General Catalyst's Novel VC Fund, Creator Economy Shift, AI Inference Cost Prediction | Nov 19, 2025

0:13Welcome, everyone, to the Information's TITB. My name is Akash Pasricha. It is Wednesday, November 19th. We have got a great show lined up for you today. First up, the debut of The Information's new finance newsletter. We'll hear from our finance editor on why he's launching that newsletter now in this moment and all the different ways that he is thinking about the AI finance ecosystem. We'll then talk to the co-heads of General Catalyst's Customer Value Fund, which has deployed$4 billion so far. And we're also talking margins and AI with Amazon Web Services Director of Technology. And finally, we are closing out the show with a conversation with Kaspar Lee, the co-founder of the venture capital firm Creator Ventures and a longtime YouTube sensation himself.

1:00It is a big show, and so let's get right on into things. Today, The Information published its first ever finance newsletter. It will publish weekly on Wednesdays and will be written by our own finance editor, Ken Brown. So we're bringing Ken on today to talk a little bit about what prompted this newsletter and what he's thinking about. Ken, welcome to TITV. It's great to have you back. Hi, Akash. Good to be back. Well, I made this joke earlier today, and I'm going to make it again. Big day and big debut for you with your column. Congrats. Oh, yes. Thank you so much. Yes, it's very exciting. It's, you know, it's fun to write.

1:34I'm an editor here all the time, and so to get to really write my own stuff is a treat. So talk to us about what prompted you to launch this newsletter now, today. Well, you know, it's no surprise that money has always been a big deal in tech, but now it's bonkers, right? I mean, we used to talk about millions and billions, and now we talk about trillions. And so, you know, just the movement of money around tech and what's driving, especially the AI boom, is it begs for someone to just look at it through a financial lens. Right. And now when you say financial lens, talk to us about all the different angles that you look forward to double clicking on.

2:13because your column today was a really great overview of honestly what I thought to be the entire story. I kind of read it and I said, okay, well, that's it. That's the story. And so what are the areas that you're looking to dive deeper on in the coming weeks? So you're never going to read another column, right? That's what I am. Well, no, it's a bit of a page turner. It's like, okay, well, I didn't know it was more complicated than that. Right. So we're going to track the AI build out. We're going to track all the money pouring into AI and watch who's winning, who's losing, how they structure all the money, and probably most importantly, where the risk is building up.

2:53So everyone talks about bubbles now. One of the things I say today is we're not there yet. And one of the reasons is because this whole big build-out explosion of AI is backed by the richest companies in the world. These guys can write checks, not for trillions, but they can write a lot of checks. They've so far paid actually for most of it. And so that is going to be what we're looking at. But then the bond market is going to finance it. Sovereign wealth funds are going to finance it, pension funds. Even your life insurance policy might be invested in the AI boom. And so we're going to track all that and look for risks, surprises, things people should be wary of.

3:33And we're going to try to explain it all very clearly. Now, you say we're not there yet in terms of the bubble. I'm curious, compared to some of the other bubbles or some of the other frothy periods that you've covered throughout your career, how does this one look and feel different compared to those? Well, the first thing is bubbles can last a lot longer than anyone thinks. You know, people call for a bubble and two years later, they're still calling for a bubble. So, you know, I'm always a little skeptical of that. And the reason is, in this case, there's just tons of money going in and there's no reason for there to be a crack because all the people who've been borrowing money have time to pay it off.

4:15So this one is different. Well, one is different because it's backed up by the richest companies in the world. Two, the businesses it's funding, say what you want about the dollars. They're growing very fast and they're having a lot of revenue now. They're nowhere near making profits. So that's a real question. But that's going to take a while to see how it plays out. So this is not—and, you know, we have another— So right now, there's a crypto bubble that's bursting, right? So it was down a ton. All these people who loaded up are suffering, you know, losing half their money. Like, that's a bubble.

4:53That's just short. It goes up, and there's no logic to it. And it was just, you know, euphoria and that's over. This is a big fundamental change. And it's not going to be smooth. There's going to be bumps, but there's just a long runway to see where this goes. Right, right. One of the things I wanted to ask you about is you outlined some of the places we might look for signals in terms of risk, you know, where the cracks come from. The stock market is one obvious place you could look, right? I mean, NVIDIA reports earnings. Earnings, let's say, don't come in as people expected. Fine. That's one way people might get the signal.

5:32But I'm wondering behind the scenes, you know, certainly thinking about the bond market and how some of these markets work in the back end. Where else are you keeping track of in terms of fault lines or cracks that could potentially show up? Right. You know, you start to see concentrations of risk, right, where people are taking on the risk that other people might not want to handle. And they may be smart. I mean, people handle risk all the time. But you look for where that's happening and you try to unpack it. So we've done a ton of stories on NVIDIA, which is backing all these AI startups. it has a real incentive for AI to grow because it sells the chips.

6:11And, you know, it's one of these things. If AI's growth slows, NVIDIA sells fewer chips, the commitments they have made might be harder for them to make. And so you just need to understand what the triggers there would be. Some of the data center builders, you know, it's such a capital intensive operation. Some of them have taken on a lot of debt and taken on that level of risk. And one thing to look for is, do they fall behind schedule? Can they not produce the data centers that are supposed to be? Can they not get the power? Can they not get the chips? That's going to be a trigger for some of those guys.

6:44Right. And last question for you, behind the markets, there are people and characters that we like to track. And there are the characters that we all know about, the founders of these companies, the executives of the hyperscalers. I'm wondering if there are people that we may not be paying as much attention to that you are actually watching very closely as an interesting character in this finance ecosystem. Yeah, so I'll give you two. One is this guy, Mark Lipschitz. He is the head of Blue Owl, which is an investment firm, mostly deals with bonds. Right. He's done a lot of riskier stuff. In the news today.

7:20In the news today. They've been in the news every day. They keep doing big deals. But they've turned out to be a big investor in these big projects. And so they're smart guys. They're taking on risks they think is reasonable. But they're someone to watch. And the other person I'll say is Michael Interator, who is the CEO of CoreWeave. So CoreWeave is a data center developer and cloud company. And they are one of these companies that's borrowing a ton and making a big bet on AI. and so far, they've done pretty well. Their stock is, they had a huge IPO. The stock is down a bunch, but the company is still happening and that is another one to watch.

8:03Right, great. Well, Ken, congrats on the launch of the newsletter. It's one that we'll be reading and watching very closely and look forward to having you back again on soon. Anytime, thanks. Talk to you soon. Okay, General Catalyst has been one of the most innovative companies in venture capital as of late with its move to buy a hospital system recently and its expansion into wealth management. The company also set up a new type of fund a few years ago called the Customer Value Fund. It has supported companies across sectors, including software, consumer, healthcare, and fintech. Joining me now are the two managing directors of the fund, Pranav Singhvi and K.V.

8:40Mohan. Gentlemen, welcome to the show. It's great to have you here. Likewise. Thanks for having us, Akash. Really appreciate it. So, K.V., let's start with you, K.V. And look, I don't know who likes to go first in these things. So, Pranav, if you want to go first, by all means, jump in. But the first question I have is who's going to explain to me what the customer value fund is? Because, look, it sounds like a bit of buzzwords up front. So, you know, we got to get into it here. Sounds good. I can take that one and KB can add to it if that works for you. Sure. So, look, the customer value fund is quite simply, if you think of a technology company, once it reaches a certain size and scale, the vast use of its own cash is to invest it in sales and marketing, the tech nomenclature for which is customer acquisition cost.

9:22And what these companies do once they reach a certain size and scale, they're constantly investing their own cash in sales and marketing to generate lifetime value, customer lifetime value. What happens with these companies is a lot of that cash constantly gets tied up in customer acquisition cost, and they have to constantly keep reinvesting because they invested today and it gets realized over some period of time, say 12 or 24 months. And then over time, they may make multiple times their money on their investment in CAC. The customer value fund was really designed to solve the problem of the fact that you have to invest it in.

9:52It takes you some time to realize your return on sales and marketing. And that is a huge cash drain on a company's own balance sheet. So what the customer value fund does is goes to companies and says, instead of using your own cash to fund sales and marketing, we will provide you a dedicated balance sheet to fund your ongoing sales and marketing spend. And in return, we only get paid back from the specific cohorts of customers we helped you acquire from that spend until we hit a target return multiple. If it works, we get our target return multiple. If that specific cohort underperforms, takes longer to pay back, or on the extreme downside does not pay back at all, we own the full downside risk.

10:25We have no recourse to any other part of the company. So it really helps companies be a lot more front-footed about growth and scale their investment in sales and marketing. how do you think about about you know it's not debt it's not equity in terms of the return what does that look like from your end yeah i mean look i think we reject the notion that there can only be two financial products i mean if you go outside of the little bubble of silicon valley you'll see all sorts of different financial products that exist in the market the right analogy to think about what's what cvf is it's really um a lot of inspiration is taken from asset-backed financing we're effectively treating your investment in sales and marketing as an asset and providing what looks and feels like off balance sheet financing.

11:04And so it's just a completely novel way to fund your growth in a way that has not existed for this industry before. And it allows founders and companies to unlock all the cash they have to constantly keep reinvesting in sales and marketing to go do something else with, something that's much more productive, that is investment in product, in engineering, in M &A, and over time, if it makes sense, even buybacks. So hopefully some of the VCs can realize the elusive DPI that everyone's been Right. KB, tell us about some of the companies that you've backed with this fund so far. Yeah, absolutely. We have backed over 60 companies by now.

11:37In the last five years, we have seen tremendous growth. And also in our companies, some of the companies we work with, Lemonade, it's a company which is public. We work with Grammarly, which is now called Superhuman. You may have seen the announcement. One of the companies that we worked with was this gaming company called Superplay, which was recently acquired by Pletica for around$2 billion. And across the board, we work with companies in healthcare, in consumer, in finance, across like in enterprise, basically across the field, any company which has found product market fit and has long lifetime values is a company, and wants to go from good to great, is a company we like working with.

12:17KB, I want to stick with you on this. When you think about customer acquisition costs. The cost is one thing. The method in which you actually acquire those customers with respect to what you invest these dollars into, I wonder if that has changed at all in the last few years, given the shifts we're seeing in customer behaviors, where they live online. B2B is obviously a bit of a different game, maybe a little less, maybe changing less quickly, I guess. But how are you guys dealing with the ways in which your portfolio companies acquire customers using that money? Look, what we always tell people is we help putting fuel to the fire.

12:55We cannot create the fire ourselves. The best people to create that fire to figure out how to grow are the founders. And we actually don't opine on how these companies should be spending. What we actually do opine on is most companies are massively underspending. What I mean by that is they need to run more experiments. The best channel always changes. You know, whether it's going to be, you know, like billboards or Facebook ads or just hiring a sales team. Honestly, we have seen a lot of companies do really well even on TV. Our goal is not to opine. Our goal is to allow the company to experiment because that is how you figure out.

13:31And as things are changing, channels get saturated, they get desaturated. You want to keep experimenting and making sure you're maximizing your enterprise value. Right, right. That is how we think of it in a very broad lens. But Pradov, I am curious, Chris, the ways of marketing from the outside, how have you observed those ways change in terms of what's effective? Because, you know, Facebook ads, they may have worked very well five years ago. They might not work as well right now. So what are the newer strategies that you're seeing companies take on to acquire those customers? You know, it really depends.

14:06It's very company specific. There will be some companies for whom meta still works really well. Ultimately, it is a very company specific thing. But to your point, we are very excited by new ways companies can go acquire customers. People are using all sorts of different techniques, not only just based on channels, but other ways of getting in front of customers. It will vary from company to company is the reality because it's going to be very circumstantial based on are you a consumer company or your B2B business? What is the right way to get in front of if you're a B2B business? The person who's your actual buyer?

14:35What's the highest propensity to buy for somebody on the consumer side? That will all actually be a function of the company itself. But to your point, and what KV mentioned to really emphasize that these companies need a diversity of channels and they have to constantly experiment to find out what works because it's growth is in highly experimental endeavor. You'll find a channel that works for a time period and then it saturates and that interim period, you have to keep looking for other channels. And a lot of that will be through experimentation. And once you find something that works, then you sort of scale it again and you keep going through that journey.

15:07KB, as you're looking for companies to invest in, in this period where it seems like valuations are high and we're in this sort of talk of a bubble, how do you tune out the noise and how do you really adjust your criteria for assessing companies in a way that sort of accounts for the froth, as people say? Yeah, look, Pranav and I are not in the business of identifying the next team jobs. I know a lot of folks are in that business and, you know, I wish them the best. We are in the business of doing math. What I mean by that is the way we actually work with these companies is we get that raw transaction level data, which is, you know, my background is technology.

15:51We work with these companies. We connect directly to the data warehouses and you're getting the raw transaction level data. We are underwriting math. And, you know, math doesn't change over time. It is the same. So for us, our criteria has always been the same. And there are companies that, you know, we pass on that, you know, I obviously won't name them that you probably consider to be hot companies because they have not found product market fit. And our definition of product market fit is very simple. It is, can you spend a dollar in your growth machine and consistently generate an ROI based on that?

16:27Right. If you can do that and if you can see that, you know, over the last, 12, 24, 36 months in the transaction data we're getting in all the statistics that we are running. That is the kind of company we want to work with. And it turns out, you know, despite all the crazy valuations, they're an insane number of businesses, which are still fantastic businesses, that in our opinions, and I would say like we feel very confident about this, are still massively under-investing in growth. Right. I don't want to add something. No, if I can add to that, But valuations go up and down based on macro environment that has nothing to do with the fundamentals of a company.

17:05And so it's actually we're quite fortunate that, you know, we are not basically indexed on the valuations going up and down or the stock market, the stock price going up and down because we do work with publicly traded companies as well. We're solely focused on a company's ability to spend a dollar, acquire a customer and retain that customer to generate an ROI that makes sense. It's an extremely fundamental product. And look, the prices go up and down. we're kind of sort of completely isolated away from that. Right. Last question for you, Pranav. I read that KV on his bio, he said that he likes fitness classes.

17:40Given that you guys are a bit of a dynamic duo here, Pranav, do you know what KV's favorite fitness class is? I would hope that he's working with Ladder and using Ladder because that's one of the companies we're invested in. But I know KV. KV, what's the class we got to go to? You know, I used to love a rumble in the marina, which unfortunately shut down. So these days, my friends have been trying to get me more into Pilates. Although I think more of a - Okay, solid core, solid core. Haven't really made it more than once or twice yet. So it is a while before I find my new fitness class looking for recommendations.

18:12Okay, great. Well, when you're both in New York, come to F45. That's the class that I love and we'll both get a workout in. Awesome. All right, well, thank you to the both of you who are coming on. We really appreciate it. It's a very interesting way of looking at venture and we'll have you back on the show again soon. Thanks so much for having us. Okay. Well, our next segment is with our presenting sponsor, Amazon Web Services. Margins have been a big topic related to the AI boom as companies all over are trying to figure out not just how to generate revenue with these new tools, but also turning a profit.

18:44I want to bring on Sean Nandy, a director at AWS, to give us his take on how companies can make the technology work for their bottom line as well. Sean, welcome back to the show. It's great to have you here. Good to see you. And I'm nervous you're going to ask me about my fitness preferences instead of AI for a minute, because I don't think I have good answers for that. Okay. I was going to say, you're welcome to answer that question if you want to start with. Maybe I'll check out your class. Okay. Yeah. All right. Well, so let's talk about AI, which I presume you're a little more comfortable talking about.

19:17You know, one of the topics that we've been talking about a lot here in the newsroom is how companies make money from AI. And I want to get your take on how you view this issue on the ground and how you're seeing customers grapple with that issue. Because from our view, there's a lot of revenue. Earnings are harder to come by. The costs are pretty high. How are you thinking about that issue? Yeah, look, I think that it all comes back to understanding ROI. And I'll give you a couple externalities that are throwing off the ROI concept. I'll start with that. When you're doing an AI solution, using Agentic especially, you have to think about, are you trying to remove costs from your business or are you trying to improve your revenue, your top line growth?

19:56Is this product driven? Is this adding new functions and features that people are going to pay for and buy? And we saw a lot of pain with software companies in the early days of generative AI, where we often refer to them as ISVs, but the software companies that sell various products to customers, they're like, I know my customers will like this thing. I'm going to build it. And they didn't necessarily think about how they're going to monetize it. Are they going to charge for a new SKU, a new product feature, or is it going to improve customer retention? And so not doing that upfront drinking sort of broke the ROI model.

20:29The second piece would be communicate that with your finance partners and teams. Is this revenue driving? Is it cost eliminating? When you have those metrics and measures, it becomes much easier to make that decision. The last piece, and this is probably the one you're going to want to unpack, is we talked on a prior show about inference and inference costs, and we've talked about all those pieces, the aspect of running AI. We see inference costs dropping by 90 % over the next year or two. And so I know that's been a hotly debated concept, but when we look at inference costs coming down, the amount of return you need to make something profitable changes radically.

21:04So lots of use cases are getting unblocked as you bring costs down. How do you see those inference costs coming down? What are the inputs into that? Because as you correctly pointed out, we in our newsroom, we've written stories about how the costs of not just the chips, but the models, right? I mean, the costs were coming down. They've kind of plateaued based on the data that we're seeing. I'm curious how you see that shaking out over the next year or two to bring those costs down. Let's talk about a couple elements. So when you think about AI today, and we often talk in the public about AI as a big broad thing, but there are many elements within it.

21:40One element of AI is actually building all these complex models. All the frontier model providers are putting lots of effort along with startups and building the next-grade model, tuning models, et cetera. And honestly, a lot of the AI consumption from a chip perspective, from a compute perspective, is driven toward training and tuning those models versus actually using them. And that's the change that we're seeing right now, the consumption is shifting to usage. And usage is what we largely call inference. So a lot of the reporting that's been done to date just looks at that broader AI landscape and says, hey, chips are getting bigger.

22:12They're getting more expensive. We're seeing higher parameter models. Models are increasing in price. Maybe even tokens cost more this month than last month. But if you look at what's changed in the Align footprint, we're seeing more and more diverse types of chips. You and I have talked about purpose-built infrastructure, chips designed for AI in different shapes and sizes. You're also seeing many different types of models. So at Amazon, we have our own Nova Micro model. It's a very small model used for specific use cases. It's text only. In the rise of Agentic, we're seeing these smaller models get used for simpler tasks, and that drives average cost down.

22:47And so one of the things you can look at, you asked me what the elements that make it up are, are you just using the single biggest chip or are you using different sizes of chips for different workloads? Are you just using one general purpose model that can answer any question or are you using smaller, more specialized models? And agentic really unlocks that. A couple more techniques I can tell you about, but that's sort of the landscape. Well, no, and this actually relates to the discussion we were having last time on the show about chips. And you basically talked about the routing of the inquiry or the request to using a specific type of chip, maybe not needing the best-in-class chip just yet given the pricing, using something that is a little more feasible or appropriate in terms of the request.

23:34I am curious, you gave that one to two year timeline. What are some of the challenges that you're seeing customers grapple with now then as they sort of work towards hitting that goal in the future? You know, I'm going to hearken back to a conversation we had many months ago on chief A officers and people things and so on. One big challenge is our mental model towards software development and product building is really changing. You know, you've heard about agile methodologies and other things, but typically you ask permission to build something, you build it, you get it out there, and you iterate on it for a while, and it survives for years, maybe many years.

24:08What we're seeing happen now with AI that's been really hard is you go build a use case, you're like, oh, this thing's really expensive. But if you iterate on it, and it gets 90 % cheaper in a month or two or three, that really changes the economics. And if it scales up and goes from being used by like 10 users to a million, that also really changes the economics and revenue flow. So I think actually the human side is hard, this rapid state of pace and of change. And the willingness to say, hey, this thing's not working. I'm going to fail fast and drop it and move to the next project. And that could be a cost efficiency project or a new product, right?