Hemant Taneja on Global AI Investing & Energy Needs, AI Team Inefficiency | Sep 24, 2025

0:28Splash by

0:34Welcome, everyone, to The Information's TITV. My name is Akash Pasricha. It is Wednesday, September 24th. We have got a big guest with us here today. Hemant Taneja is the CEO of General Catalyst. He is here coming on in just a minute, so do not go anywhere. You do not want to miss that conversation. We've also got our friends at Atlassian coming on for a discussion about why individuals are getting value out of AI, but teams are not, at least not to the same extent. It's going to be a great show, so let's get right on into our first guest. General Catalyst is one of the most storied firms in all of venture capital, and Haymuth Deneja has helped steer the company for much of its 20-year history.

1:14He has also been the CEO of the firm for nearly five years now. The company's portfolio has included Stripe, Circle, Anduril, Canva, and Mistral. I want to bring on Haymuth to talk about this moment for the AI sector and how his company has evolved over the years. Hey, welcome to TITV. It's great to have you here. It's great to be with you. Thanks for having me. So look, there's a lot to get to. I want to sort of start with a bit of my confusion about the current state of the AI sector. And one of the things that you've said time and time again is that we are peak ambiguity right now. And on this show, we try to make sense of the ambiguity, make things a little less ambiguous.

1:53And so hopefully you can help us do that. I want to start with a column that our executive editor wrote last night. Martin Pierce, he wrote saying that, you know, Sam Altman put out this blog post saying, the growth of AI services has been astonishing. Okay. And that's the first part of the story. But then on the flip side, you have, you know, a lot of commentary around how businesses are not seeing the value from AI as they would have hoped, not the least of which is that MIT study. And so you kind of have these two different sides of the story. and I'm hoping you can help us reconcile these two things.

2:29Yeah, look, you have to think about what it's going to take to create diffusion of AI in the enterprise. Let's work backwards from that, okay? For an enterprise to really effectively use AI, you have to have four things in my view. One is you need to make sure your data infrastructure is ready for AI. That's a complex problem. It doesn't get solved overnight. Remember, the old Transformer zeitgeist only happened a couple of years ago. You also need to have models trained on your data. And so they're applicable to your secret sauce, if you will. And the last thing is you have to think about workforce transformation because you're going to have agents, AI agents and humans in your workforce now.

3:08And some agents are managing the humans. Some humans are managing the AI agents. And so it's like a whole transformation. And the last thing is you need leadership in the top desk of the courage to drive that diffusion of AI. So a lot of what you see in the MIT study and others is, you know, these companies take a shot at trialing one of the language models. They can see what the art of the possible, but then they get stuck. And they get stuck because it's actually a place where you have to have all four of these ingredients to really drive diffusion. So, okay, so you're saying that businesses need all four things, and perhaps that's why they're not seeing the value.

3:45You know, the other part of the story that was kind of interesting is there was a report from Bain that came out yesterday that talked about the extent to which businesses would need to spend more on IT services to justify the capital expenditures that these cloud companies and AI companies are spending to invest in capacity and compute. And the takeaway from the report, it was quite interesting, was that, you know, they're not projecting enough revenue growth to justify that capital expenditures. So how do you think about that, if that's the case? Look, I think the equation becomes hard if you think about the market size in the context of software budgets of companies, and that's really what you're going after.

4:29But if you start to think about the fact that the ultimate revenue opportunity is the labor budgets of these companies, that's when it really starts to make sense. Because if these models are going to actually effectively diffuse in the enterprises, they're going to take a lot of the labor content away. Now you're talking about a much, much bigger market in terms of what is going to create profitability around these staggering numbers. I do agree with you. These numbers are so large. But the belief here really is that the budgets you're going after ultimately is going to be around labor, which obviously creates a whole other set of issues I've talked about a bunch.

5:04But that's where this is headed. Okay. And so when you say labor market, say more about that. What exactly do you mean by that? I mean, so take all the innovations in the coding market, Cloud and other products, are you really going after what is the software tooling budget for developers or are you going after how much you spend on developers as a company? Right? So the market size in terms of what your revenue opportunities when you're providing those products is 100 times bigger. Let's say you spend a percent of somebody's salary on their tooling. You're actually going after that 100 % because as these models get better, you're starting with entry-level engineers all the way to over time, what we think of as 10X engineers, these AI models are going to be able to do it.

5:47And so that's your time. And so now all of a sudden the justification for how much infrastructure you're going to spend to create this market starts to make sense. Obviously, there's a big leap in there that these companies are truly going to diffuse these technologies, but the raw potential in terms of scope is there eventually. But then what happens to the labor itself? What happens to the people that were doing these jobs? I mean, there you go. A really interesting thought experiment that one of the CEOs of consulting companies you would know told me is a client asked them, how do we go from 50 ,000 employees today to 100 ,000 in five years, but only 10 ,000 of them are humans?

6:26And so I do think companies that effectively diffuse AI into their overall operations will be market leading and will have fewer people than they do today. And that is where the budget opens up to subsidize the staggering compute infrastructure investment that we're talking about. So, and, you know, I want to come back in a little bit to general catalysts and sort of the strategy you guys have. But very quickly on that note, you guys are about 300 people right now at the firm? A little over that, yeah. Where do you think you'll be three years from now in terms of headcount? You know, it's very hard to say, but I mean, I'll tell you, we are fully replatforming ourselves to be AI native.

7:07There's a massive initiative inside the company. And as I think about our priorities for next year, I'm pushing every part of our organization to say, why do we need to hire more? How do we get more out of making our teams do the interesting work and take a lot of the boring or entry-level work and have the models do? And we're early in our journey, but I'm hopeful that we will definitely bend the curve of how much we have to hire to do our business. Now, ours is sort of a unique business. It's not what most of the enterprises look like. Companies in our industry are not headcount-heavy to begin with, but very much focused on figuring out how to bend the curve.

7:48So, I mean, 300 now, three years from now, are you, is it 400, 500? Is it flat? Probably, if it was up to me, if we actually use AI effectively, I think it shouldn't be growing more than 15 % to 20%, even though our aspirations as a business are to grow far greater than that. So I think you start to see that efficiency with AI. And so you're saying that you won't be hiring as quickly as you use AI in the organization? As long as we're effective in deploying AI. Just like all those issues I described for everybody else, we have to go through the same challenges. So one of the other parts I wanted to ask you about is, coming back to the news of this week, we're seeing a lot of investment in capacity and compute.

8:32and there's a lot of money flowing around. And one of the things that you mentioned, you were on CNBC yesterday, you talked about sort of the bumpy period that could come up as a result of all this investment in capacity and the likelihood that in a few years, we might be in a place where we've over-invested, we have over-capacity and there might be some problems that come with that. What exactly is this bumpy period that you were referring to? Talk about that. Look, I think we went through this with the CLAX when we were putting the telecommunications infrastructure in place as well. The reality is when you have companies trying to go on a market, they will over-provision the infrastructure they want from their vendors.

9:13So if you think about the clouds today and the amount of data center capacity they're provisioning for and energy capacity, they will over-provision because they don't really know how that sector is going to scale up to meet their needs. And you'd rather, and these are not all deeply committed contracts, but the infrastructure that's getting built. So I think in some cases I've heard there's two to three X more overprivileged in some of the hyperscaler plans. And so if that's indeed happening, then there's going to be a period where we're overbuilt. And you'll see some of those infrastructure investments not go very well, but over time, same thing's going to happen with the computer infrastructure that happened with the internet infrastructure that will grow into it.

9:57Because there's long-term, obviously, what's getting built today will get dwarfed in the context of what we'll have 20 years from now. There will be this period in three to five years where we'll feel like, gosh, there's oversupply of compute infra because the industry as a whole over-provisioned. It's hard for me to not imagine that scenario. So, okay, so let's dive into that scenario, though. So in this three to five years from now, let's say we have all this compute that isn't getting used because we built all of it. Who gets hurt there? I think there's a lot of infrastructure capital going in there.

10:31There's a lot of data center development companies that are popping up. There's a whole industry that's popping up around this. And I think those companies are able to effectively navigate that period from a liability standpoint. This movie has played out before. Right. And I'm just trying to understand here who should be most worried about this. It sounds like it's the companies developing these data centers who should be most worried about it. What about, I mean, the big tech companies, look, they can afford or they have ways of affording some of this investment. What about startups in this bumpy period?

11:09What happens to that? Yeah, look, I think the startups, look, the infrastructure, the whole infrastructure industry that's coming up is the one that has to navigate how that infrastructure drives utilization, you know, to justify the investments that are getting made. I think for startups, that issue isn't so pronounced. I think they will go provision capacity, the rent capacity that they need in the context of their business is slowly growing to it. I think it's more the infrastructure companies that are provisioning for these large hyperscalers and labs. And then, you know, what's the actual demand going to be in the end?

11:48You don't really know. let's talk about you know sort of projecting you know this idea of of over provisioning or building over capacity as you said and this is not the first time we've seen it i mean you know one sort of question i have is a why do you have to build over capacity in the first place and the second thing is why don't you just narrow your ambitions a little bit you know instead of building capacity for all of AI, why not stick to sort of one particular space of AI, like healthcare, for example, you know, where we know that the use case might be a little bit more proven out. How do you think about that?

12:25Look, I don't think the technology company is going to, you know, marginalize their ambition in that regard. I think it's much more what, if you're a hyperscale trader, you just don't know how this infrastructure set of companies their ability to deliver. So you're just sort of over-provisioning, not because you will overspend in the end, but it's because you want to make sure you have the compute that you need to really meet your needs. The idea that you'll see these industry-specific models come up on top and provide solutions that are vertical solutions, that's already happening, right? We incubated Hippocratic AI, we did physics, actually, industrial.

13:04There's a lot of these companies that are already starting to sort of leverage the models and build on top of them solutions that are suited for those industries. Going back to my framework of the four things for enterprise transformation. I think that's naturally happening. And that to me is in some ways the best places for the AI talent in the startup world to go tackle. Because those are somewhat more tractable problems to build a company around from scratch. Right. And what about the environmental costs that would come with building this overcapacity, as you say? Look, I think if you think about the AI compute opportunity, it goes strapped heavily to the energy opportunity.

13:40Right. And the trade-off you have there is in the short term, if we're going to win an AI, you actually don't have an option but to lean into natural gas as the most abundant resource. That's really what we've got the most of in a lot of ways. And obviously, in the 10 to 20-year period, you want to see the renewable infrastructure with baseload solar, really scaling to take a lot of the natural gas energy workload, you know, from a scaling perspective. And long term, obviously, we have these moonshots with, you know, fission and fusion that folks are working on. We're invested in Pacific fusion in that regard.

14:16But that's what to me is, you know, could that make a difference in 20 years? So I think there's like a, today, can we use natural gas as cleanly as possible, medium term? can we really get renewables at scale in a reliable way for this industry? And long-term, can we drive energy abundance with nuclear? I mean, that to me is the framework. So when you say that the AI opportunity is the energy opportunity, I mean, do you see yourself increasingly investing in energy startups? Could you see yourself setting up an energy fund in the future? What would that mean for your business? We're heavily looking into it.

14:53And I've invested in the - No, no, no. I've invested in the Cleantech 1.0 era very unsuccessfully and have a lot of learnings from there. So I think to me, and we've started to do some investments in that space. As I mentioned, Pacific Fusion's example of that, Fourier is another one that's a hydrogen generation company, onsite generation of hydrogen. and so there are solutions we're going to slowly invest in but the act of company building and scaling those is still a new art from my standpoint so we're being careful in how we move and I do think if we think comprehensively and take a step back the opportunity is as much about can you create the right technology solutions as it is about what is the roadmap for scaling the infrastructure so that it meets the needs of computers so we're taking a step back and really comprehensively looking at it and see how do we help founders tackle that space.

15:47Right. What do you make of the current policy in the U.S.? I mean, you talked about renewables, right? We've seen this sort of rollback on a lot of clean energy initiatives in the U.S. How do you square that with your comment that there's a lot of opportunity in the renewable space and in clean energy? Look, first of all, my framework for advanced energy that really will get deployed at scale is got to be clean, affordable, secure. It can't be just clean, but inexpensive. And I do think we need to use this new demand to drive towards sustainability. So I do believe in that. But in the short term, you know, if the US is faced with the choice of be environmentally friendly or win AI, which I think is really the choice that's in front of us, I think the cost of missing AI is too large to make that trade off.

16:40And that's really what's happening here. If you think about most of the growth in the US, most of the investment, it's all... The United States is making an enormous bet on artificial intelligence, like a venture bet we're making in what creates our next phase of prosperity, when you really think about everything that's going on as a country. You know, it's exciting and it's scary, just like investing in a high-risk venture capital project. And so in order to really make that come to bear, you need energy. And so you're not going to slow it down. So I can see why the administration is rolling the regulations back.

17:16But I do think in the long term, if we're not, while winning AI, also making investments with that long term, we do have this clean, affordable, secure infrastructure. In the end, we're not going to win the long term because there is a cost to carbon in my view. And I do believe in sustainability matters. And so how do we make those short-term, long-term choices very intentionally? At least we're using that framework to get our own investment strategy into space. So I just want to understand this. So you're saying these policy rollbacks with clean energy, your stance on it is, you said you could see why they're doing it.

17:54So are you saying that you're okay with it because we'll have to worry about the environment in the long run? Is that what you're saying? No, no, no. Look, my own view is that we should not lose momentum on sustainability. but I think because the rest of the world follows us as well and I do think it's going to slow things down but I understand why that's happening and I'm sort of looking at the environment to agree if that's what it's going to be how do we make sure we focus on the short term win in AI that we need I'm sort of putting my being an American hat on for a second but in a way that we're also investing in this infrastructure that is economical because it's got to be economical for it to be used to scale.

18:38And the good news is we have so much demand in energy that when you have new demand, you can absorb innovation. And so can we intelligently create an energy transition roadmap for the country that also allows you to win an AI and make those straight-offs? I think that's a lot of the way I think about our investment strategy in this state. And before we talk, I want to talk about Anthropic, which you guys have a large position in. China is pulling ahead in renewables. What do you think that means for us? so look i think i think we um we underestimate china in a lot of ways i don't think they're that behind in uh ai i think they're going to be very pragmatic in the way they catch up in the chips race you've seen the wabi announcements for example and they're very far ahead in uh energy i mean a while back i used to say they're building a coal plant a week but they're also doing the largest deployment of uh renewables i think they're actually getting these technologies all to a point of viability where the energy advantage will also give them an AI advantage.

19:36So this is one of the reasons I'm saying that we can't lose sight of the fact that we need to create these new sources of energy and enable them at scale in the US because in the long term, if we're going to win AI, we do need that transition. We are going to run out of gas. And so we have to make those investments today and why I think it's important for our industry to be backing founders in that as exciting as all the focus is on AI infrastructure, and that's where most of where we're focused as well. I think it's very important to have some of your attention on getting this ecosystem to be scalable, affordable.

20:14I want to talk about Anthropics. So we've seen OpenAI now coming out very loudly saying, we are going to build our own capacity and we're going to throw a lot of resources at investing in data centers. Does that kind of up the ante for Anthropic? I mean, do they need to be doing the same thing here and investing in capacity to the same extent? Look, one thing I'm very impressed with, with Dario and the overall team at Anthropic, is that they're very focused. They actually don't pay much attention to what others are doing. They have a lot of confidence in their own strategy. I think they've been very capital efficient.

20:50If you think about the amount of enterprise value they've created in terms of the capital they've spent, the surgical sort of investments they made in the products that they care about. So I don't think they will be influenced by what others are doing. So if they decide to do it, they've got their own copy. I think that's a team that's very much an independent thinker and at peace with their own strategy and their own values. And I admire that about them. I think they're executing very well. But you don't think they're going to move into building their own data centers in the near future. how do they keep up with their own compute needs then?

21:28I didn't say that. I'm just saying, I think if they think they need it, then they'll build it. But they're just not going to react to what others are doing is my only point. So did you see that coming in the next year? Let's see. I mean, I'd let them dive. I want to be respectful of letting them talk about what they want to expose. Right, right. I'm just trying to wrap my head around, you know, we see one giant pursuing this strategy, You know, I can only imagine, you know, what that would mean for another. But I take your point that, you know, you want to focus on your own game. Let's talk about general catalysts.

22:04This is such an exciting time for the company and the firm. What is the strategy behind all these expansions into buying a hospital chain and moving into wealth management? You know, just talk about what you're trying to do here with this strategy. Yeah. Look, our true north is to help founders build enduring companies, right? Companies that are market leaders and are scaling and compounding for a long time. So one way to look at it, everything I see coming out of GC is, you know, what is that doing to GC the business? But the more important way to look at it is to say, how does everything enable a core advantage for the founders that we back?

22:45Okay, so when we focus on a deep and sincere effort in building the best relationships with ambitious people around the world, our familia, as we call it, that catalyzes opportunity for our founders. If you think about all the innovations we've done in our capital stack, that's because we think as technology company building matures and scales, the innovation required in our industry is not at the three axes of state sector and geography in terms of equity funds, which is what the industry has traditionally done. It needs more capital solutions like our customer value fund. We created for funding sales and marketing and our creation fund for funding a lot of the inorganic and the roll-up work, as we've talked about in the past.

23:33So that's the capital innovation for the founders. If you look at buying a hospital, if you look at some of the other things that we're doing there, or you look at our General Catalyst Institute, it's to help make the founders be more effective at engaging with governments and policy. or having access to distribution. What's going to happen with our 23 health insurance partners that we're working with on their transformation? It's an opportunity for our companies, our founders, to go be part of an ecosystem that can drive that transformation in these complex industries and really act as a system.

24:06So that's the distribution advantage. All those things. So every time you hear something that sounds muddy, go back to how does this really support the founders that GC backs? So all of this is to give you more knowledge, more tools for the founders. Okay, got it. So, you know, as GC sort of spreads its wings a bit, I have to ask the question. It's come up time and time again. We're not. We're not. We're not. We're not. You should not know you to ask. You didn't even let me ask the question. Okay, well, I got it. So GC is not interested in going public. GC is not going public. Okay. What about Andreessen Horowitz?

24:55Do you see that happening? I think you have to ask them. They're an amazing firm and you should ask them. Do you think it happens? I actually don't know. I really honestly don't know. I don't know in my own opinion. The broader question I have here is, you know, this idea of venture funds going public. You know, let's just broaden it out. I mean, do you see any venture firm going public in the next three years? Do you even think that that makes sense, you know, in sort of this next phase of venture capital that we're on? Look, I think, so let's think about it from first principles. If somebody is going to go public, this happened in PE.

25:36I think platforms have been created and they do go public. So I think, could somebody go public? Yes. Is somebody going to go public in the next three years? the most well positioned probably is our friends at Andreessen Orvis, but you have to ask them. I've never had that conversation with them. No idea. And it certainly is not going to be us. I'll tell you that. It's just not our priority. We're really trying to innovate in going back to the peak ambiguity. So we're really figuring out what are all the unfair advantages we can create for our founders to come out of this peak ambiguity with scale and success.

26:08That's ultimately what's driving our our innovation roadmap. I mean, this whole notion of venture capital firms becoming more like PE firms, what do you think of that comparison? I think it's, it never makes sense to me. And I'll tell you why. Why would we want to be like PE firms? The best companies, the biggest companies, the most profitable companies are in the venture capital portfolio, not in the PE portfolios. And the companies that are the most resilient and relevant for the future are in the venture portfolio, not in the PE portfolios. They all have to go through transformation. We're actually helping a lot of the companies in the peer portfolios become viable with a lot of the enterprise transformation stuff we talked about.