Why Sequoia's China Spinoff is Struggling, Plus Apple's AI M&A Strategy | Aug 26, 2025

0:12Welcome everyone to the information's TITV. My name is Akash Pasfritz. It is Tuesday, August 26th, and we have got a great show planned for you today. We've got the CEO of Decagon coming on the show to talk about how he's thinking about AI agents in the current moment. We've got a big story about Apple and how it might be changing its tune to M &A in the current era of AI. We've also got the founder of an AI CRM application that is officially coming out of stealth today. But I want to start with a big story that our Asia Bureau just published this morning. Two years ago, Sequoia China split off from Sequoia Capital and set up a new firm now known as Hongshan.

0:53The company raised a huge chunk of money to invest in startups in China. And today, our Asia Bureau published an in-depth story looking at what the company has been investing in and also how much of its giant war chest actually has been invested. I want to bring on the information's Asia Bureau Chief Jing Yang to talk about her reporting. Jing, welcome to the show. It's great to have you. Hi, Akash. Thank you. So let's go back in time a bit. Just give us a very brief history here of Sequoia China, Hongshan, and what the mandate of this new firm actually was. Yeah. So Sequoia set up the China arm back in 2005 in a partnership with Neil Shen, who at the time was an entrepreneur.

1:38And at the time, it was the mid-2000s, and China was at this coming-of-age moment. Many VC firms in Silicon Valley believe that there is a lot of money to be made from backing startups in China. And then they were correct. And Sequoia Capital, especially, is correct in making that move. So in the ensuing decade and a half after the setup of Sequoia Capital China, Sequoia China, under the leadership of Neil Shen, made a lot of really big and smart bets. I'll just list a few. For example, Bydance and Chinese food delivery giant Meituan and also PDD, which some of our audience might be better known as a parent company of Temu.

2:30And then all of these, you know, Sequoia, China, and Nielsen spotted these companies at a very early stage. And then, you know, in the case of Meituan and PDD, for example, when these companies went public, eventually returned very handsome, like billions of dollars back to Sequoia Capital and its investors. For example, when PDD went public in 2018, the returns that Sequoia Capital got were equivalent to the returns that Sequoia got. in Google's IPO in 2004. Holy moly. And then, obviously, things changed. You know, the relationship between U.S. and China changed in the last few years. That sort of forced a lot of Silicon Valley VC firms to have a reckoning of the new reality.

3:18And as a Koya Capital being probably the most successful in their China investments, was caught in the crosshair of, you know, regulatory tensions, you know, DC politics. Right. And then, so long story short, in 2023, Sequoia Capital announced a milestone breakup where it will separate. This was huge. I remember, I mean, this was a big breakup. Yeah. It's what I call the watershed moment of US-China venture capital decoupling. Right. Because then we saw several such breakups also happened. It was really Sequoia that started this trend that sort of telegraphed it's the end of an era. Right. And so let's fast forward a bit.

4:08So the breakup happens. Hongshan, which is the new name of the new firm formerly known as Sequoia China, they raise$8.8 billion for their first fund, I think it was. what have they been investing in since then? This was the bulk of your reporting. Tell us about it. Yeah. So just to add a little bit of context, the$8.8 billion actually raised in the summer of 2022. That was, if you remember, that was 12 months roughly before this split up, right? So when that huge war chest was being raised, it was still raised under the brand name of Sequoia Capital. And then quickly enough, 12 months later, the split up happened.

4:54And what does the name Hongshan mean first? Actually, that Hongshan is exactly Sequoia in Mandarin, right? So for Sequoia capital and its China arm, in China and in Chinese, it's being referred to as Hongshan. So when Sequoia China went independent and needed a new name, the fact that they went for Hongshan, which, by the way, means that in Chinese, they didn't have to change their name, right? Or the English branding part, right? They could go business as usual, really. I mean, and they just raise a whole lot of new money under the same name, so to speak. Yeah, and then they can continue to, for example, in China, right?

5:36They can continue to, there's no disruption in terms of a brand image, in terms of when they reach out to funders or when they network, right, we're still Hongshan and then Niu Shen is still the commander-in-chief. Right, right. So then they just adopted the English moniker, which is, you know, the opinion, the Mandarin Chinese for Sequoia. So what our story reviewed, the main review, is that that$8.8 billion raised three years ago, it's now only about a quarter of it has been deployed. Well, what does that mean? If we were to compare the previous track record of Sequoia Capital China and the cycle of which they raise funds and deploy them and raise again and deploy again, three years having deployed only a quarter is quite slow compared to its own previous track record.

6:34And then we dug the deep into why that is the case. Okay, so lay it on us, so why? Yeah. So it's quite a complicated set of reasons. Obviously, back when they raised the funds three years ago, which was a year before the breakup, I think a big chunk of the vast majority of the$8.8 billion is actually earmarked for the growth fund and expansion fund. And those deals are just almost non-exist now in today's China for several reasons, right? The Chinese economy hasn't really, you know, failed to stage the post-COVID rebound that many people have hoped. And then, you know, if you remember back in 23 and 24, globally stock markets, public markets haven't been doing well.

7:23Public market investors haven't been very generous with how they value tech companies. And that is the same case with Chinese tech companies. then that sort of then trickled down to the valuation of private companies, sort of relatively mature private companies, companies that are doing growth and late stage pre-IPO rounds of funding. You just can't, you know, have like that many deals that raise, you know, multi-billion dollars. This kind of deal just didn't exist anymore. But NIO and NIO Shen and many other venture capitalists didn't have the principal back then in 2022. They didn't see me, right?

8:00And then one of the other things that you mentioned in the story was that they've actually expanded now outside of China as their new focus. Yeah, that is the focus in addition to, you know, the Biden China. So basically, in order to pick up the pace and also bear in mind that, you know, when Hongshan was still part of Seqoia capital, they kind of, you know, their realm is kind of just China adjacent, right? Now that they're on their own, theoretically speaking, the world is their oyster. They can go anywhere they want to go. And that's almost what happened. In the last two years, Hongshan set up offices in Tokyo, Singapore, and London, and they started doing deals in those markets.

8:44For example, one of their biggest buyout deals, which, by the way, is not the kind of typical deals that, say, Sequoia Capital in the U.S. would do, is they took a controlling stake. Our reporting shows 80 % of stake in a Stockholm-based order you can make a Marshall. I don't know if you guys have heard of Marshall, but I personally am a big fan of their speakers. But anyway, that deal is sort of like a representative of what the new Hongshan is. And they took a stake in Eight Sleep, the sleep company. We just had them on the show, actually. I was surprised. Yeah. So we also talked in our story a little bit different in the ASLEEP deal.

9:30So ASLEEP's recently announced round was about$100 million. Right, right. Round was, you know, Hongshan participated in the round, but in order to not attract any scrutiny from the U.S. government from CFIUS, they had to make sure that they don't invest too much so that they don't get a board seat at ASLEEP. Right, right. And then this is kind of the reality, right? Hongshan has increased its best in markets like the rest of Asia and in Europe. However, in the US, it's still sort of off limits in a sense that they have to be very careful in the type of companies that they invest. For example, ASLEEP does incorporate technology in its company, but it is not, you know, I say, for example, a large language model developer, which is very sensitive.

10:23Let me jump here. Let me sort of zoom out a bit. This was a fascinating story to me. It had a lot of detail in it. One thing that I want to get your take on is it's slow to invest. That was kind of the headline, right? You talked a lot about what they're investing in. But why should readers really care about this? If a venture fund is going slowly, who knows? It could be there was less opportunity, the time wasn't right, the market conditions weren't there. But why should readers really be paying attention to the pace of Hongshan's investments right now? Yeah, big picture level. If you have an interest in keeping up with what is happening in China's venture capital and innovation land, then there's no better person and no better firm to follow than Nielsen and Homshen.

11:15And then that is just the high-level picture, right? How he's doing is, you know, two years on after getting on his own and also the kind of best that he makes, right? There's a lot of people that actually, you know, because he is, I think, arguably the most celebrated venture capitalist in China. And in China, his name is synonymous with venture capital. And in the U.S., also very highly regarded. For example, he was three years in a row ranked as number one in Forbes Midas list, higher in those years than many famous venture capitalists in the Valley. So that's why. And at a more micro level, I think it's, you know, the$8.8 billion, there's also a significant of American limited partners, such as University Endowment and foundations, that are investors, have been investors in that huge war chest.

12:19And the last question for you before we let you go, those investors, and I'm not talking just the American investors, but the LPs in the fund, have they expressed any kind of discontent with the pace of investment? Yeah, so the picture based on our reporting is kind of mixed. So I spoke to some LPs who sort of feel like, you know, based on Hongshan's previous track record, the deployment of the current series of funds has been slow. So maybe, you know, bear in mind, LPs have no right to asking their managers to return the money once they committed it. But they were secretly hoping that maybe you can consider return a part of, you know, the capital back to us.

13:03And, you know, if you don't have, like, a lot of opportunity to deploy the capital quickly, to sort of put in a bit of a goodwill. So then the next time you come back to fundraising, then, you know, it's a good relationship building, right? But then there's also some other LPs who are very confident in NIO based on his track record, who thinks that it's still too early to tell, you know, venture capital, the entire venture capital industry is going through some changes right now. I'd rather give it NIO and Hongshan, you know, some more time before I make any decisions. Right. Well, Jing, thank you so much for coming on the show.

13:42We appreciate it. And I should say a special thank you because you are actually on vacation right now, but you made time for us. So thank you very much for coming on. That is Jing Yang, our Asia Bureau Chief at The Information. Well, moving topics now. Apple has never been an avid dealmaker as it relates to M &A. But this morning, the information published an inside look as to how Cupertino has become a lot more open to acquisitions in the era of AI. It has even looked at some of the biggest AI names in the space. We've written about a lot of them here at The Information. I want to bring on Aaron Tilley, who was one of the reporters who wrote that story, to talk all about it.

14:20Aaron, welcome to TITV again. It's great to have you. Yeah, gosh, thanks for having me. So talk to me about what Apple's track record has historically been with acquisitions. Yeah, so they've, as you've said, they've never been a big acquirer. They've had this sort of deep aversion to big M &A. They've acquired a lot of companies over the years, but these have been very small deals, like in tens of millions, you know, maxing out at, you know, a couple of hundred million. And there's a few exceptions there. Of course, Beats in 2014, the headphone. I remember it well. Yeah, it was a big announcement.

15:023 billion. That was very unusual for the company as sort of an anomaly. But besides that, they've really done very small deals. These are deals where they're finding some sort of gap in their technology that they need to fill with a certain team or a certain sort of IP that a company offers. These are very small deals. They're extremely frugal and they just haven't done that in the past. And why have they not been so acquisitive in terms of, you know, these large deals like other companies have? Yeah, I think it's this history in their DNA of frugality that has kind of stopped them from doing the big deal and looking for smaller tuck-ins.

15:47And then I think more than that as well is just sort of the cultural aspect of M &A. I mean, whenever a big company is acquiring a smaller company, there's always going to be some challenges there, getting these two companies, their cultures to align and figuring out how that works, the integration challenge. But Apple, that's especially acute. Apple's culture is extremely specific. It's about seniority, how long you've been there. And when you're coming in as an outside company, you really have no sway. And so a lot of sort of challenges there for new companies and people end up leaving quickly.

16:30So it's just kind of oftentimes not a successful integration for Apple. So, right. Right. So, OK, so we have this long history of Apple, culture of frugality, like you said. Now AI comes along and it seems like the tune is changing. Talk to us about what you found in your reporting. Yeah, so I think they see, like everyone in tech, the significance of AI in this moment with LLMs and these large models that we've never seen before. So they recognize that and there's some interest and there's conversations internally around maybe big deals be done here. Specifically, they've brought up perplexity and mistral and specifically Eddie Q, their Apple, the services head of the company, Eddie Q has sort of been a key part of that kind of like investigation, promoting that as a potential deal here.

17:35But yeah, they're just looking around because, I mean, look, every few years when some big new trend happens in tech, the topic of what is Apple, the Apple needs to finally open up the checkbook. They need to be open to big M &A every few years this happens with a new big tech trend. And the question is, is AI that much different? Is it that much more significant in this era? So, you know, they're debating that internally. And, you know, you said this happens every few years as the new technology comes along. In your reporting, did you sort of get a bit of a temperature check on are they leaning more towards possibly doing a big deal or is this very much just considerations right now and sort of some people championing the idea?

18:24Yeah, it's some people championing the idea. To do big M &A at the company requires some sort of consensus among the sort of upper echelon of executives there. And Eddie Q is just one executive and not the person who's really adding their AI strategy. So I don't think he's going to be able to decide just on his own to do that sort of big M &A. So I think it's going to be a conversation inside the company still. Right, right. And last question for you. I mean, what will you be watching for here? I mean, obviously, there's a question if they do an acquisition, if they do some of these big partnerships, that's something to watch for.

19:03But, you know, as it relates to open questions right now, I mean, you know, have acquisitions helped Apple in the past? Or do you sort of expect that it'll just sort of chug along and, you know, we might see something, but it'll be quiet if anything does happen? Yeah, Apple has a we build it here mentality. And their aversion to big M &A, I think is smart. I think they're a challenging company that, you know, does not open outside influence with open arms. So I think it's smart that they kind of avoid that and maintain that posture. Because I think to, you know, make that sort of transformative deal, I don't, I'm not super confident it's going to work out.

19:48Right. And these would be big bets too. We should say, you know, a lot of these companies, Perplexity, Mastral, I mean, you know, we don't know what they would go for, but these are companies that are obviously raising a ton of money. And so it would be a big bet, as you pointed out. Aaron, it was a very fascinating story. Thank you for coming on the show and telling us about it. That is Aaron Tilley, who covers Apple for the information. Well, one of the easiest applications of AI agents to wrap your head around is having automated chatbots answering customer service requests instead of needing to employ real people to do the job.

20:20Decagon has become one of the fastest growing companies in that space. And I want to bring on CEO Jesse Zhang to talk more about how he thinks about agents in this current moment. Jesse, welcome to TITV. It's great to have you here. Hey, gosh, thanks for having me. It's great to be here. So look, we're trying to sort of ask big picture questions here on the show a little more. And so I want to ask you perhaps the biggest picture question you could ask an AI founder in this day and age, which is that you're in the space of AI agents. What does Jesse Zhang's vision for the future of the internet look like with AI agents?

20:58Yeah, so if you think about our AI agents, right, at the core, what they're good at are conversations. And I think that's something that we've realized that AI and Gen AI models are very good at, is having conversations with folks because that's what they're trained to do. And what we have really set out to do is really build a nice and clean system to construct these conversations for large enterprises, right? So if you think about, you know, an airline or a bank, you know, I use an airline, I want to book a flight, or I want to upgrade my seats, or I have questions about my loyalty points. This AI agent can just help me resolve all these things and take actions for me.

21:33And I can just get what I wanted immediately by just texting a number or calling a number. And in the bank example, I lost my credit card. I can just go get a new one immediately. So that's really the style of AI agent that we're interested in. And you mentioned customer service. That's often the best place to start because organizations can really easily quantify the ROI for themselves. Like, you know, how much money am I saving? How much new revenue am I driving by making my customers happier? Right. And ultimately, that will expand, right? So our whole vision is that, hey, eventually, you know, in a couple of years, there's not going to be that much difference between the types of conversations.

22:10You'll have a single AI agent that almost serves as like the front end, like concierge for your brand. Right. The all-in-one. All-in-one. Exactly. Right, right. So, you know, this is kind of an interesting point you raised because I was talking with with a couple of colleagues in the newsroom. And it feels to me like agents, you've got these sort of application layer companies that are building agents for specific use cases. And the defense, the way I take it, when they get asked the question, well, how do you compete against, for example, an open AI coming in and building something? The defense is, we're really specific.

22:47We're solving this one problem. But then inevitably, you get this company that starts with one problem and then you start to expand. And then you get sort of closer to, like you said, this sort of all-in-one agent, which is sort of the concierge type of product that we might be going towards. And so when I think about that, I sort of think about, okay, well, then doesn't that sort of specificity or the argument that this is how we defend ourselves against the open AIs of the world, doesn't that specificity kind of go away? I mean, how do you think about that? I don't think the specificity ever matters in this scenario.

23:21So I actually don't think specificity is the thing that gives you protection against platforms and so on. I think what really matters is like how deep is the software or the application you're building on top of the models, right? And to that point, yeah, maybe the points other folks have made is that specificity allows you to build more depth in the application. And that's what gives you protection. And you just think about kind of, you know, the models themselves, right? They obviously have created a massively valuable technology. Everyone's using it. It's amazing. But there's also kind of risks with, you know, this API business because, you know, there's a bunch of different ones.

24:03They're all kind of interchangeable. And so they're kind of driving the prices down. They're driving the margin down. And so they know that moving into applications is healthy because, you know, if you're building an application, you're solving a business problem. And so you have, you know, potentially more margin there and more business to take. And so if the application is more of a thin layer on top of the model, that's that makes it more attractive for them to build and easier for them to go into. Right. And so you see all the coding agents from the labs and of course, those are very difficult to build.

24:37But the application like the sort of the sort of stuff you need to build around the AI is relatively small because you're kind of delivering a user product that people can just download. Right. And the goal, if you want to stay, I guess, more independent long term and not cross paths with the labs is you have to be building something that has like such a deep level of application where there's all these features you have to build. There's all this tooling. There's all this stuff that has to go in there for, you know, for example, for enterprises to buy you. And I think that gives you more of the staying power.

25:11And so I think the depth matters a lot more than the specificity. Right. You talked about sort of some of the margins that some of these bigger players are dealing with. As it relates to your own business, I'm curious, what are the biggest cost drivers for you right now? And what are the biggest moving parts that you're trying to sort of figure out cost wise, broadly speaking? Yeah, I would say for any application these days, the biggest part of your cost is going to be model inference. So, you know, calling models, using them across the stack, right? And this could be like outside of the agent itself.

25:44You know, for example, in our space, there's a bunch of very valuable tooling that our customers care about. For example, simulating test conversations before the agent even goes live. Reviewing the conversations afterwards to see if there's any learnings and if there's any insights and stuff like that. So those all cost model inference. and that's where most of the cost comes from. Right. And the levers there are over time, the models will naturally get cheaper and they'll naturally get faster and so on. So that is helpful. But then I would say we're also seeing a trend in the industry of people fine-tuning more models because once you know exactly how your agents are structured and you know the tasks each model needs to do, you can pick off, you know, this very specialized ones and fine-tune smaller models for that.

26:26They'll probably have better accuracy and way faster, way cheaper. Right. So that's generally how applications are going to go. You talked about inference costs. Well, I was listening to a podcast you did a couple, actually, it wasn't a couple months ago. It was in January. Okay, it was kind of funny. I listened to the podcast you did in January, and it feels like decades ago, you know, you're talking about literally like 01 and GPT-4 as if it was like eons ago. You know, I'm listening now. We've come so far. One of the things you mentioned that I found funny was you talked about the latency that it takes for some of these products.

27:02And a little trick that I heard you say is that's why you say sometimes a chatbot says, give me a second to look up your information. And when I sit in there and I'm using Decagon on ClassPass, I believe it. I'm like, okay, let it work. It's looking at my information. I get it. I did not think that that was latency happening in the background. So that was kind of a funny realization. We talked about cost. I want to ask you very quickly about pricing. How do you think about pricing at Decagon and pricing in AI agents, probably speaking? Pricing AI agent, I would say you generally want to tie it to the value that it's providing.

27:39And this is a little bit different from the classic SaaS model. I mean, in the classic SaaS model, you're also charging based on the value you're providing, right? But with most SaaS, the way you quantify the value is, you know, seats-based or, you know, you're paying kind of a recurring subscription based on the number of people using it. And with AI agents, it's, you know, if you have an autonomous agent, there's no one quote unquote using it. There's folks managing it, of course, but you want to tie it to the output. And so for us, the output is essentially the number of conversations it handles, right?

28:07So if it is kind of being this front end for an airline or a bank and people are having tons of conversations with it, you know, the number of conversations is kind of the unit work. So the way we generally price is we work with our partners and we're generally working with large enterprises here. So I work with them to scope out the use case and here's how many conversations you'll, they're almost paying for like a bucket of conversations for the term. Do you price it at all based on like, it's a number of conversations, but then, you know, if the agent actually did the thing that it set out to do, is there a success metric there?