The Secret to Raising $9B

0:00What does it take to be a mega fund in the world of venture capital? Today, we're talking about a megafund you may never have heard of, Lightspeed Venture Partners, who have done a terrific job of flying under the radar despite raising$9 billion alongside some strong exits this year. In this episode, we're going to dive into what makes a megafund using the example of Lightspeed. We'll look at their business model, the startups they're invested in, the IPOs that they're leading, and how they're seeing significant growth while seemingly spitting in the face of the AI bubble. But first, we're going to discuss the new developments in the relationship between Amazon and OpenAI and the shifting allegiances in AI.

0:40This is the Newcomer Podcast.

0:49There was a time where one of the most exciting things that could have happened but didn't happen in Silicon Valley would be that Amazon would buy Anthropic. Amazon, a big cloud service provider, invested in Anthropic, doesn't seem to have figured out foundation models. But we've really put a nail in the coffin, I think, this week of that idea. Amazon just turned around and struck a huge deal with OpenAI. Tom, you cover these cloud companies, the foundation models and the deals they're striking. yeah what do you what do you make of amazon's decision to strike an enormous deal with open ai yeah well we should have some caution here that it's just sort of in discussions right now so i don't know it's in talks it hasn't closed yet yeah but i mean the origins of this basically was a couple of months ago amazon and uh open ai announced that they had signed a multi-billion dollar cloud computing deal with aws where it was going to be hosting some of open ai's models on AWS.

1:58And it was sort of the first big non-Microsoft deal that OpenAI struck once it got out of its exclusivity agreement. And, you know, in case you were worried after Amazon and OpenAI struck a deal that it wasn't circular enough, fear not. Fear not. You know, people always wonder with OpenAI, how are you going to pay for it? The answer is more money from the cloud computing companies that they work with. So it looks like OpenAI is in talks to get around$10 billion dollars maybe more from AWS and you know just to like take some pride in the way this has all progressed we did a story I think it was a newsletter item a couple of weeks ago where I talked about how Amazon had really started to shift away from its anthropic relationship but for a time anthropic models were the Amazon bet and that was the way it was discussed internally I was told Jeff Bezos was going around talking about anthropic as the Amazon bet and it's clear that they have shifted allegiances.

2:57A little bit behind the scenes, when I wrote that, AWS was very upset with me. And they said that to say that Anthropic is no longer, you know, our big bet and one we're committed to is absolutely wrong. And they pointed to all these press releases that Anthropic would make claiming that, you know, you know, Anthropic is still our primary provider. Sorry, AWS is still our primary provider. We use their chips. Yeah, that's clearly changed. That's no longer the case. There's been a shift there. Yeah. And I think, as we wrote at the time that newsletter item like i think there's been shifting allegiances across the tech world where once microsoft was the exclusive provider for open ai and then they invested or plan to invest billions of dollars into anthropic amazon obviously made their big play with anthropic now they seem a lot more attached to open ai uh google is kind of relatively neutral but more focused on their own models and i think it just goes to show that there is just such an insane need for compute right now and an overall belief in AI that there's no reason to pick a horse.

3:57You really kind of just need to bet on the industry at large and just take as much money and give as much money to keep it going. Amazon's also trying to get OpenAI to use its chips here, right? Yeah, that's part of the deal. Amazon is building chips that they want OpenAI to be using to diversify off of NVIDIA GPUs. Also in Anthropik deals of AI world too, they announced a couple of months ago, a deal with Google to use Google's chips. So Google has a partnership with Anthropik as well in this, even beyond, you know, focusing mostly on Gemini and their internal products. They do have some ties to Anthropik.

4:33Maybe Anthropik somewhat shielded Amazon first by doing stuff with Google. A Google deal. So then we have to circle back around here. Yeah. Yeah. Well, but also keep in mind that Google had already invested money in Anthropic before. They're a significant investor. And I actually my understanding is that most Anthropic training and inferencing, at least training I know, is done on TPUs, on Google's chips. So there was already like a deep relationship there. I mean, the fact with Amazon is that they're kind of like the rich kid that has to buy their friends. No one likes their chips. they just aren't very good i'm sorry i'll i'll say this strongly on the podcast i'll have anyone come on here and disagree with me but every conversation that i've had with people in the space says that like the tranium and inferencia chips that amazon makes are like a huge tier below certainly nvidia's chips but also google's tpus and so they gotta they gotta get people to use them somehow and 10 billion dollars is not a terrible way to do it and if there's a core financial engineering risk with AI, it's that I invest in you and then you use that investment to do me a partial favor by pretending like there's enthusiasm for my product, even though there isn't really enthusiasm, right?

5:52So in the case of chips, I, Amazon, invest in you, open AI, and then you turn around and use my chips, show a little enthusiasm, and then maybe I get a market premium on my stock. My stock goes up. That's great for everybody involved. And who cares, right? And similarly, this is the cynical case, the cloud companies benefit from open AI, Anthropic, et cetera, spending on these enormous training runs, right? So they want to keep this foundation model game going. And so keep the money going there also so it can turn around and spend money on us. So there's definitely this cynical argument that foundation models like OpenAI and Anthropic are so valuable in part because the people investing in them want them to be, because they're big customers and they're validating a lot of the products they're creating.

6:50And this is why, to me, this is the bull case for Google because Google's on both sides of it. If you think foundation models are going to be revolutionary and not only revolutionary, but capture a lot of the economic value of what happens in AI, Google has Gemini. If you don't and you think all these models are sort of ludicrously spending on training runs that are just like lining the pockets of cloud providers, well, Google has one of those too. So I think, yeah, good to be in both camps. Yeah, they have the chips, they have the models, they have the customers, and they can force these AI models down the throats of these customers by putting it in all the different, you know, crevices of their products.

7:30Sorry, that's a gross image. But I think, like, you know, this all leads to what I thought was a fucking hilarious video that Greg Brockman, the co-founder of OpenAI, just made where he, I don't even know what it was for, but it's basically this video where he talks about the importance of compute. And it's like, as time goes on, we just need more and more compute. The funnier part of this video to me was him saying at first, we tried other ways to build our models. Open AI did not set out with a thesis that compute was the path to progress. It's that we tried everything else and the thing that worked was compute.

8:09We really wanted to find another way to grow our company without just demanding multi-hundred billion dollars of compute deals. But we can't. This is what it is. So please, it's like a PSA. It was like for only a dollar a day. Yeah, he was getting quirky with it, you know. He was doing like a, oh, how do we do it kind of tone. I don't know. It was only for only a dollar a day. You too can keep open AI afloat and give us the compute that we need Please reach deep if you have GPUs if you have GB 200s If you have anything that you or your friends might have please donate them to open AI because that's what we need to keep going I don't know what the fuck this video was.

8:44I don't know who it was for I honestly if you start to look back at this era from 10 15 years from now If it does happen to go sideways or doesn't turn out the way people expected this video is gonna be one of them to me That's just like what were we thinking? How did we not understand that there was like something slightly concerning at the core of the way these companies run? To be clear, my point of view remains that we are imbuing everything with the capacity to be intelligent. And there's going to be a huge market for that. And as they get more and more intelligent, they're going to get even more valuable.

9:15But what that means for the near term value of open AI, I have no idea. You know, glad that's not my job to figure out. But I do think these things all come back to the actual, your belief on the value that AI is creating or not, right? If it's creating a lot of value, then it makes sense to build chips and spend money on chips. It makes sense to do these huge training runs. Right. Yeah. And I'll be fair to Brockman in his argument was that like our products are so successful. They're in such demand that we need compute to be able to provide power to it that, you know, we can build more of them because people want them so much.

9:55It's not that speculative. It's like, we need this now. It's for our products to work. My only counter to that would be like, what products outside of chat GPT? right you know like and what you know you you do these training runs and then everybody catches up with you tomorrow right well i think that's also that's that's the that's the business question that is completely separate from the value that ai this technology creates and is transformational technology i think we're all in agreement here that it is very useful and is very impressive and there is a ton of utility for it but where the business moat lands and how these deals as they are structured financially work out, that remains a little bit still murky.

10:38And as you'll see in the segment coming up, that's why they need all the money. That's what the money is for, is to make the compute, pay for the compute that OpenAI so desperately needs. And so everything has to be mega-sized from that. Up next, we talked about Lightspeed's$9 billion fundraising as a venture capital firm. So we'll dig into that. Stay tuned. Hey, folks, I'm Ammar, product and design lead at Google DeepMind. Have you ever wanted to build an app for yourself, your friends, or finally launch that side project you've been dreaming about? Now you can bring any idea to life. No coding background required with Gemini 3 in Google AI Studio.

11:15It's called Vibe Coding and we're making it dead simple. Just describe your app and Gemini will wire up the right models for you so you can focus on your creative vision. Head to ai.studio slash build to create your first app. to do. We're about to talk about the biggest, most important venture capital firm that maybe you've never heard of. Depends how much of an insider you are. Lightspeed Venture Partners just raised$9 billion, which even in the world of venture where we get numb to big pools of money, this is a venture firm with a lot of money to invest in startups. Here at Newcomer, we have published some of Lightspeed's old performance numbers.

11:59So we have a good sense of what companies they're investing in that gives the limited partners that back venture capital firms the confidence to re-up with Lightspeed. And Madeline has been doing some reporting on Lightspeed to understand this latest fundraising can help break down all the numbers. And Tom will weigh in with his spicy quips. Tom can weigh in with his hottest takes. I can really dig in on Lightspeed circa 2016. Yeah, exactly. You cared a lot. I mean, so if you are not in the venture business, you might have heard of Lightspeed because they were a big investor in Snapchat. Jeremy Liu, who we all know, invested in Snapchat.

12:41Then I think as Tom helped break, had a falling out with Lightspeed, right? Did you break that story? I actually didn't break that story. It was our departed former co-host on Deadcat, Katie Benner, which is a real blast. Oh my God. It was a very different era where the New York times would have written a story about like a fallout between a founder and the whole, I mean, we can't, we should get to your story, Madeline, but if you really want to, a trip down memory lane, the whole history between Lightspeed and Snapchat is a very interesting one. Not even just the falling out, but the way that they made that investment.

13:13So we can, we can talk about that then, but it's like a very different circle back to that. Yeah. It's like a very different. Lightspeed was and is mostly a firm that invests in enterprises, businesses selling to other businesses. There are exceptions. They're a big investor in a firm, Max Levchin's payment advance company. But a lot of the big investments are firms like Grafana and Navon and Fair, you know, that sort of are the backbone of or not consumer facing companies. Well, Lightspeed, I think, you know, wants to be known for their enterprise deals more. Their ties to Snapchat cover the press.

13:57And so people think of them as a consumer first fund and they're quick to tell, you know, we do enterprise deals. And to their credit, they had significant IPOs in the last two years. They were big backers of Rubrik, Netscope, and Navon, which all went public in this time of IPO drought recently. So they've had big exits in the last couple of years, which is impressive. If you say they're the firm that flies under the radar, they're returning money back to their LPs, which is great when you're fundraising. Madeline, you want to break down this$9 billion fundraise and give us a sense of where all the money's going to go?

14:32So this$9 billion fundraise is broken up across six different vehicles. It's all kind of multiple funds under this one umbrella to hit the$9 billion total. But the early stage fund is kind of split into Lightspeed 15A and 15B. And the way you can think about that is 15A. What marketing geniuses. They've got a number, they've got a naming scheme like Sony headphones here. Yeah. Well, 15A is 980 million. And like the title says, it is for seed up to series A, maybe a smaller B deal. Lightspeed B is$1.2 billion. And that is going to go to things that are more entering product market fifth growth series B onward side.

15:16However, it's not the growth fund. It's just like in the age of AI, these series Bs and series A's and billion dollar seed rounds, they wanted a lot of capital to be able to properly allocate in these rounds and double down on concentrated bets. And when that moves earlier and earlier with these AI rounds that everyone wants in, they wanted capital for those two, wanted capital to back those. So that's how those are structured. So those are two earlier stage rounds. So those are the two early funds. There's also Lightspeed Select 4 and Lightspeed Opportunity Fund 3. Select 4, a$1.8 billion fund, that is for doubling down on bets that they want to get a higher ownership share in that are more growth stage.

16:02So think of it as a kind of medium growth fund. And then Opportunity Funds have been actually, the past Opportunity Funds was where they first backed Anthropic because they did not get into Anthropic until the Series D. But then they ended up co-leading and then leading later Anthropic rounds as they kept raising. So they did get a larger ownership share later on. But that's what the Opportunity Fund would be to double down. Which right now seem like great investments. I know. They've been marked up a lot. There's been this chatter in the zeitgeist right now about if you're not in these key AI deals, you're failing behind.

16:38But they're mostly mentioning Cursor, Merkur, OpenAI. high. Lightspeed is not necessarily in those companies, but they are a pretty significant shareholder in Anthropic, which for, as you can tell by now on this podcast, we're kind of fans of Anthropics style at this point, you could say, in how they do things. And it's a strong growing business with top models. So the two new funds they raised with this fundraise are the Lightspeed Co-Investment Fund, which is about$600 million, and a single investor vehicle that was$1.25 billion. Those exist for existing LPs to tack on more money to double down on specific bets.

17:20So those basically exist because of how big AI rounds are getting. There are certain LPs who want to have even more exposure to top AI deals, and those funds exist so they can opt in to give Lightspeed even more money on ones that they think Lightspeed has an advantage of getting into further rounds on and getting additional ownership. So those are what those two additional structures are, which total it up to$9 billion, which is pretty hefty fundraise. Was one of them, so sorry, one of them is a sidecar? There are two kind of sidecars. Like are these real funds? These are real funds, but they basically exist because certain LPs wanted to double down more on later stage deals.

18:03And it's a way of balancing the pro rata rights in those deals to where the LPs wanted to participate more. So all of the different terms across some of the, you know, some of the more favorable to, but you have to be an LP in the early and growth funds if you want to get into these special two additional funds. So it's the same LP base. What's interesting to me with this and like, I don't have the perspective, but like do other funds portion out their funds or do other firms portion out their funds in such a regimented segmented way? or is this like basically a new marketing scheme for lps where you're just like you can have access to the select fund but if you want further exposure to later stage rounds you can be select plus you know the sidecar whatever they call it like is this all just kind of a different way of divvying things up so you can essentially get more from an lp well there's a lot going on that we don't know but one reason you'd want separate funds is venture how are venture capitalists compensated I think that's worth a discussion here, right?

19:02So venture capitalists are compensated based on management fees and carry. Management fees, that's the two of the famous expression, two and 20, means that they get 2 % a year of the assets that they've raised. So you have bigger and bigger funds, you get larger and larger management fees. And so part of what's appealing about mega funds is that you have so much money that the management fees alone become very valuable. And who cares how well your investments do? Because you can make, you know, if you do the, we should do the math off 9 billion. I mean, it depends. Not all these funds are the same fee structure and we don't, they don't like announce the fee structure, but they're making hundreds of millions certainly in management fees alone.

19:50Now you have to run a bigger firm, you pay employees. Right. The overhead goes up among the number of partners, that sort of thing. So it does get diluted out, but it's a pretty good market if you can make sums like this off your management fees. So if you're thinking about management fees, you want to have a big, you just want to raise a bunch of money, right? That's great for management fees. On the carry side, you like having different funds, right? So if my venture fund does really well, and I only get paid as a venture capitalist if I double the value of my holdings or if there's some hurdle that I have to reach, I don't want that to be dragged down by the fact that then we double down later at bad valuations.

20:35And so we have a bigger and bigger pool of money we have to pay out until we reach our hurdles. So to hit carry, it can be advantageous to split out funds. Split out funds meaning like having early, having late, having a middle tier. Right. Early could be in the money, but opportunity might not. And so then it's like, oh, we make money off. Whereas if early and opportunity were the same fund, it's possible we don't make out. And LPs know this. And this also could be in Lightspeed's case, they have 15A and 15B, which even splits the early level farther. So there's like seed A, and B, C. So before you even get into growth, they're tiering it out that way.

21:21It's possible they have some overall promise to LPs. I don't know how much these things are operating standalone versus they're saying overall we will make the money back but but yeah i don't i don't think it's mostly about marketing and they're probably let you know some lps concentrate in areas they like better and you know there's there's a lot of negotiating but you don't think there's a way that you can you know get like more money from an lp by offering them exposure to multiple funds within the firm. Like it just feels like it's like one. Yeah. It's just like one chunk of money going to one fund versus like, you know, it's just like a psychological thing of little bits will add up to a greater number, but it doesn't feel as psychologically, you know, like a huge chunk that's coming out of your fund.

22:12Like you're, you're, I don't know, you're doing different strategies. Yeah. Yeah. It just feels like there's an appeal there. It also, if you wanted to get a larger ownership share of one specific winning bet, which seems to be what everyone wants to do in the AI race, right? Like everyone wants to be an open AI, everyone wants to be an anthropic with these kind of winner companies. You could have exposure as an LP from the growth fund, but you could also be in the co-investment fund and get more that way. So you could say, you know, they could sell it like, Hey, if you want to get even more in anthropic, you could opt into this additional fund and it's doing so well.

22:46Why wouldn't you do that you know i think this new fund was broken in the new york times uh and the way that the times frame the story was like ai investment is going crazy right now you need to be able to raise huge amounts of money in order to participate in these mega round seeds or you know hundreds of millions of dollars and billion dollar valuations right off the bat and this is just a response to the size of rounds that AI companies are demanding. Is that basically what's going on here when you see Lightspeed doing this, you know, astronomically large round comparatively to the other things?

23:21It's just like, this is AI, this is the bubble, this is just like the table stakes for participating? I think, you know, there are two things going on in terms of why Lightspeed is able to raise this amount of money. One, limited partners, the people and firms and institutions that invest venture capital firms are sort of fleeing to the best of the best. Related to the second phenomenon, which is companies are staying private forever. And so there are these clear winners or apparent winners, companies like OpenAI, Anthropic, Cursor. Databricks. Clean. Databricks. Didn't Databricks just raise like a Series L recently?

24:01Yeah, it's like 160 billion or something. I had never heard of a Series L before that. Right. So they're staying private, which means that only sort of connected investors can really get access. And so firms like Lightspeed are well positioned to do those deals. And so they need way more than traditional venture capital money to invest. But what about AI specifically? Like, is it just, I mean, is it, are we seeing rounds that are fewer and larger? Like they're not really passing the hat around that widely. And so you just are entirely reliant on, like you're saying, these big name firms to give you allocation.

24:40You just don't have that much option as an LP. Yeah, that's basically part of this too. There is hardly ever been, at least in my time reporting VC, just a clear kind of consensus view of what to invest in from all these firms, right? People have collectively decided that there are the winners in AI, not for applications, the jury's still out there, but on the foundation model level, a lot of people think there's still a ton of value in these companies. There's not that many of them. They're very expensive operating companies. So the money they need to raise and keep growing is more. And so then the big players need more money to fund them because there's a sort of consensus view that these are the companies that we need to be backing right now.

25:23So it's kind of reinforcing. Yeah. I mean, you know, there's been some, it's almost like a meme at this point, but I've just seen it enough on X to decide this is like, I think people are discussing that there are like only so many firms or sorry, there are only so many startups that really matter right now in AI. And the ones that typically get named are like Anthropic, Cursor, OpenAI, Mercor, for whatever reason, gets, gets slumped with this group. Yeah, people love Mercure. Okay, I get it. I mean, you know, there's value in sort of, we basically do like reinforcement learning in human kind of, they're basically like replaced scale as like kind of the go-to service.

26:04So sure, whatever. I don't really see those as tech companies, but they certainly have venture backing. But you basically, if you're an LP, you want exposure to those companies and you're going to get pissed off if you don't get it. And so like the value that all of these firms basically have is say like, oh, we can get you in it. Don't worry. Don't worry. We'll get you in there and we'll get you in there in a big way. And it is incredibly concentrated. I mean, I don't know, Eric, like you were, you know, covering late stage venture in like the Uber era when there really weren't, you know, it was Uber and Lyft.

26:33I mean, was it the similar sort of competitiveness where there was just like a small number of firms that had access to those crazy Uber rounds. In the hedge fund world, people think it's sometimes crazy that it's like you can raise this money with great fees and then you turn around and you invest in Meta, in Alphabet. It's like anybody could do that. It's like, what, you're a genius? You pick which one of the seven most important companies and go big in it and then you collect fees off of that? And I think there's a degree to which a only marginally more sophisticated version of that is now happening in venture capital, which is, oh, you're going to invest in open AI?

27:14Now, VCs have the advantage of saying, well, you, schlub, can't get in open AI. Only I can. I don't totally, even though I follow this stuff very closely, it's hard to come up with a non-cynical reason why these companies gatekeep their investor base and therefore allow their investors to basically collect fees on what are, you know, you know, OpenAI is desperate for any money it can get, right? So why does it restrict the investor base? I think the answer is just you create FOMO, you create this like sense that I need to get in by making it a little bit difficult by having prestigious people invest.

27:56And the people investing are, you know, bullish optimists who have an incentive to run around and say, this is a great company. And so instead of just doing a fair market price, you sort of restrict the investor base and you create FOMO. And if you wanted to do it the normal way, you go public. So yeah, it's been this weird situation where insiders are getting a pretty good deal. Now, sometimes they get it wrong, right? If you dump all the money into the wrong one of these winners, sometimes it doesn't work as well. There are certainly companies - risk, right? And the traditional venture model was that you had a bunch of different bets that you based on your research that could grow.