Rob Ward | A Top Venture Capitalist Analyzes the AI Landscape

0:00The CMO Confidential podcast is a proud member of the I Hear Everything podcast network. Looking to launch or scale your podcast? I Hear Everything delivers podcast production, growth, and monetization solutions that transform your words into profit. Ready to give your brand a voice? Then visit IHearEverything.com.

0:22Rob Ward:Welcome to CMO Confidential, the podcast that takes you inside the drama, decisions, and choices that go with being the head of marketing. Hosted by five-time CMO Mike Linton. Welcome marketers, advertisers, and those who love them to Chief Marketing Officer Confidential. CMO Confidential is a program that takes you inside the drama, the decisions, and the politics that go with being the head of marketing at any company in what is one of the most scrutinized jobs in the executive suite. I'm Mike Linton, the former CMO of Best Buy, eBay, Farmers Insurance, and Ancestry.com here today with my guest, Rob Ward.

1:03Today's topic, a top venture capitalist analyzes the AI landscape. I know you're left. Which part was it air quotes, the top part? Yeah, top, top. Well, you know, we have a lot of leeway, journalistic leeway. Now, Rob is a co-founder and general partner at Meritech Capital, a highly successful late-stage venture capital firm in Silicon Valley. He's worked there for over 26 years and invested in companies like Facebook, NowMeta, Snowflake, NetSuite, Zipcar, and Cloudera. Full disclosure, I've known Rob and his firm for many years and consider him to be one of the best at explaining what technology actually does.

1:48Welcome to the show, Rob. Thank you, Mike. Great to be here. Long time listener or first time caller, though. First time caller, right. I hear you, Seattle. So first question. Rob, where are we really in AI adoption? You've seen adoptions over 26 years, massive ones. Tell us where we really are today. Yeah, it's the right question. My honest take is we're still so early. And I say that not because there hasn't been a lot of activity, and we'll get to that. And there's certainly been a lot of hype. Um, but, and so it's not, you know, I'm not saying it's early in the sense nothing's happened. A lot is, is happened in a very short time period.

2:37It's kind of a remarkable to think chat GPT only came out three years ago. Right. Yeah. Um, but, and, and obviously, you know, we've had AI for a long, long time in, in the sort of ML sense of, of machine learning sense of AI, but generative AI being relatively new. And I'll also say there are some specific areas where it's moving very quickly, right? AI for writing code, you know, companies like Cursor, like Anthropics, Claude, you know, Cursor went from zero to a billion in over two years. Legal AI, another area where you're seeing a lot of activity, Harvey, Lagora. Harvey's over$100 million in annual revenue.

3:23search, glean, rapid prototyping of web apps, lovable. So, I mean, there are some companies with staggering growth rates. I mean, and just to put that in perspective for your listeners, we used to think getting from zero to a hundred million in revenue in about eight to 10 years, you know, that was best in class for decades. And it kind of get down to, you know, that was sort of the Salesforce DocuSign era. Then it got down to like three to four with Wiz and CoreWeave. Now it's like a year. These companies I just mentioned, many of them, it took one year to go from zero to a hundred. So there's a lot of explosive growth, but it's specific to certain areas.

4:05And those are businesses that are either targeting developers, early adopters of new tech, or businesses where the fit between generative AI is just hand in glove, like legal, perfect example, right? It's texted out, sort of tailor-made. Beyond that, it's still a lot of experimentation, especially when you talk about the enterprise world, right? Enormous pressure to do something, right? I mean, enormous. Give me a win. Let me get out there in front of Wall Street and say that we're an AI company. But it's a slog, right? Getting widespread enterprise adoption is not easy. Getting these deployments into production really hard.

4:49But it's coming, right? 90 % of IT budgets are going to be up next year. And of that, 90 % of them think they're going to increase their AI spending. And so, I mean, don't get me wrong. And the vendor spend, that's where the real action is, right? I mean, as in, I can't... When you say vendor spend, tell our listeners what you mean by that. I'm talking about, you know, the hyperscalers, the companies that are actually building out these data centers to run AI and then the AI platform providers themselves. The best statistic I've seen on this topic, there is a renowned venture capitalist, a guy who's been around for forever named Roger McNamee.

5:36And he made the point that by the end of 2025, the tech industry will have invested about three quarters of a trillion over the last three years into LLM AI and that infrastructure supporting it. And if you believe that we're, you know, that's a trend line that will invest a, you know, relatively comparable amount in the next year, that the AI industry will have received more capital than has been invested in the rest of the tech industry from the dawn of Silicon Valley, right? Sort of the mid-50s when, you know, the transistor was invented. And I mean, that's just a staggering. But the valuations are going with it.

6:18and the hype is going with it. So how do you as an investor think about these valuations? Because it's like, you know, and also the circular investing thing that's going on. How do you look at that? I mean, as both an investor and an industry observer. Yeah, yeah, it's challenging to say the least. First of all, people are like, is there a bubble? The bubble question. We might as well address it because I know it'll come up at some point. There's no doubt this is a bubble, right? I mean, something like, if you look at the public market, something like 80 % of all gains this year are coming from here.

7:05If you just pull out, the public markets are not really performing very well, but it's totally clouded over by AI. And by the way, there's also this huge bifurcation. If they say you're an AI winner, you trade it, you know, many multiples of revenue. Oh, yeah, we're changing our name to CMO Confidential.ai. Perfect. AI Confidential, even better. And in the private markets, as you point out, you're right, the valuations are extraordinary. And, you know, I've been around long enough. This doesn't end well. And it's not just that valuations are going up, by the way, it's that there's a lack of oversight creeping in.

7:46It's that rounds are happening at an unbelievably accelerated rate. When you say rounds, he means investment rounds where people are just taking money without, I'm going to say, as much due diligence as they might historically have applied. But again, to try to make it, to put a point here, it's not unusual now for if we're investing in a company, let's call it the Series B round. Before that Series B round has closed, another investor has come to the company and said, hey, we'd like to put money in it 2x the price of the round that is just, it isn't even closed. And so, you know, you're having the Series C happen right after the Series B.

8:31It's absolutely, you know, just not a great fact pattern.

8:41I'm not done with the terrifying part. Okay, go ahead. But the other thing we have to worry about is there's this phenomenon of what is known as AI washing, right? The practice of every company, just what you were doing with your own podcast name. Yeah. Exaggerating your company's AI use or capabilities. And every deal is an AI. Oh, no, Rob, we use it for the newsletter and some other things. So we're really adopters. And I will say, you know, the technology, the underlying theme that's going to come out here is this technology is moving so rapidly. And therefore, what is really successful today, there's a lot of question about that long term viability of that business.

9:30And remember, I can't I'm not a public investor. I can't just decide tomorrow. Hey, I don't really like this business anymore. There's a better one over there. I'm going to sell it. I'm in this thing, you know. Yeah, because when a VC invests, they're in. They can't get the money out until another round or someone buys them out. So these are illiquid securities. And yeah, a competitor can leapfrog your technology. We're seeing it already in AI, you know, again and again. And the switching costs are so low. So the cost of being wrong here is just, you know, extraordinarily high. I mean, just to give you a sense for this durability of revenue concern, we looked at a legal AI company a few, 18 months ago, two years ago, and we did a diligence call with a really well-known legal firm in Silicon Valley, one of the customers that you absolutely like to have.

10:28And the feedback was extremely positive on this AI software. And we got to the end of the call. And of course, we said, well, so you're going to buy this, right? Because he was just running a proof of concept where they get to use the software for several months and try it out. And his response was, oh, God, no. Why would I do that? I feel a little thin veneer that sits on top of this AI software between the users and the software. And once the POC is done, I'll go grab the next great thing because there's going to be something that will be way better. And I'll just unplug this and plug that in.

11:05And my lawyers won't even know the difference. And that POC is proof of concept. Yeah. Yeah. So that's. And I guess there's one more thing I'd point out, unlike in 2000 with the Internet bubble. The startups had a big advantage because the incumbents were basically catatonic, right? They were asleep at the wheel. It wasn't really a big fight back. That is not the case this time. You know, who's leading the charge here? It's Microsoft. It's Google. It's Amazon. And these are formidable competitors. competitors, right? Highly ambitious. They've got great leadership, you know, very, very strong. Limitless resources almost.

11:46Yeah. Yeah. But, okay. So that's all the negative. Let me give you the reason though that people then are still, you know, leaning in or jumping in. VCs, we live for platform shifts, right? That is, it's not, it's not wholly untrue to say that you, as a VC, you make all your money in these periods. Like you, and in between, you sort of just try to keep your head above water until that next big - Yeah, and a platform shift would be, yeah, the digital revolution, mobile, 2.0. Yeah, cloud computing, mobile, farther back, obviously the internet, farther back - Social, yeah. There's just been waves of them, right?

12:27And it's not also, you know, really unusual to say that this is the mother of all, waves. One of our guests, the authors that wrote AI First said, this is the biggest prize in the history of capitalism. You got it. You got it. So you are not, as a venture capitalist, you do not want to miss out on this is your moment to shine. Maybe nothing like this ever again in your career. And so, yeah, that's why despite everything else I just said, there is an unbelievable amount of activity. And to be fair, there are all kinds of really interesting startups doing great things and succeeding. There's also some advantages to this AI wave that startups never had before.

13:16I mean, part of the challenge or the trade-off, I guess let's put it this way, is startups always were able to deliver a much simpler product experience, but the incumbents would win still because they could configure things and customize things in a way that was really useful for that individual purchaser. Consulting firms and folks like that really make the playing field more difficult for startups. With LLMs, now they can enable configurability and customization in a really rapid manner. And so it's given startups equal footing, so to speak, in terms of fighting the battle on the large incumbents battleground.

14:05So it's a super terrifying time, but it's a super exciting time, too. So, you know, we talked a little pre-show about how you gauge how these companies turn profitable, especially during the investment phase. And then also, if you're going to adopt these as a company, the switching costs may be small now, but if you really get everybody trained on one or two of them, how's the switching costs later? How do you guys think about that when you look at profitability, sustainability, like companies that are going to win and companies that are more focused and are going to get eventually passed, bought or crushed?

14:47Yeah. To be fair, and we did joke about this to begin with, we don't really look at profitability. I mean, the farthest out we will look at is when will this company become cash flow positive? You know, becoming profitable even as often, you know, many, many, you know, for at least several years beyond that point in time. And I will tell you, there's always a bit of a trade-off between growth and profitability. It's easier to get profitable if you grow at a slower rate. Because almost by definition, when you're growing more quickly, it means you're having to hire more developers and hire more salespeople, and you're layering in more operating costs.

15:33Now, that has definitely flipped a little bit here with AI. And a big part of it is because companies are able to get to market so much faster with such lower levels of headcount.

15:53There's a famous story about the cursor coming to market, and it was like a three-person team that actually brought the product out. originally. That's a little bit of a hyperbolic example, but it's directionally right. And we see this even in our own portfolio. It's not unusual to see companies that have gotten to tens of millions of revenue and their headcount is 10 people. You will see some of these businesses become a little bit more profitable, certainly at an earlier stage. Not to say they still aren't raising huge rounds because of these huge, these, these big valuations. And I might also say venture capitalists are partly to blame.

16:41There's a, there's a very terrible term going around called the foie gras effect. And it's basically larding startups with capital until they, until they burst. And because again, the average venture fund has grown so big and has so much capital, they're just desperate to find places to put it. So even these profitable or relatively efficient companies are still raising massive - That's a foie gras bubble. I mean, the foie gras thing is now stuck in my head. So we're also seeing all the circular investing. You have the foie gras thing. We just had the you know, the core weave announcement where they lost like, you know,$40 billion in valuation almost in no time.

17:32How do you think about all this circular investing? And then, you know, Amari's law, you know, that you're underestimating now, and it's going to have an effect later. How should our marketers be looking at this when they are evaluating what to do? Yeah, it's a problem. It's real. I mean, don't get me wrong. Real dollars are being spent and chips are being shipped, but a lot of it is circular. Now, it's a little better than what we saw in the dot-com boom, right, where it was started. But that was just trading money. I mean, that was circular. There were no real assets. Here, there's some physical assets, right?

18:12So it's much closer to sort of the Nortel Lucent vendor financing deals back in the day. Again, those didn't end well either. Balance sheets are a little better to begin with, but the sheer scale means that even with healthy balance sheets, it's eye-opening. And, you know, the scary part is so much of this data center spend, it's not even, I'm not even sure, you know, it's going to pay off even if it gets done and gets done properly. And here's what I mean by that. As an example, there's this company called NeoClouds that are like Together and Lambda, and they basically offer, they're cloud providers, they're startups that offer, you know, a GPU-centric architecture, right?

18:58tailored for AI workloads. And you'd say to yourself, why in the hell, why are these companies, you know, able to succeed? The hyperscale, you know, the Amazons, the Microsoft Azure, the GCP, Google Cloud Platform, they've got tons of data centers. The reason is those data centers are not optimized for AI. It turns out the value of a data center is it's highly dependent on the specific use case, right? And back to what I said earlier, but, you know, AI is changing really rapidly. So will all this construction that's happening today even be the right kind of data center architecture in five years?

19:35You know, that's far from clear. That's a big question. Let me layer on another sort of scary thought. Again, this technology is progressing incredibly rapidly. So how do we really know what that demand is going to look like? Because this speed and scale of improvement means it's impossible to forecast demand, right? I just have to ask this. So essentially, if you were building ships, and I'm doing an analogy, and data centers are ships, and you are building battleships. Exactly. It's a perfect example. And five years from now, you realize, oh, this is going to be an air war. I should have built carriers.

20:15These battleships are not usable. You have to scrap them and build aircraft carriers? Yeah. Great analogy, right? You can be parading your great white fleet around the globe like we did in the early 1900s and then suddenly realize, uh-oh, this really isn't where the puck has moved. And the proof of how tenuous this is starting to become is just look at the way the finance world is starting to react to these deals. credit default swaps boy we all thought we'd never hear that term again after the you know the the mortgage in a financial crisis of a decade ago the purchase of those by the way has exploded in the last tell everybody how credit default swap works in just a couple sentences you're you're betting against something right back in the financial crisis you bet against the housing market the big bet that paid off spectacularly was that you know we were we were you know sell you know, packaging together and selling these mortgages, you know, that were all subprime, but is packaging them together and washing them and saying they were prime.

21:25And so it's essentially a bet insurance policy against an investment. So you pay a premium to the seller and it pays out if that borrower, in this case, the borrower defaults on a loan. So you're betting - So back to my analogy, I'm saying you're going to be building all these battleships and I believe you're going to need carriers. So I am buying a credit default swap against your battleship. Yes, exactly right. And the biggest targets now are Oracle, they're CoreWeave, they're all the businesses that are the less credit worthy in this great game that's going on right now. So if I'm sitting there listening to this show, and we have a lot of C-level people listening to this show and also a lot of agencies and consultants, how should they be picking vendors or picking people to play with in this market?

22:24Yeah. I think that the best thing today is to choose someone that's a, you know, fits under the umbrella of a trusted advisor, right? It's not, this is not a technology bet alone because again, as I mentioned, things are moving so rapidly. What you really need is somebody that you trust can help you navigate, you know, those, those twists and turns. And, and the other part of it is you need somebody who can help educate your team, right? This is, again, incredibly tumultuous period. You need, you need somebody who has exceptional client experience, right? You need somebody who can help you with change management.

23:05This is, you know, we can get in the topic of why some things are succeeding and why some things aren't. I think this is a big part of it, right? You have to have that approach. You can't skip that. You can't, don't skip the people part, right? This is, people are - Yeah, we did a whole show that the hardest part about AI is the culture. Yeah, change management. I mean, it sounds boring, but it's absolutely true. And it's doubly true here because expectations are so out of control, right? People think this is black magic. It's going to solve everything. And so they start applying it everywhere all over the organization.

23:48Not the way to adopt AI, right? Small, focused use cases. But some of our other guests have said, you got to have a use case. You got to actually watch it really hard. And then you got to put this into the org in a way the org accepts it. but we've and we've also had a couple guests say don't be buying small little answers like go with a frontier model and and then if you want to buy some small answers great but don't be messing around with a million little startups if you're a big company what do you say to that yeah I I think that there's a there's definitely a time and a place for that but again it gets back to who's going to be there, who do I trust can deliver a successful outcome for me.

24:37Sometimes startups are the ones that will give you the most attention and give you the most bespoke solution that allows you to win. The other thing I'd say that people skip over is the people that see the real results aren't just the ones that have the most advanced models, but it's the ones that can manage manage the complexity of the data layers. Said another way, there's no AI strategy without a really sound data strategy. Give me a good example of someone managing the data layers really well. Well, here's the way I would say. First thing is, you need to have incredible context. You need to know, I mean, AI is at some level a black box.

25:23And so therefore, how are you going to trust the output of these models unless you trust the inputs. You know where the data is and you know who touched it and whether it's the freshest data. The puzzle analogy is a good one, right? The data are the pieces of the puzzle, but the context that you need is the image on the box, right? It allows you to sort of put the pieces together and transform it into action. And, And, you know, shameless plug, we've got a portfolio company named Atlin, and this is what they do, right, for large organizations. They provide that data context layer. And it's very different, by the way, for AI than BI, right?

26:04You know, people say, oh, you know, I've got a data strategy for business intelligence. Sorry, that's BI. That was sort of the world of data, you know, before we had AI. Right. But it's totally flipped, right? In BI, structured data, rows and columns was really easy. The unstructured stuff was hard. Now it's the reverse. Unstructured is actually easier for AI. That structured data, you need a context layer. Because here's an example. I could say to you, what are our high-value orders that are at risk? And how's the AI going to know what I actually mean by that? What's the idea? Or yeah, if I said I want to go on vacation, it could give me 100 vacations, but it would have to know what I kind of wanted.

26:51That's right. And by the way, you know, the AI is never going to say, I don't know. Right. It'll hallucinate. No matter what you ask it, it always says, that's a great question, Rob. I like how it treats me that way. Yeah. My colleague and your close friend, Paul's Madera's favorite phrase, right? Frequently wrong, but never in doubt. I mean, never in doubt. Yes. Hey, so if I'm sitting out there and I'm watching VC patterns because, you know, I want to pay attention as a marketer or a business person. What could I be watching in this industry to just increase my knowledge base? Yeah, yeah, yeah.

27:38I think, you know, it's, here's the way I'd say it. I would say, you want to dive in, right? You know, now's the time if you haven't already started, you know, yesterday was not too soon. and the way where we have seen success, at least within our portfolio companies and even with our firm, and I'll talk a little bit about how we use AI, but as a marketer, start again with those easy wins, bring AI to initiatives that you're already employing, that you're already doing manually, but you can't properly scale, right? Without hiring a ton of humans. So you're sort of rate limited without using AI.

28:26So what do I mean by that? What's in that bucket? You know, targeting, right? More data leads to much more accurate sort of lead scoring. Content personalization. And we've got a great company, shameless plug number two, called Clay that does this. You're doing a great job of getting these in subtly.