Raghu Raghuram: AI, Robotics, and the Rebirth of Infrastructure

0:00We had an offer from Cisco, and it was to basically put a hole in the backyard and bury the company. And so I took the long walk up to Raghu. Do you remember this meeting? Yeah, yeah, I remember the meeting. And I'm like, listen, we have this offer from Cisco. We would love to be in the hypervisor. And you know what Raghu said? No. You understood leverage at the time. Yeah, I'm not a super genius, but I'm not that dumb. If you want to understand how enterprise software really gets built and scaled through multiple platform shifts, there's no better person to learn from than Raghu Raggerom, A16Z's newest managing director and general partner.

0:36In this episode, A16Z co-founder Ben Horowitz and general partners Martin Cosado and David George sit down with Ragu to talk about what it really takes to build and reinvent great companies across generations of technology. We go deep on the lessons from Netscape and VMware, the strategy behind one of the most successful acquisitions in tech history, and how AI and robotics are transforming infrastructure all over again. Let's get into it.

1:03Raghu, at Netscape, you met Ben and Mark for the first time? Indeed. So what was that like? And what did you think of them at the time? And how are they now compared to then? Wow, loaded question. It's a good thing. Ben, I don't think you rescind offers, do you?

1:23By the way, Mark's not on the pod, too. Yeah, Mark's not on the pod. So just keep that in mind. You're a lot more mellow now than before. There are some circumstances. That is also true of him as a board member. Yes. His bedside manner has improved. Yes, has improved. In my defense, I would just say that. We were in this very weird situation where we came out with the browser, and it was like the biggest kind of software hit ever at the time. It put everybody on the internet. And then Microsoft, who had 97 % market share on the desktop, So basically a complete monopoly. Macs weren't a thing or anything else.

2:02And they decided their whole mission was to put us out of business. And we were selling the browser for money at that time. It's$50 a browser. And they did things that, like, I really haven't seen since. So they would put bugs in Windows to break our client, our browser, so it wouldn't work. they set their download thing on their downloader that had a bug in it between like 13 megabytes and 13.3 megabytes, and our download was like 13.1 megabytes. It did stuff like that. The craziest was when we cut a deal with Compaq, which was the biggest PC vendor to kind of bundle Netscape with all the Compaq PCs, and we did a big announcement with them, was front page of the Wall Street Journal, above the fold, the whole thing, which was a big deal in those days.

2:56And the next day, Compact called us and said, you have to let us out of the deal. And we're like, why do we have to let you out of the deal? It's like a great deal for everybody. And they said, Microsoft is withholding Windows 95 from us unless we break the deal. We'll go out of business. Like that type of tactic. So we were under a lot of pressure, just long to say. And so then the strategy at Netscape that I hired Raghu into was basically to replace all the browser revenue with server revenue. But the problem is we really didn't have any server products. And the browser was, you know, when it peaked at like$250 million a year.

3:39And so it was a lot of revenue to replace in a very, very short order because it went from$250 to zero over a two-year period. and so anyway so poor Ragu gets hired in by me who was feeling very urgent at the time I would just say and you came in right out of grad school right? you were straight out of grad school yeah okay yeah but I didn't have time for a lot of that slow training type of stuff it was all like very like military style training it was wartime all the time yeah what are some of the stories of the war that you were in? no I mean I think if I think back Ben was on top of everything like you wouldn't believe.

4:20You couldn't miss a comma in your data sheet and Ben would catch it, right? It was crazy. People used to dread going into reviews with this guy. But the second thing, though, in his defense, because we had no server products, I think you hired, what, six or seven product managers at the same time. None of us knew any product management. Some people are like one or two years in the industry. Some people are like me. so remember there was this paper floating around on the internet for a long time still is good PM, bad PM of course it's Ben, that's yours right? Ben wrote that I wrote it for those six guys it wasn't meant to be published that was for you it was your training document congratulations so that is actually the second thing to say about Ben systematizing organizational behavior.

5:15I think all of that is, that's true of him then and true of him now, obviously, now he's written books about it. All right. So I want to fast forward a bunch here. So obviously we have Martine, Ben, Raghu. We got to talk about Nysera. So Martine, obviously you're the co-founder and the inventor of SDN. Ben was the investor and board member. And obviously Raghu was the champion of the acquisition at VMware. So I'd love to hear the dynamic across the three of you. I mean, listen, the short version is that Ben saved my company from going out of business by investing in it and then helped run it. And then because we love Raguse so much, we gave him a big discount on a great company.

5:56Oh, here we go. It was a premium like none other. I love that it's positioned as a big discount because it was like the highest priced acquisition of its time. Yeah, I think we had. It was, but if you look Look at the revenue numbers that thing did after the fact. Raghu got it very cheap. Yeah, that's right. In retrospect. Very cheap. But it was weird because, yeah, we only had, I mean, I don't know. We didn't really have, I don't even want to say what our revenue was because I don't know if it was really even revenue. We had some checks. Yeah, yes. We had some checks. Yeah. And by the way, so part of the story is we had an acquisition offer from Cisco.

6:35And we didn't want to sell. And so I thought, if we don't get into the hypervisor, what we sold is we sold networking that would go in the hypervisor. But like the entire hypervisor market was VMware. And so I'm like, well, if we don't get in the hypervisor, we kind of don't have a market. I'm like, there's one person that controls the hypervisor, one person. And that was Ragu. So I took the lonely trip up to see Ragu. Hold on, this is when you had an offer from Cisco, though. We had an offer from Cisco, and it was to like basically put a hole in the backyard and bury the company. And so we knew that because it was threatening hardware.

7:06And so, like, I took the long walk up to Raghu. Do you remember this meeting? Yeah, yeah, I remember the meeting. And I'm like, listen, we have this offer from Cisco. We would love to be in the hypervisor. And you know what Raghu said? No. You understood leverage at the time. Yeah, I'm not a super genius, but I'm not that dumb. It wasn't a nuance. It was, no, that's at the time. I thought we were basically done. We'd have to sell to Cisco. So then how did it progress from no to? Ask Raghu. By the time I'd gotten home, Shikar had called me. So there must have been some internal conversation. No, so part of saying no was because we wanted to acquire the company as well.

7:43Yeah, it was no to the BD deal. Yeah. You have to understand. So it was kind of, hey, we can gang up and compete with Cisco. And Ragu was, no, not gang up like that. Gang up like I will swallow you. Yeah.

8:00I'm buying the cow, not the milk. Yeah. Yeah, but I will say, there was no scenario in the world where it made sense for me to have a big market for Martine. Of course, an independent, yeah, of course, yeah. But more positively, we knew we had to get the next product that was going to sort of rule the enterprise like the hypervisor did. And this clearly was it. And Pat Gelsinger, to his credit, he was the CEO at the time, my boss, he was as equally convinced about it as I was. And so was Joe Tucci, who's CEO of EMC, which held 80 % of VMware. And so there was no hesitation from that point of view.

8:40Yeah, because we got it done over a weekend, actually. Yeah, yeah, it was some of the fastest corp dev work. Because I think also the VMware team and Ragu realized that if Cisco actually realized what was going on, they would buy it at any price. because it was the only actual threat to Cisco. And speed, I mean, that's a lesson that carries through, especially against bigger companies, Cobb Dev and Acquisition, and so on. Speed matters a lot. Speed matters a lot. Yeah, getting a deal done over the course of the weekend at that size, the highest price paid ever, is very fast. I will say, I mean, Regu has so many strengths as a leader, which many people know, product, et cetera, but it's hard to overstate how complex the integration was because the reality was is the VMware had an internal team that was doing work just as good.

9:34Like literally, they were just as talented, just as good. Some areas they were stronger, some areas we were stronger. Now what we had that they didn't have is we had all the Linux stuff just because we were doing this open source stuff. But we were a smaller team and we were a team that actually got this big acquisition. And so Raghu was in charge of integrating this smaller team with a large payout with a larger team. Both teams had done this great work in having that work out. And listen, we managed to do it. Yeah, we managed to do it. Yeah, so how did you actually do it? I mean, it took a lot of great leadership from the Masira team as well, right?

10:08And the VMware team that was in there to get to work together as one. And then overall, it turned out the value proposition that mattered to the customer had elements of what VMware had built and what Martine had built. So once you realized that, then it became easier. The tricky part was whose base product do you use? And that was the hardest part of it. But once we solved that, then the rest became possible. I will say, when we were selling, I put in a call to Diane Green, who was the CEO before, and I said, like, okay, I would love some guidance. Like, is there anything you can, you know, point me to?

10:47And she said, keep your sales team. That was it. That was kind of the guidance. And I didn't appreciate that until after we landed, which is if, and by the way, VMware is phenomenal. Like Raghu and Pat were phenomenal. They kind of allowed us, you know, to do our thing for the first few years pretty much unencumbered, right? And then over time, you have to integrate kind of more seriously. So the thing is, when you have your own sales team, you can actually lead your product team much better just because they're directly doing the enablement. They directly see the customer pull. They directly have access to customers.

11:19So it's much easier to unify a team from the market on end if you actually control that market. So I honestly think one of the keys, in addition to the leadership, in addition to, you know, the hands-off approach, was just we were able to keep and grow out our independent sales team. And then over time, of course, we rolled that back in. And let's face it, the category was not built out. The product suddenly was not built out. So without an independent sales team, we would not have been able to build a product. Yeah, we would have gotten the vine. Yeah. Yeah. Okay, so then brag a little bit, because this is one of the most successful integrations and, you know, financially acquisitions probably in the history of technology.

11:57So what happened in the coming years afterwards? Yeah, I mean, I think there were two dimensions of value to VMware. One is the independent revenue from this. I mean, this cut to about$2 billion in revenue. And this was, by the way, license. And you sold it for$1.3 billion? $2 billion. This is why it was a discount record. Yeah, get your mind around a less than one times revenue multiple. Yeah, so I was just going to say, the second is the multiplier effect on the VMware core franchise. Yeah, sure. Because it made, I mean, Microsoft's product was slowly getting better, open source was getting better.

12:32Yeah. And now we've been able to redefine the core virtualization stack in the enterprise to say you've got to have this and you've got to have networking, right? And so just purely on a multiple revenue point of view, that it paid for itself 10 times. Massively. I will also say there was a portion of time where we were like just the BU that was, you know, part of the necessary acquisition was a significant portion of VMware growth. Yeah. Right. So not only was it just like the revenue alone, it was actually, if you looked at the overall aggregate revenue added, yeah. Yeah, like, I mean, basically the growth was coming from the...

13:04It was a meaningful contributor to the growth rate. I forgot the numbers, but yeah. It was 46%. There you go. I'm sure you don't know. I remember. I remember saying just something around there. To defend the sales price, though, I remember when we had the meeting with the bankers, the dot for our multiple was like literally off the chart. Like of all the deals that had been sold, nobody had that. Like it just didn't even fit on the chart. How much revenue was it? There was one other dot that we beat, which was Yammer. Oh, yeah, Yammer with Microsoft. Yeah, it was Yammer. But they had like 20 million.

13:51Slightly less successful integration. Yeah, yeah. God bless David Sachs for getting that deal done. Yeah, yeah. Okay, huge success story. Okay, so that's Nysera into VMware. Obviously, you were at VMware and then became the CEO ultimately. So talk about your experience building the company. At the time when we bought Nysera, we were on a product expansion path. So obviously, I mean, Sarah wrote a great piece on the platform. Platform, yeah, what it takes to be a platform, yeah. It was as though she was at VMware, right? Because it literally follows the playbook. And so once you get to a certain level of saturation in your core market, you got to build out the adjacent markets, right?

14:36And that's what we were up to. We built management, then we got Nisira for networking. We did our own thing on storage and then security. So we were progressively building that out. So that was our task for the following seven, eight years. What that did is it vastly expanded the on-prem franchise. Because even though cloud was coming, we were continuing to dominate on-prem. And as you well know in the enterprise, things never go fast. They always take their time. It doesn't really go away. New stuff comes on, but yes. Yeah, exactly. So I think that was one big part of the mission. The second is we initially tried a variant of the cloud, cloud review, add our own cloud thing, which did not work.

15:25And then we recon the industry around what we call the hybrid cloud because most of these enterprises, they had some stuff on AWS and stuff on us. So we said, let's go partner. And that was another radical move. Because people thought this was going to be a blood path between the two. Yeah, of course. So I was doing that, right? And then we had to get the developer franchise, so we went and acquired Pivotal. Yes. So it was basically building out the product portfolio, first as a GM for all these data center businesses, and then as a COO. It's also crazy about it. is like the amount of lessons that you saw in infrastructure software successful.

16:08Like, you know, under Palmeritz was trying to go to the application stack, which is very tough to like skip parts of the stack. Then you went back to doing adjacencies. You did organic product build-out like vSAN, which became billion-dollar product lines. You did inorganic acquisitions, but it became billion-dollar product lines. So I feel like, you know, like the broad scope of business strategy was embodied at VMware at some point. I think more than any other company. Like you can see it in hardware companies like Cisco, but a few other companies had to play that many games. In reality, if you're going to be around for more than 15, 20 years, you've got to play all those things.

16:40Yeah, that's true. Fair enough. Because the market's changing you so fast. But yeah, I think if you have a consistent theme around how to do it, then it's quite possible. Yeah, and actually VMware, interestingly, was born before the cloud. Yeah. So although it was kind of a cloud technology, it was like a very, it was almost like not even first generation. It was like pre-first generation cloud technology. So the cloud immediately put pressure on the business. It's this very difficult product cycle circumstance. That's the business model too. Yeah, even like most people's first experience with VMware that were around at the time was actually as a consumer product.

17:24Yes. Like it was literally like, it was like, no, it was really like the first time I used was like the late 90s and I was like, I'm running Red Hat on my Windows machine. That was it. And so you go from a consumer product to this, and then of course you did server consolidation and it became a deep enterprise product. And then of course there was all the OEM stuff that happened early on. It was a very, very complex business. Yeah, it was a complex business. And by the way, that's where the founder vision comes to play. Diane and Mendel, right from day one, knew that they had to get to the server.

17:53And they actually charted their path. They said, hey, we're going to get known with a desktop product, which by the way is going to work out all the bugs in the system, and then we're going to go into the server and then into the data center. So you've got to give them a lot of credit for that. Yeah, for sure. So at exit, how large was the company? How much revenue? How many people? So we were$13.5 billion in revenue. How big was it when you joined, roughly? I think it was$10 to$15, about$40 a quarter, something like that. $40 a year, something like that. $40 million? 40 million a year. Or something like that.

18:30To 13 billion? Yeah. What a ride. That's incredible. Yeah, again, with the cloud in your face, like that's the most incredible part of it. Yeah, yeah. No, it was. And Microsoft. Well, Microsoft. And Microsoft, yeah, of course. VMware was sitting on top of Windows, right? I mean, like, you know, Netscape. Yeah, what was the name of Microsoft's crap? Hyper-V. Hyper-V. Hyper-V. Hyper-V. Yeah, yeah, yeah. No, I mean, I think after that Netscape, the thing that gave me most satisfaction was beating back HyperV. Yeah, I was going to say. I'm like, you're not going to give me a second time. Yeah. Give me a second time.

19:05I love that. Yeah. So, yeah, so$30.5 billion, about 37 ,000 people. Incredible. And$69 billion in enterprise value. $69 billion. Wow. Yeah. What a run. Okay, so I now want to shift topics. You have this incredible set of experiences. is you know compute, network, storage, maybe better than anyone experience-wise. What are you going to do with us here? Yeah, I mean, I think I would say fundamentally two or three areas where I'm super excited by. One is like we were talking about the other week, everything from the foundation model down all the way to the power station is going to get reinvented.

19:47Yeah, of course, yeah. Right? There is just no other way you can supply all the compute that AI is going to need, right? And that opens up all the opportunity in the world. For the last 15 years, you really did not have opportunity as a hardware vendor because everything was being done in-house by the hyperscalers. Now there's neoclods and the world is moving so fast, even too fast for the hyperscalers. OpenAI is going to build a lot of things on their own, Oracle, et cetera, et cetera. So now there's a legitimate infrastructure market that's opened up for innovators. So I think I'm super excited by that.

20:27I think that's going to be a significant area for both early-stage companies as well as late-stage companies. Yeah, and helping late-stage companies navigate that market as well. Yeah, there's going to be new types of service providers, etc. I think that's one. The second one, and this is more in partnership with you on the growth side, the enterprise transformation will happen, right, through AI. I think that's going to open up a lot of opportunities. And then the third is adjacent areas to the data center, like robotics applied to the data center, generally robotics as a whole. Those are all areas that are going to be physical AI, in other words, significant areas that I think are going to be a lot of exciting things, a lot of fun to go study and work on.

21:14So, you know, one thing that really occurred to me, you know, as Ragu was talking to us about joining is, where there's this very unique point in history where even very small companies have the problems that would draw on the breath of someone like Raku. Like, think about someone like Cursor. You know, like, the ecosystem includes, like, kind of incumbents at scale. You know, the company is at scale. How do you manage platforms? How do you manage the clouds? All of these things are at the sophistication that you would actually see a VMware, right? And it's just because AI is growing so fast. And so one thing that I'm – I mean I've said this many times.

21:53Rig is the best techno strategist in the industry. And I strongly believe that. But also he has this wealth of experience when it comes to partnerships, OEM deals, dealing with incumbents, dealing with competitors, dealing with talent, like the whole thing. So it's like if you can distill all that down, it's actually applicable. International expansion. International expansion. Yeah, exactly. Literally, it applies to the companies in the portfolio that aren't that, yeah. So, like, you would think, like, oh, this is some, you know, this is only going to be growth stage stuff. And it's absolutely not the case.

22:22No, no, no, no, no. Well, it's not. This is so relevant. That's the dynamic. They are early stage companies that reach the problems of big companies so fast because they grow so fast. Totally. And so this is the magic of organic demand and, like, the market pool that they're all seeing. In fact, you know, as, you know, Ragu was coming on board, I mean, probably every night I would call him asking for some help with, like, you know, a 10-person company. I'm like, can you help with this intro? Can you talk to us about whatever? And so this is the reality of what we're living in. Everything's being changed all at once.

22:51The problems are the problems that you would face as a very serious senior CEO. And so it's great to actually have that kind of experience and talent around the table. Yeah, early stage companies are having large company questions being posed to them. Well, it used to be like year 10 or 15 when you're facing things like how do I be multi-channel, multi-product, international. and now it's like year one and a half. No, you see M &A, like, they're working very well, these very large cloud partners. I mean, we have companies that will, like, raise money and then broker a$100 million GPU deal, right?

23:26I mean, this happens all the time. With tiny companies, like very small companies. And then you're relying on these, I mean, all of these things that you have to manage, you know, so. Yeah. Well, and the thing that's still true is none of the founders, a year into running the company or two years in have any of those skills. Right, of course. Yeah, and so it's just such a boost to have somebody to talk to. I mean, it's transformational because the number of mistakes that you will make right there and blow yourself to bits is just insane. Well, and we have to change too, right? Like the way you build a company is different.

24:05The technology stack is different. The type of people you hire are different. And so we also have to be different. And so one of the things that for me is great about having Ragu here is there's a lot to do on that front, you know, to kind of improve on every dimension there. So I need help too. Yeah, exactly. It's both the combination of the companies growing really big, really fast, and facing big company problems really fast. At the same time, the companies stay private so much longer now. So we have to evolve to that dynamic as well. So one of the areas, Regu, that you mentioned that you are intrigued by and you have obviously great experience in is robotics.

24:51I'd love to hear your state of the world of robotics right now and what gets you excited over the next few years. Yeah, I mean, I started looking into the data center construction and all that, right? And server construction. and you find that, number one, all those industries are highly manual today, believe it or not, right? These AI servers that people are building is like 70 % of labor or 80 % of labor, right? Wow. And if you look at the bazillions of GPUs that have to turn into usable infrastructure and all the power that has converted, the only way that's going to happen is there's going to be a lot of robotics, right?