chips. Now they are blocking companies from using the NVIDIA chips they already buy for new AI data centers. For example, Bydance, NVIDIA's biggest Chinese customer this year, is basically sitting on a stockpile of NVIDIA chips it can't use. So this is Beijing forcing a full breakup from the US AI ecosystem. They are pushing companies onto homegrown AI chips from Huawei, Cambricons, Alibaba, and others. But most Chinese chips currently can only handle inference workload, not training, which is a compute-intensive process that actually builds AI models.

10:26So there's a lot there that I want to double-click on. And I think one of the pieces I wanted to get your explanation on here is, can you explain to us domestically in China how much competition there is from those players to NVIDIA? You mentioned Huawei and some of the other companies that are developing their technology. How good is it? How much demand is there for it? Where do they fall short? Talk about that. Sure. The regulatory pressure forcing tech companies off foreign trips like NVIDIA is definitely creating a captive market for local chip companies. So we can start from Huawei. Huawei is a clear leader in China's AI chip sector.

11:11They just unveiled an ambitious three-year roadmap with three new AI chips using local high memory, high bandwidth memories, and a whole new design structure, which is basically confirmed two stories we previously did about Huawei. But no matter how hard Huawei tried, its chips are still weaker than NVIDIA's best offerings in single chip performance. Each Huawei's chips designs, especially their flagship Ascent 910C, deliver about one-third the performance of NVIDIA's most cutting-edge chips. So they are trying to compensate with scales and quantity by throwing massive numbers of these inferior chips together using advanced interconnect technology, scaling up to more than 10 ,000 chips into a single supernode, and eventually maybe a million chips in superclusters.

12:13So Huawei's offerings can theoretically match or exceed NVIDIA's server's performance. Just because they're putting so many of them together. Exactly. So basically the company, it's leveraging its networking expertise and China's power supply advantage to achieve competitive system level performance to compensate single chip performance limitation. And how does that compare to the H20 and the H200 chips now that have become so relevant to the NVIDIA story? How do the Huawei chips compare to what NVIDIA can sell in the country? So for Huawei's, so most Huawei's chips, they are on sale in the markets actually better than H20, which is the best chips NVIDIA can and is allowed to sell to China.

13:04But for the rest of the chips, like H200 and BlackWild chips, Huawei's chip is still far, far behind in single chip level. So that's why I say they were trying to compensate, trying to go with this quantity over quality route. And for these other companies, Beijing-based Cambricon is another key player. It's stock already tells the story. its stock surged over 100 % in the past three months as the Chinese investors are betting heavily on China's technology self-sufficiency push. Then you have other tech giants developing their own chips. Alibaba and Baidu have both started training their own AI models with their own in-house chips.

13:55Alibaba has been deploying their own chips called Zhenwu for a smaller AI model since earlier this year. And while Baidu is testing new versions of its flagship Ernie AI system using its Qunlun chips. So Alibaba's Genwood chip can perform as good as NVIDIA's H20 in their internal testing. Alibaba is also working on a chip for AI training, which in their internal testing can perform close to NVIDIA's A100. A100, it's NVIDIA's chip released in 2020. So you can, and there's other smaller player. Beijing have also fast track some smaller players like MoreThreads and MetaX response to UX export controls.

14:45You talked about ByteDance and Alibaba. They obviously are customers of NVIDIA and they're trying to use that technology themselves to the extent they can actually get a hold of any of it, given the export rules and then the regulatory hurdles domestically. But when you talk to people on the customer side, people involved in ByteDance and Alibaba who are trying to get these chips, are they frustrated that they can't use the best-in-class NVIDIA chips to operate their own AI products and services? I would say certainly yes. I would say absolutely when we talk to the AI developers and engineers, they want NVIDIA chips.

15:33The customer's preference is crystal clear. We just look at the buying behaviors. Bydance, Alibaba, and Tencent, all these big tech giants together, ordered more than 1 million NVIDIA S20 chips in just three months earlier this year. So roughly a four-year supply. So they're just trying to secure them before potential restriction. It's just Beijing creates a different reality. So we already mentioned a few restrictions introduced by local regulators. So you can get the detentions. So customers want immediate performance and ecosystem. But the government is essentially mandating their domestic boots.

16:18I want to ask you about this point here you made in your story recently. You wrote that Washington's fear is not just that Beijing will achieve semiconductor self-sufficiency, but that China will export an entire alternative technology stack that weakens U.S. leverage. Can you explain a little bit to us this alternative technology stack? What exactly is it that China is trying to build and what is it that the U.S. is so fearful of? Right. So Washington's concern, I would say, goes beyond just China making its own chips. Like, I would say China making its own chip may be actually the smaller worry for Washington.

16:59The real fear could be China develops not just chips, but an entire ecosystem, the semiconductors, the servers, the AI models train and can be run efficiently on those Chinese chips. the cloud service, the software, even the technical standards in the future. And then China can export this complete package to countries around the world. So why does it matter? Because right now, US leverage comes from the fact that if you want cutting edge technology, you have to play by the American rules. Companies need access to Nvidia's GPU, to US cloud services, to American AI models, it basically gives Washington an enormous geopolitical influence.

17:46But if China offers a parallel system, let's say, for example, Huawei's chips powering DeepSeek's models running on Alibaba's cloud infrastructure, now other countries have a second choice. Maybe it's not as advanced, as good as the American ones, but it's probably good enough for many inferior applications. Right. And then from the customers' end, if they're able to get everything domestically, not just the chips, but also the models, the cloud services, to the extent that they have built a parallel ecosystem, essentially, I mean, why not do everything domestically? It'd probably be easier for them.

18:25Yeah, exactly. And then for China, they can sell these packages to other countries. They can sell this to other countries. Maybe they would be willing to sell a good enough package, which is and with much, much cheaper costs. And some maybe crucially is sometimes can come with strings, less strings attached with like alignment with US foreign policy. So I would say that - Let me ask you this then, if that is the risk, how valid is that concern right now and how risky is it? Is there evidence that China actually is marching towards that parallel ecosystem? I would say this is already happening maybe in global south.

19:12We can see already some of the Chinese tech stack deploy in Southeast Asia, Africa, Latin America. And we can see increasingly AI infrastructures being deployed in those areas that would shape how these countries and districts develop their AI models and their applications using a more cheaper AI tech stack. Last question for you, Channer. What are you watching for in this story as it develops moving forward? Tons of things. The reporter says with a sigh, oh my gosh, Akash, if I could tell you everything. Yes, if we can have like a whole day, I can tell you tons of things. I think the two top things related to the US-China tech race.

19:59The first, we will try to keep an eye on Chinese companies, if Chinese companies, how or if Chinese companies can come up with new workaround to bypass US export control, to get access to advanced NVIDIA chips or cutting edge shipmaking equipment. And secondly, I think our Asia team is also keen to figure out, like we mentioned in some of the story, that if China has secure enough chip for training AI models. And if not, how companies are tackling the training chip shortage, whether they are trying to improve it on model level or trying to access to more smuggling chips or like try to improve the local alternative on a system level.

20:48We're just trying to figure out how they can tackle an issue like that so that they can catch up with their U.S. competitors. Great. Well, Channer, it is great to have you on the show. I know it's late for you, so we will let you get some sleep, and we look forward to having you back on the show very soon. Thank you so much. Thank you, Akash. Talk to you soon. Okay. The information released its 2025 holiday gift guide, and it features some standout products, so we invited a few of the founders from the list to come on the show to share the stories behind their businesses. First up is a pick from our editor-in-chief, Jessica Lesson, who added the board game called Board to our guide.

21:27It is a digital gaming console that sits on the table in front of you, so you're playing face-to-face with your family, but it still has physical game pieces merging digital gameplay with in-person interaction. It's not cheap. It is on sale for$4.99 right now, but it normally retails for$6.99. We're going to talk about that. But the person bringing this vision to life knows a thing or two about blending the digital and physical worlds. Bryn Putnam previously built Mirror, the interactive at-home fitness system acquired by Lululemon in 2020. Joining me now is Bryn, the founder and CEO of Board.

22:01Bryn, welcome to TITV. It's great to have you here. Thanks so much for having me. Let's talk the business of board games. I'm excited for this one. You built a fitness company. What made you want to start a board game company? Really, it was where I was in my life. I sold mirror to Lululemon in 2020, stayed on for about a year working at Lululemon. And then when I exited, I was in really just a different life stage personally. I had now two young kids, ages two and nine, as well as three grown stepkids, 17, 19, and 21. And we were just struggling as a family to find ways to connect. We were either playing games like Candyland to try to accommodate the littlest kids, or we would try to play video games with the older kids and they would smoke us because none of us could keep up on modern controllers with all the buttons.

22:51And so it felt like there really had to be a better way to bring the best of board games and video games together and bring people together playing face to face. And that's how Bored was born. Did you play it at Thanksgiving last week? We did. Bored got a lot of love at Thanksgiving. That's a big, it must be a really fun Thanksgiving. God. I want to talk to you about the game itself. So look, you have, it's like an iPad that sits on the table in front of you. You've got the physical pieces. Like, what exactly are the variations of games that I can play on the board? Yeah, so it's a 24-inch screen in a beautiful wooden frame that sits on your table.

23:27And what we spent two years doing is teaching the frame how to recognize physical pieces. So if you work with a normal iPad or tablet, it can only recognize 10 fingers. But our screen uses embedded AI to detect not just an unlimited number of fingers touching the device, but also physical game pieces. So it knows this is the spaceship facing this direction sitting over here, or this is the knife slicing the tomato. And what it enables is people to use physical pieces as a controller to interact with the digital screen. Got it. And what was the hardest part about this technically? I mean, look, I've never tried to build a board game, but you're an expert in all things sort of mixed reality here.

24:09Is it the piece detecting or the board detecting what piece it is? That seems to be the hardest thing that hasn't been done before. Yeah, so no one before has been able to have a screen that recognizes unlimited finger touches, gestures, as well as physical objects. So building that technology took us multiple years and was quite challenging. But also we were building 12 original games as part of our launch portfolio. So anyone who's worked in entertainment, I think, knows that finding the fun is quite challenging. So not only unlocking this new kind of technology for play, but also making sure that the games were really fun and engaging for the full family.

24:46So everyone from the littlest kids to the oldest grown up can play together and find joy. And I'm curious, and we're talking years out here, but do you plan to open the development of these games up to outside developers? Are you keeping it all in-house right now? How do you think about that? Yeah, Bored is a platform. So we were so excited after launching to have over 2 ,500 inbounds from developers within the first week, people really wanting to build their own physical digital experiences on Bored. And so we're already doing that. developers are starting to build games that will be available in 26 in our store.

25:22And we're also really excited about the possibility for the average consumer to build their own experiences on device. So we really want to build UGC creation tools. So Bored goes from being a gaming console to a broader platform for everyone to create their own experiences. Right. Now it's$4.99 right now. It just came out about a month ago. It's$4.99. Prices are going up, I think, later this week after the holiday sale ends to$6.99. Who is the customer here? Because it's not cheap. What's been really exciting to see is the customer really on day one has been everyone. Within the first week, we were in every single state, except for Wyoming.

25:58Wyoming held out a few extra days. But beside Wyoming, we were in every state. And we've had just an incredible range of buyers from people who are in tech, excited to build their own games, to really the average mom and dad who are looking at this to be the family Christmas present to bring people together. And so it's really hasn't been sort of a coastal early adopter product. It's from day one has really been quite mass, which has been really exciting for us. And I'm curious how you're approaching the marketing here. Can you talk about which channel has been most effective? I say that it's been a month.

26:32I know that. But, you know, from your experience at Mirror, now bringing that here, which marketing channel are you investing most into? Are you leaning heavily into D2C? Or do you have to rely on the retail platforms now? How do you think about that? We're doing a little bit of everything. So it is a D2C product. You can find it at our website, board.fun. We also have a holiday pop-up at Camp, which is a premium toy store that has nine locations nationwide. And you can find board there and do a demo. But what's been really exciting is sort of the visuals and the storytelling of play bringing people together has really resonated.

27:12And so it's just been about where can we get that story out, be it Twitter or Instagram or on Hacker News. So it's been kind of all over the place through press. And last question for you, and I'm sure you get this a lot. I mean, I thought, and I don't have kids, so you know this better than I do, but I thought the whole idea was we were trying to get kids off of screens. Yeah, I mean, I think that's a question of what we're asking the screen to do. So there's screens that pull us apart, everyone on their own phone or their iPad sitting in the corner in an isolated way, flipping between different tabs sort of distractedly.

27:48And then there's a screen that brings people together to create, collaborate, compete. And that's really what board's about. The screen isn't a passive, independent experience. It's something that's creating really great core memories for people you care about. Got it. Well, I look forward to giving it a shot and playing a game. Brynn, thanks for coming on the show. We really appreciate it. And good luck in your next game, whatever it is. Thanks for having me. Okay. Our next pick from the 2025 Holiday Gift Guide comes from the Information's e-commerce reporter, Anne Guillen. It is a subscription for Flighty, the flight tracking app that delivers real-time updates on delays, cancellations, gate changes, and more.