00:00:02
Speaker 1: Bloomberg Audio Studios. Podcasts, radio, news.

00:00:07
Speaker 2: This week on the podcast, my extra special guest is Glenn Kacher. He is the founder and chief investment officer at Light Street Capital. He's got really a fascinating background and a great track record. He worked at Julian Robertson's Tiger Management, eventually ended up at Roger McNamee's Integral Capital Partners. He's put together really a fascinating focus and track record. One of the few hedge funds located right in the middle of Silicon Valley, focused on AI and technology. I found this conversation to be absolutely fascinating. And I think you will also, with no further ado, my conversation with Light Street Capital's Glenn Kacher. Glenn Kacher, welcome to Bloomberg.

00:01:07
Speaker 1: Thank you.

00:01:07
Speaker 2: Before we get into Light Street, let's talk a little bit about your background. You graduate from University of Virginia School of Commerce with a bachelor's in commerce and eventually get an MBA from Stanford. Was investing always the career plan?

00:01:25
Speaker 1: It was. I started really looking into that industry. I read a book by Peter Lynch called while I was in college, one up on Wall Street. Or beating the street. It could have been the first book, actually. And I was just caught by this idea of the search for great companies, great ideas. And the way he told the story of finding these companies and researching them, it was really a journey of a detective trying to figure out what would matter in the future. And that really captivated me and my interest in becoming an investor.

00:02:04
Speaker 2: So in between UVA and getting your MBA at Stanford, you work at Julian Roberts Tiger Management. How do you get to Tiger at 22?

00:02:16
Speaker 1: A very fortunate opportunity. So one of the teachers or instructors at McIntyre School of Commerce at UVA, was a former Tiger management partner in Michael Bills. And Michael taught several finance classes there for a couple of years. He had taken some years off from Wall Street after working at Tiger and then before starting a fund-to-funds business that he's run very successfully. And he suggested that I take a look at it. I certainly knew of Tiger. Tiger was It seemed about half of the investment staff actually at one point or another attended UVA. And so a lot of the guys there, you know, sort of knew what we were capable of as young guys coming out with finance degrees from UVA.

00:03:12
Speaker 2: And Robertson was legendary in 93. Was he still running the ship? Oh, yeah. Very much in charge.

00:03:18
Speaker 1: Very much in charge, yes. You know, I was there from 93 to 96 full-time, and then still when I went to Stanford for a year, I worked for Tiger as well, and Julian would occasionally wake me up with a 6 a.m. phone call when I was in business school.

00:03:35
Speaker 2: 6 a.m.

00:03:35
Speaker 1: East Coast. No, 6 a.m. my time, 9 a.m. his time, just a half hour before the market. So he had some discretion there, but we had some great times. learning and talking through the technology industry at the time, investing in companies like Dell, Microsoft, Compaq, and Cisco were some of the.

00:03:59
Speaker 2: So really, right out of college, you're full-on technology? Did you look at other spaces?

00:04:04
Speaker 1: I worked briefly looking at financial institutions with Rob Pitts there, and we had a great time doing that, but I was certainly more interested in technology and I had really studied that industry prior to going to New York. And so it was a better fit for me following that industry. And, you know, I think two or three months into the job, I ended up sitting two chairs away from Bill Gates at an analyst meeting at the sort of after the meeting dinner. And at that point, I knew I was in the right spot.

00:04:41
Speaker 2: To say the very least. So after Stanford, you end up at Roger McNamee's Integral Capital, and you stay for 13 years, and you're really less of a public markets analyst and more of a venture sort of banker. You either lead or co-lead venture investments, and the list is pretty impressive. Agile, ArcSight, Blue Nile, Epiphany, Extensity, Fortify, Interwoven. log me in, OpenTableOvertureGoTo.com. What's the common thread? Is it just, hey, that's what was hot in the late 90s? Or what tied that list together?

00:05:21
Speaker 1: Well, the amazing thing about Roger was he really focused on saying, look, we can't cover every company in this industry. We were a small team, much like at Tiger, there were two or three of us looking at tech at any one time. And At Integral, even though we were a tech-focused firm, but we had four or five people total. But even with that number, you can't cover the entire industry. So you have to focus in when you're investing and say, where is the change really happening most quickly? Where is it most dramatic? That disruption equals opportunity as an investor.

00:05:57
Speaker 2: That's a theme that comes up over and over in your career. Identify the disruption and get in front of it. before the existing companies realize what's coming down the pike.

00:06:09
Speaker 1: It's great to be early, but not too early, right? I mean, that's also an important part of it. Right.

00:06:14
Speaker 2: I started on a desk, and early was equal to wrong, at least when you're trading public equities. Not only do you do all of these privates where you've co-led, is this right, about 46 deals? 46 deals at Integral over 13 years. I read something you had said about that, And you said, you know, the takeaway from all these private venture investments is you don't buy the second or third best company in the space. You always buy the best company you can. Give us a little details on that. What's the thinking behind it?

00:06:48
Speaker 1: Well, experience, right? I mean, you see the movie over and over again, whether it's private investment or in the public markets, that, you know, the old saying was, you know, the number one player is going to get two-thirds of the market. number two player might get 20%, 25% tops, and everyone else fights for the scraps, right? And, you know, there's the ability to make higher margins and have the dominant market share is just so dramatic. And I think in technology, we've seen the power of that, you know, the ability to sort of compound that lead is definitely there. Now, You also see in technology that you can get disrupted, right? Incumbents are the real innovation, and these disruptive changes tend not to come from the big companies but the smaller companies. There are exceptions to that, and we can talk through that. AI is kind of an interesting test case and the semiconductors behind AI, but there's real power into compounding that lead.

00:07:56
Speaker 2: So let's talk about those moats and the winner-take-all situation, is that primarily a technology phenomena? Is it a modern era phenomena? Or is this companies that develop a unique moat regardless of the space they're in get to capture most of the market share?

00:08:15
Speaker 1: Well, I think you've seen in mature industries, whether you look back at, you know, GE and Coca-Cola, you know, you've seen certainly there's advantages to having that dominant distribution and market share. But in technology, I think it's more a story of getting in front of your competitors and investing more. You have more dollars to invest back in the technology and to grow that lead and that compounding of advantages or compounding of innovation at the early part of a market's development can be incredibly powerful. And then that gives you the opportunity to put in place other kinds of moats that do kind of block your competitors from coming along. You know, there's a lot of discussion today around NVIDIA. A lot of people sort of assume NVIDIA is going to lose their massive market share in AI accelerators, which is roughly 85%. And certainly, I think the move to inference is an opportunity for competitors to change what's going on there. But I think people right now are, for instance, underestimating NVIDIA's opportunity to innovate as well.

00:09:30
Speaker 2: So let's define some terms for some of the lay people that might be listening. Compute and inference. Explain what those are. Explain how they're investable themes.

00:09:42
Speaker 1: Sure. So the training compute or the chips, the AI accelerator chips and Today, NVIDIA dominates that still with their graphics processor chips. And that, you know, those chips originally were made for gaming, for doing very rapid mathematics that have to do with calculating physics and lighting, shading in video games. It turns out that the same kind of mathematics are incredibly well positioned to do the math around AI. And then, so you're training a model, an AI model that will be able to make judgments. And then when you're actually using that model to ask questions and, you know, or have it solve problems and actually execute those problems, that's called inferencing, right? And so inferencing can be done on a more simple chip. So people have kind of used a phrase XPU to X out the graphics and let's say this is the next generation of chips that can be used to actually solve the problems with those models that are built.

00:11:05
Speaker 2: Meaning the compute and the inference are going to be on the same chip.

00:11:08
Speaker 1: They can be done with the same chip, but you can have a more specialized lower-cost chip, usually an inference with more memory, for instance. And there's different approaches and software to execute that with a lower-cost chip.

00:11:23
Speaker 2: So it sounds like our alphabetical evolution has been CPUs, then FPUs, GPUs, and now XPUs. What's beyond that?

00:11:34
Speaker 1: Well, I think that's why we use the term X. You know, there's TPUs, you know, Google's version of their AI chip. We've got Tranium, et cetera, and others, competitors. So there's lots of flavors. You know, you also saw, for instance, NVIDIA by Grok, which was another approach to inference. So there will be many flavors and many opportunities and ways to innovate in inference. Because ultimately that will be a larger market than the training market.

00:12:09
Speaker 2: Really interesting. So I usually save my mentor question towards the end of our conversation, but your list of people you've worked with and worked for is just so incredible. I want to get it out early. In addition to Julian Robertson and Roger McNamee, there was Philip LaFont, Steve Mandel, Chip Morris, who was, I think, at Blue Ridge. Alger, Viking, Lone Pine, Impala, Matrix, KOTU. That's like a murderer's row of modern investing names. What did all these legends have in common and how were they each different?

00:12:49
Speaker 1: Well, I think the focus for, you know, we had a great team there at Tiger Management and so many of us went on to start our own firms and many of them in sort of modeled by what we experienced at Tiger and seeing how Julian did it. I think Julian was just such a inspirational leader and was so values-driven and really focused on, hey, we want to work with the best people. That doesn't just mean the people around the table with you and your investment staff. That also means the CEOs that we backed and the CFOs of those companies. We looked for people that we thought were of high integrity and if there was any question about the integrity of those CEOs and CFOs, we were out. We just weren't interested in that company. And then Julian was, there were no shortcuts, right? It was, you got to do the work, explain to me why and how we got to the conclusion that this company is one, positioned incredibly well, and two, it has a real opportunity. There's something fundamentally changing in their industry or in their product set that's going to change their trajectory. And then the last one was, Hey, let's, let's use our power and, and, uh, success to help other people.

00:14:10
Speaker 2: Right.

00:14:11
Speaker 1: And so the combination of that, uh, of those, um, principles was very powerful. We, we all, I think many of us wanted to, uh, see if we could do, do something similar. And, and that was really powerful. And then I was lucky to go on to work with Roger, uh, and John Powell at, at integral capital and chip Morris, all three of those guys came out of T-Row price. And we worked with Kleiner Perkins. We were in their building. So we were surrounded by some other incredible investors that just saw things early and really invested in great entrepreneurs, people like Jeff Bezos and the founders of Google. I was lucky that I was able to see so many inspirational people and things happen early in my career and just wanted to try to do it on my own.

00:14:59
Speaker 2: Really, really fascinating. Coming up, we continue our conversation with Glenn Kacher, founder and CIO of Lightstreet Capital, discussing the firm's founding and launch. I'm Barry Ritholtz. You're listening to Masters in Business on Bloomberg Radio. I'm Barry Ritholtz. You're listening to Masters in Business on Bloomberg Radio. My extra special guest this week is Glenn Kacher. He is founder and chief investment officer of Lightstreet Capital. The firm is a technology-focused hedge fund and private investment firm located in Palo Alto, which is a good place to start. You launch in 2010. The great financial crisis is still dominating the news flow. What was the original pitch to investors? Sure.

00:15:52
Speaker 1: The original pitch was, look, we've The game board had kind of been reset in terms of making money. Multiples were low. And we saw the emergence of kind of four things. Mobile, with the smartphone really growing. At that point, it was becoming a dominant platform. Social media, most of it was private, but we saw Facebook emerging and Twitter really redefining media. Cloud, the development of taking the internet technology and using it for business and the ability to propagate applications everywhere that the internet was available was incredibly powerful. And e-commerce, the ability to sell goods anywhere at a very low cost and with the back end that Amazon and others had built. to get products delivered within a day or two to many locations in the globe. Those four things were incredibly powerful. And then ultimately we saw things like the sharing economy come out of that. You couldn't have had Uber and Lyft and DoorDash without having e-commerce and the mobile phone and the ability to get those companies distributed through the mobile universe. So there was a real emergence of these four powerful things, mobile, social, cloud, and e-commerce. And it was really redefining what we could do as consumers and business people.

00:17:32
Speaker 2: I love how you described the firm. We are the Silicon Valley home team, one of the few hedge funds living and working at the center of the technology universe in Palo Alto, 100% focused on tech opportunities. The first time I read that, I was like, that can't be right. There has to be tons of hedge funds out there, like not many hedge funds in, in the center of the VC universe. Cause all of those successful venture investments eventually go public.

00:18:03
Speaker 1: Yeah. There's a relatively small number of public market, uh, managers out there and, and a good number of, of, uh, you know, you've seen Philippe, what he's done at, at Coto has been amazing. And, and, um, at Tiger Global, Chase has done incredibly well. Uh, and, and, uh, Whale Rock out of Boston with Alex. And so you've just seen the success of those guys. So I'm not saying it can't be done, right, by any means. But there is a real advantage to living and working in the place where the innovation is centered. And I think when you see this fundamental innovation like we're seeing now with AI, it really draws that advantage of geography back to Silicon Valley. I think there's a small number of great AI entrepreneurs, and they want to be in the same community with one another. And so that's a real advantage for us.

00:19:01
Speaker 2: Yeah, I kept hearing that San Francisco was over. It's dead. The city is, you know, on its last gasp. We were there in the spring, and the city is just, it's a boom town. Like, I know there's a little bit of a boom and bust, West Coast, gold rush mentality. And each new cycle of technology kinda works its way through but it's to anybody who steps foot we were down by the embarcadero the city is just absolutely on fire what's it like is this feel like the late nineties in terms of the amount of uh... human capital intellectual capital and actual money sloshing through That's.

00:19:42
Speaker 1: A great question. I'd say more in the mid-90s, probably. I think that we're at a point where this is very fundamental, low-level technology. We've seen something of a renaissance in the hardware industry, and that hardware innovation really matters when you're trying to scale at the rate.

00:20:08
Speaker 2: Meaning semiconductors or everything around it or the whole ecosystem?

00:20:12
Speaker 1: Semiconductors, networking, down to printed circuit boards. You have to innovate at sort of every level of the stack in order to grow at a 10x, 100x rate. And the acceleration required in order to provide AI cycles at a competitive price is huge. incredibly challenging. And, you know, the amount of demand that's out there is incredible. So the need to scale is back. And I think, you know, it's pretty interesting, you know, what we've seen, you know, I think in the early 2000s and the semiconductor industry was allowed to consolidate and the capital was provided to do that. And you saw a company like Avago and and Hoctan, you know, really organize the industry and do some horse trading of properties to other semiconductor firms and really rationalize that industry. And so, as AI has emerged, what it's done is it's really taken advantage of the fact that there are a small number of companies that compete for a massive market. So, AMD and Broadcom and NVIDIA and TSMC, of course, in Taiwan on the back end, and then, of course, the semiconductor capital equipment companies like ASML, those companies just have very large market share and have huge demand and huge moats and advantages.

00:21:55
Speaker 2: So let's talk about the first four companies you mentioned, Taiwan Semi, NVIDIA, Broadcom, and AMD. That's about 40% of the public portion of your portfolio, or at least it was a few filings ago. I know you're not a big fan of revealing too much of your portfolios, but that's a fairly concentrated portfolio. Tell us the thinking behind having such a dominant emphasis on those four semiconductor companies.

00:22:25
Speaker 1: Sure. Well, it goes back to what I was saying earlier. You want to focus your capital in the place where you see the most innovation. And And right now, that's at the core of accelerating computing in order to do AI. And right now, NVIDIA's got 80-plus percent market share in the network GPU market. AMD is certainly coming up in that. And as we move to agentic AI, which is a very important innovation that's happening in AI and is really driving that next leg of growth, there's certain advantages that AMD has, because they also are one of the two major players in the CPU market for desktops and servers. So that explains why AMD matters a lot. And Broadcom and what they've done with Google with their TPU over the years is incredibly impressive and has gotten them now opportunities with OpenAI and some of the other major AI players. So that's certainly great exposure and then TSMC makes the chips for all three of those companies and the ability to Kind of win no matter who wins is and and really have a massive Oligopoly monopoly almost for TSMC We certainly want to back that company as well.

00:23:57
Speaker 2: So those four companies plus Microsoft you described in 2024 as the AI5, and while everybody was focused on the Mag7, the AI5 significantly outperformed the Mag7 that year. Is it still a concentrated holding, all five, and how does that thesis hold up today?

00:24:19
Speaker 1: It's a great question, yeah. I'd say the company that's kind of been in and out of our portfolio, mostly out, has been Microsoft. And their early lead with OpenAI, they, in our opinion, kind of fumbled that.

00:24:36
Speaker 2: And hence, giving an opening to Anthropic?

00:24:40
Speaker 1: Yes, for sure. And so the uptake of Microsoft's AI that was somewhat powered by OpenAI really didn't work that well. And that was a real miss for them. And ultimately, they pulled back on on their development and funding of their AI efforts. And I think they're now back in the game. But at the same time, what we're seeing now is, you know, for instance, Microsoft is the largest security company in the world. And one of the things that we've learned is that AI creates a lot of security vulnerabilities for businesses. So any business, you know, is going to need to invest more aggressively in their cybersecurity defenses. And so that will be a big benefit for Microsoft. So that's a huge advantage for them. But I think what they've done and the repositioning that they've done on the Azure side of their business has been very impressive. They've also rationalized some of the spending that wasn't going as well in their gaming business and are sort of pulling back there. So I think they're repositioning the company well after they sort of blinked on AI and, um, you know, we're, we're, it's back in our portfolio at this.

00:26:04
Speaker 2: So when we talk about agentic and we talk about the major AI players, is this going to be a duopoly? Is this going to be anthropic and open AI or is it going to be a little more wide open than that? Yeah.

00:26:18
Speaker 1: It's, I think, uh, in this battle's happening in real time between those two, uh, leading companies and, um, Google certainly still a player with Gemini and their advantage in distribution with their massive success, of course, in the search engine business. And they've also now they're backing Apple's AI efforts as well. So they have a real distribution advantage. So I wouldn't count Google out and they still have great technology.

00:26:51
Speaker 2: By the way, their notebook LLM is outstanding. If you want to upload a giant a file, a book, or anything, it's unbelievably accurate and fast. I've been really impressed with that.

00:27:06
Speaker 1: Yeah. Their ability to innovate is stunning. But the real battle that's emerging today is open source models that, one, are cheaper than the closed, anthropic, and open AI models. because they're free. You can download them for free and run them on local hardware, or you can engage with them on others' commodity hardware in the sky. And those open-source solutions are really battling these more expensive frontier models from the two big companies. So we will see. I think the early signals are that... There's a place for both of these solutions broadly defined. There's also some regulatory questions, open models. You really can't regulate very well because you can install them on your own software. You can adjust them to work how you want. There's questions about how to make sure these are engaged safely and in the wild. But there's also not a lot of choices around for regulators, too, because those are in the wild.

00:28:21
Speaker 2: So I want to combine what you've said about Microsoft and security and open source. Is it fair to say that security-aware enterprises are going to be steering clear of open source because of the various security problems and the duopoly of Anthropic and OpenAI is going to be where the big players are going to end up, if for no other reason, if there's a hack, it's a defendable decision?

00:28:53
Speaker 1: Well, there's two questions. There's using AI within your four walls and being able to provide the proper controls to make sure that it doesn't get to your data that is sensitive and that it doesn't somehow leak that or distribute that. The second is what a bad actor can do with an open source technology from outside of your firm trying to break into your firm. So those are the two things that you have to, you know, account for with your cybersecurity spend. And so there's lots of opportunity for whether it's CrowdStrike or Palo Alto and Microsoft, as we talked about. But you've got to protect those. And, you know, in addition, when it's your internal to your organization, understanding what the roles are of the user of that technology or the open source technology and what they can access as a user, you have to make sure you... that you honor those restrictions as utilizing an agent system.

00:30:06
Speaker 2: So we're talking a lot about public companies. Let's just look at some of the private venture investments Lightstreet has made over the years. And this is quite a list. Uber, Lyft, Slack, Pinterest, Toast, Harry's, Everlane, Box, Blackbuck, EasyCater. In 2018, at the IRISONE conference, you presented Palo Alto Networks, far, far cheaper price than where it is today. At a later IRISONE conference, you presented Farfetch'd. All of these have become giant winners. The key question I have to ask is, what does investing in VC teach you about public companies and vice versa? What do you learn about public companies that are useful when evaluating a venture opportunity?

00:30:59
Speaker 1: Sure. In the venture companies that we invest in, and even the ones we don't invest in, there's real value into understanding what's happening in the industry. The advantage for us as an investor is that when we meet a CEO or founder of a company and try to understand how they're solving a problem, they're starting with a blank sheet of paper. They don't have ties to some incumbent solution and incumbent set of customers that they've been trying to keep happy for five, 10 years usually. So they're able to be most aggressive in adopting new technology. And so what we learn is that we can apply in our private investing and our public investing is what matters to them, what what technologies can solve the problem with no constraints, you know, around keeping their long-term customers happy. So that's a real advantage. And I think, you know, when we, as in 2022, 2023, as AI was really emerging as a category, when we were talking to some of these early stage firms about, okay, how are you developing your AI solutions and which semiconductors and infrastructure and service providers are you using. That gave us a real insight into NVIDIA and AMD and Broadcom and Marvell as potential investments for our public side.

00:32:41
Speaker 2: Long before people were talking about it in the mainstream, you're hearing this directly from these clean sheet venture startups. Yeah.

00:32:50
Speaker 1: I mean, there's one great story when I was at Integral. Bill Joy was a partner at Kleiner Perkins for several years.

00:32:59
Speaker 2: Previously at Sun, if I remember correctly.

00:33:01
Speaker 1: One of the four founders, sure. And Bill, I can't remember the exact year. It was early to mid-2000s. And he was talking about this group of engineers that he came across. I think it was at Caltech that were utilizing the GPU technology. to do early AI calculations. And so this was 2005 or 2006 or something like that. And, you know, we... So I always... And, you know, the conclusion of that team and of Bill himself, one of the great pioneers of Silicon Valley, was that GPUs would be the best chip architecture to do AI calculations. So I always had that in the back of my mind. And over the years when we would visit with NVIDIA we would ask about AI and Jensen would talk about it. And it was a tiny, tiny product and solution or end market for them. And at that time, crypto mattered a heck of a lot more. But it was very fortunate in the back half of 22, crypto crashed at the same time as AI was taking off. And so that.

00:34:13
Speaker 2: Gave us- It was that simple.

00:34:15
Speaker 1: For them or- They were always working on these things, right? And it's really about market adoption more so than they're addressing it, right? And, you know, it just so happens that these things coincided. The stock market was much more focused on what was happening with crypto that drove the stock down and not as focused on this emerging opportunity in AI. And so as AI took off in the back half of 22... we were able to build a great position in NVIDIA.

00:34:47
Speaker 2: So let's talk about that run following 22. You guys had one of the best three-year runs of any hedge fund in recent memory. I'm looking for my exact numbers. 21 and 22, the whole market got slacked in 22. 21 was a rough, you're down 26% in 21, down 54% in 22. and then come screaming back in 23, 4, and 5, you're up 46%, 59%, and 37%. First of all, how much are you just holding on for dear life when you see numbers like that? What's it like to live through the regular sort of drawdowns that technology goes through? How much beta are you How much volatility are you experiencing and how do you manage around that?

00:35:46
Speaker 1: Yeah, it's very challenging. I mean, I think it was a very frustrating time, obviously, for us in 21 and 22. We came off an incredible 2020 where we played the COVID market incredibly well. We're short going into COVID emerging. It got very short, the market, and then had a tremendous run backing SaaS and e-commerce through that period of the world being in kind of a quarantine. And, you know, it was a difficult transition coming out of that for us. And so it was a really rough time. Software really got hit in 22 over the course of a month or two. And, you know, we had to reevaluate what we were doing, you know, And that was tough, you know, it's a tough time. And so I think the ability to step back and say, okay, AI is emerging and these are the incredible companies that are, you know, very well positioned for it. You know, I think, and they were trading it, what we thought were attractive valuations. And so we've just been, you know, solving for looking forward over the next, you know, six, 12, 18, 24 months. since then, and it's been very fortunate that we've been in the right place as AI has emerged.

00:37:14
Speaker 2: So let's talk a little bit about that philosophical look, and obviously AI and software is a perfect example of what you've described as long the disruptor, short the incumbent. And it's not just SaaS versus AI anymore. You could be long Uber, was an example I saw you discuss once, and short rental car companies. Walk us through those kind of trades philosophically.

00:37:44
Speaker 1: Yeah, well, you know, we don't necessarily do paired trades, but if we think there's a well-positioned solution like Uber at a certain period of time and think that's, you know, benefiting from this merger of trades, of e-commerce and for them and mobile and dominant market share, you know, we'll go along that. And if we see a company out there that's getting displaced or substituted, you know, there's short opportunities. We look at them as independent opportunities, frankly. So I think sometimes the market or the press around the stock market tries to simplify things into a, hey, this is good, this is bad.

00:38:32
Speaker 2: Um, you know, if only it was that easy, right? Yeah.

00:38:34
Speaker 1: I think, you know, I think, I think sometimes, uh, that, that leads to things getting overdone. I think software just in the last, uh, month or two has really had an incredible, uh, bounce back. The, I think people, the SAS POC cop SAS POC ellipse, if I can say it, um, you know, that, uh, that view that software is, is doomed as sort of a, a huge simplification, right? I mean, I think if you look at the history of what happens with incumbent technologies is if they solve a problem really well, they can stick around for a long time. And I think until very recently, many brokerage firms and banks are running mainframe solutions still because it works. And you know, when you get a new technology, you want to take that new technology and you want to apply it to do new things that really get you an advantage versus your competitors. You don't want to take a new technology and say, what's the boring business process that we've automated and it really works really well that we can apply this new technology to? No one does that, right? That would be a waste of innovation in a lot of ways. So those core systems don't tend to get swapped out. So you get these opportunities for bounce backs. And we're taking advantage of the doom and gloom as well as the excitement about the new things. That's what we have to do.

00:40:10
Speaker 2: Really interesting. Coming up, we continue our conversation with Glenn Kacher, founder and chief investment officer at LightStreetCapital.com. discussing the current environment for AI and beyond. I'm Barry Ritholtz. You're listening to Masters in Business on Bloomberg Radio. I'm Barry Ritholtz. You're listening to Masters in Business on Bloomberg Radio. My extra special guest this week is Glenn Kacher, the founder and chief investment officer of Lightstreet Capital, a technology-focused hedge fund located right in the heart of Silicon Valley, in palo alto so i have so many great quotes of yours uh... i want to uh... through by you i'm gonna start with variant perception we look for a mismatch in perception reality timing matters but there must be a thesis about when and how the mismatch resolves itself also that sounds pretty easy that's all you have to do tell us a little bit about identifying that variant perception.

00:41:16
Speaker 1: You know it uh... I started this by talking a little bit about why I got excited to be an investor from the beginning. And part of it is being a detective, right? And going out, talking to people firsthand, working with my team of investors that work at Light Street Capital every day. And we all operate in the same way as Roger would say, everybody goes out for a pass. And go out, talk to the people that matter, talk to the customers, talk to the suppliers. talk to the innovators themselves. And that's how we try to get it done and get the real story. I think we're in a situation today where AI is, you know, now being cast as sort of this evil empire that is going to, one, cost people jobs and, two, you know, It's crazy evil people overspending and it's certain you know like it's it's it's gonna crash and burn eventually and that's really not the story of AI the story of AI is that the end users are Self-selecting every day in their browser or now with agent software or or their development tool to build more software and And they're saying, this is how I can get more done quickly and well with these tools. And that's what's driving the demand that's creating the capacity build of AI compute. And so we look at that and say, you know, there's a mismatch in the way AI is being perceived today. And that will reverse, but you have to figure out when.

00:43:10
Speaker 2: So that's a productivity story, it's an efficiency story, and obviously it's a profitability story.

00:43:16
Speaker 1: It's a demand story.

00:43:18
Speaker 2: Which kind of raises the question, you're focused on the core AI players. What about everybody else? Forget the Mag7, the next 493 and the S & P 500. What does this mean to the rest of corporate America?

00:43:35
Speaker 1: Well, it... I don't think you can forget the Mag7. But what does it mean? But we'll put that aside. You know, what does it mean for the rest of corporate America? I think it's gotten their attention. It got their attention pretty quickly. And, you know, I think if you talk to anyone on the board of directors of a public company or the CEO and top managers, they're saying, gosh, you know, we hear bad AI. We need to come up with a plan. We need to figure out how we're going to harness this tool and make it work for us. And so that's the task at hand. I think it's still early to say, well, this company is doing a great job with AI, so we should buy their stock. That's not, to me, a great.

00:44:31
Speaker 1: You know, thesis for today for investing. But, you know, I think that everyone that I talk to in corporate America is very focused on, hey, we've got to take advantage of this tool.

00:44:43
Speaker 2: You mentioned demand is really surprising everybody. I want to say it was the second quarter, even Jensen Huang at NVIDIA was surprised for his expectations for the AI infrastructure spend by 2030. I think he bumped from $ 1 trillion to $ 4 trillion. That's just a Forex giant set of numbers. Are we running the risk of over-allocating to AI the way we did for things like fiber and go down the list of every new technology that seems to get over-allocated? At what point does this become, is this explosive upside demand going to, when does a coyote step off the cliff and not realize he's gone a little too far?

00:45:36
Speaker 1: This is the big question everyone's battling with today. And I think, you know, the MAG-7, we mentioned a minute or two ago, they have really become the key partner. I think if you look at Amazon, you look at Microsoft, Google, Those companies are partnering with Anthropic and OpenAI in order to fulfill on building this compute stack and the infrastructure to run AI. And the question is, how far ahead of demand are they planning? And the reality is they're not ahead today. They're behind.

00:46:23
Speaker 2: They're playing catch-up now.

00:46:25
Speaker 1: They're playing catch-up. The negative doomers are expecting them to over-invest, but today that's just not happening. There are bottlenecks, right? There are real bottlenecks, and it's quite well discussed, that have slowed down the ability to build. And you've got companies that are in control of some of those bottlenecks, whether it's memory... companies, which we like as well, or whether it's Taiwan Semiconductor, they're actually, they can only invest so fast. So today, demand is still running way ahead of supply. And so this doomerism that has grown up around AI, in my mind, is misplaced.

00:47:15
Speaker 2: Let's talk a little bit about the bottlenecks. I use Gemini, I use Notebook, I use Chad, I use Perplexity, but really, Claude Pro has become my favorite way to engage. And I've noticed just going from Opus to Fable, like an order of magnitude, faster, deeper, better. It doesn't feel, and these are coming along like every few weeks, it doesn't feel like this much of a bottleneck. When you say bottleneck, what are you referring to?

00:47:49
Speaker 1: Well, I think the bottleneck drives the pricing higher. than it needs to be today, right? And so, no offense, but you're probably not paying for your clawed traffic. Bloomberg may be paying for it. No, I'm paying.

00:48:05
Speaker 2: Well, my firm is paying, and it's $ 200 a month, and then we just did a whole enterprise thing, and it hasn't been crazy. Like, I keep hearing about people just going crazy on credits and spending a year's worth of credits in a month. We're pretty reasonable and a little aware of our spending, but it's not like it's $ 100, 000 a month. It's fairly reasonable for the output you get. Right.

00:48:36
Speaker 1: I've been surprised. We have our own software product and stack that we have developed on for 15 years where we run our entire research process. And so we're constantly improving that. We're also doing analysis and sentiment tracking, et cetera, of sources of data that we buy. And it's not cheap to do that.

00:49:06
Speaker 2: Are you spending $ 50, 000 a month, $ 100, 000 a month? What does it look like? What is a typical hedge fund in the tech space? Not necessarily yours, but what do you think people are spending? I know I'm only scratching the surface for what I'm doing.

00:49:20
Speaker 1: Well, for programmers, it's not uncommon to spend $ 100 a day. It can get expensive, you know, and that can add up quickly. Sure.

00:49:31
Speaker 2: $ 30, 000 a month is not nothing.

00:49:33
Speaker 1: Yeah.

00:49:33
Speaker 2: All right.

00:49:35
Speaker 1: You can spend a lot more than that, too.

00:49:36
Speaker 2: Well, you know, a couple of months ago there were stories about, wait, we had a whole budget for a year and it's gone in four weeks. Is that the bottleneck, being able to service the super clients, the hyper users like that?

00:49:50
Speaker 1: Well, that's where this demand for the open source solutions comes in, which are far cheaper, right? And so the ability to load it up on your own hardware and have it run and be able to also adjust the weightings of the model and, you know, train it on your own data, those are all very powerful opportunities for investors or, you know, just any kind of business. So, yeah. You know, being able to do more for less is certainly attractive.

00:50:24
Speaker 2: So another quote of yours, you were talking about the AI build-out, and you said, it's a 10-year cycle of demand. The bear case is that CapEx gets cut the moment returns disappoint. Tell us a little bit about why you think this demand cycle is going to go a full decade. Yeah.

00:50:45
Speaker 1: Well, I mean, I think we're changing the entire stack of computing. The only thing that looks like this that we've experienced before is the move to the Internet architecture from client-server. And, you know, these computing cycles happen about every, you know, 15 to 25 years since, you know, since the development of computing. And the way the technology works... is completely different. You know, in the old school of technology is search and retrieve or create, search and then retrieve model where you stored things in databases. And here in the AI world, the technology is essentially creating a custom solution, custom to your question, custom to your data every single time you use it. It's just a much more complex and compute-intensive model. The ability to have custom solutions and custom answers every single time you need data is so much more powerful. If we follow history, these things tend to to 15 years to become a quarter of the total capacity of the industry. So to say that it's going to take multiple decades is not much of a stretch.

00:52:24
Speaker 2: So where are we? Are we in year four or five now of a 10 to 15 year first leg?

00:52:30
Speaker 1: Yeah, exactly. We're kind of a third of the way through the first leg. I mean, if you look at the way technology develops, it sort of goes in three cycles. You know, your big infrastructure development years take, you know, five to 10 years, let's say. Then year six, year six through 16, let's say, that's when your platform or OS really gets developed and put into place. And then the applications kind of come in years 11 through 21, and applications become the dominant place where businesses invest and the innovation happens. It's at least a 15 to 20 year cycle that we're looking at.

00:53:18
Speaker 2: I'm kind of fascinated by the energy demands and the build out of these giant data centers. And I'm curious, what are your thoughts to the political pushback to where these are located? A couple of states have already banned them. I never saw the politics against tech morphing this way. How do you look at that as a investment risk?

00:53:45
Speaker 1: It's a real risk that, you know, any bottleneck that slows down the adoption of your technology is a problem, right? We're investing in NVIDIA or Taiwan Semiconductor saying, okay, here's what we expect. And in our view, the numbers are still significantly better than Wall Street's looking for. However, we have to bear in mind, you know, Is there an obstacle that's going to get in that way? Today, it's, in our view, not a big enough problem, but it's an emerging problem. And I think the way, as an industry, we have to get around this is that we have to explain the places that invest most heavily and most aggressively. You know, if you look at Northern Virginia, not far from where I grew up, you know, that is the data center capital of the world. And that opportunity and what's happening, what's happened with tax receipts in those communities that have all these large data centers and the demand for blue collar work in order to build those data centers, whether it's electricians and plumbers and construction work. It's a, it's a massive, uh, you know, shot in the arm for those economies. And then the, the tax revenue is, is an ongoing, uh, ongoing, uh, payment that happens over many years. So I think it's, uh, it's a little bit, uh, it's sad that, that, uh, some of these communities are not as, um, positive about, uh, about the opportunities. I think they're just not well-educated by uh... their elected officials uh.

00:55:32
Speaker 2: i'm not surprised that it's in virginia or new york uh... i'm enormously surprised when you see push back in places like texas which is big enough that you can stick a data center out wherever this juice and and nobody has to see it here it be concerned about it uh... but it it it keeps raising the question of cost of electricity and People seem to be concerned. We let a data center in here, our electrical costs are going to go up. How should we as a tech-savvy nation of investors respond to that concern about electricity?

00:56:13
Speaker 1: Yeah, absolutely. The source of electricity needs to be behind the meter, right? So the firm that creates the data center, if there's not enough existing energy, then they have to provide the energy.

00:56:30
Speaker 2: So run a gas line, set up your own generator, and you're off the grid.

00:56:36
Speaker 1: And if it's close, look, if it's close to a residential area, there's actually a data center that's being contemplated in San Mateo, California, not far from Palo Alto. And their solution is to put Bloom Energy servers behind, which are powered with natural gas. with almost no emissions. And they're incredibly quiet, almost no audible sound. And you can put a Bloom Energy fuel cell behind the meter. And even though that's the plan, residents have rallied against it because they've heard data centers are bad. They're just not educated on what the solution is and how it will not impact their energy prices. And there will be no emissions and no noise.

00:57:29
Speaker 2: And, and we have midterms coming up in November. Is this the sort of thing that once we get past the next group of elections, this will fade or is this really an ongoing challenge for the industry?

00:57:42
Speaker 1: It's an education challenge. Yeah.

00:57:44
Speaker 2: Yeah.

00:57:44
Speaker 1: We've got it. We've got it. And it's from local to national, right? Each project has to explain this is the, This is the decision we're making around procuring this energy. These are the number of jobs it's going to create. These are the tax revenues it's going to generate. Here's our existing energy situation. This can go on the grid without much of an impact, or we're bringing our own energy. So it's both a local and a national solution.

00:58:14
Speaker 2: Really interesting. And the MAG-7 keeps coming up. When we met in the spring in San Francisco— You liked Amazon, Google, and NVIDIA. I don't recall what your thoughts were on Microsoft. You weren't a big fan of Meta, Tesla, and Apple. How do you see the Mag 7 today? Is that still fairly consistent?

00:58:38
Speaker 1: That's fairly consistent, yes. As I said earlier, we've put Microsoft back in our portfolio, and so I'd say that they're back in the good category. The challenge for Apple is to get their AI solutions tuned up and working well for the consumer. If you think about your mobile phone, it's in a very unique position. It has both your personal and your business data to the extent that you're not a small business person. And the security is there to separate those two things. And so that device has the ability to optimize and recommend actions or solutions. to you as a consumer that address both your business life and your personal life. And that's a very unique position that Apple's in. And obviously, you carry it around and it's on most of the time, if not all the time. And they have a real opportunity to bring AI solutions to democratize them for consumers in a very complicated but elegant way. And so, If Apple can get things right, that should accelerate their opportunities or earnings over the next couple of years.

01:00:04
Speaker 2: They don't have a great history with it. Siri has been nothing less than a total embarrassment for a decade. I mean, I'm not revealing any secrets here. Everybody knows it's garbage. And there was some criticism of Apple for not jumping in with both feet. to become a hyperscale and spend tens billions dollars what they've done with google is worked out great for both companies hey what's a couple of billion dollars a year to apple and to google it's pure profit is the same sort of setup teeing up where it's a win-win for apple to integrate google's technology into the.

01:00:46
Speaker 1: Iphone potentially but it's execution based Mm-hmm.

01:00:50
Speaker 2: Isn't that always the case?

01:00:52
Speaker 1: It is, but their strategy, this is a very consistent strategy where they are the, you know, they were not the first smartphone, right? They waited. They watched what Nokia did, what BlackBerry did, and RIM BlackBerry, and then they came out with a more elegant solution after those guys established the market.

01:01:16
Speaker 2: Second, mouse gets the cheese. Is that the thinking there?

01:01:19
Speaker 1: Well, If you have a big bank account and users that really will wait around until you solve the problem in a better way, then it works.

01:01:28
Speaker 2: Last question before we get to our favorite questions. So I'm not going to ask you about 20 years out or 10 years out, but five years out, what does this technology look like? What's going to define AI for the consumer and business customer in 2031? Agents. Agents.

01:01:48
Speaker 1: The ability to have the technology working on problems when you're not directing it. That is incredibly powerful. It leads to users consuming 5X the tokens that you would consume just directing AI as you would a search engine. And so the ability to have your agent... or agents working on your personal life and solving problems as they come into your inbox or into your messaging solutions with your family and friends. And then on the business side, the same thing, solving problems for you, solving problems with your coworkers and teammates. It's incredibly powerful, this technology.

01:02:44
Speaker 2: To say the very least. All right, let's jump to our favorite questions. that we ask all of our guests starting with, and I already asked, but I got to ask a little more specifically who were the mentors who shaped your career?

01:02:58
Speaker 1: Well, you certainly have to look at Julian, uh, Robertson and, and, um, you know, he just has his, the, the example that he set and how to run a, you know, an investment business with integrity and, um, but you know, and, and, and intellectual honesty and, principles. And so that was incredible. Roger and John at Integral Capital Partners were just great as I got out of business school and was in my early 30s, really, you know, those key years of learning, again, how to run a firm and make great investments. And they gave me the opportunity to both succeed and fail in some of those private investments that I made. Those are going to be the key people that really shaped my career.

01:03:51
Speaker 2: You mentioned the two Peter Lynch books, One Up on Wall Street and Beat the Street. I know you read a lot of other research. Any other books worth mentioning these days?

01:04:01
Speaker 1: You know, I pulled a book off the shelf recently, Empires of Light, which tells the story of the propagation of electricity and the battle between Edison, General Electric, Tesla, Westinghouse.

01:04:17
Speaker 2: AC and DC.

01:04:18
Speaker 1: Yes. And, you know, incredible story. And I think at the end of the day, it was really interesting. And Edison really pushed that AC was dangerous. And to the point where he promoted it for the electric chair, because it made it made AC look bad and dangerous.

01:04:40
Speaker 2: Didn't one of them electrocute an elephant to show how dangerous it was?

01:04:43
Speaker 1: Many different animals. Yeah. And a prisoner. And it didn't go so well, actually. The first electric chair didn't work extremely well. So to scare people and say AC is bad, and you look at what's happening today with AI and people taking this incredibly powerful technology that is going to change the world and is already starting to change it, and making it this evil empire. It's pretty fascinating. And I think the other side of that is that, you know, in the end of the day, Westinghouse won out with, you know, steady execution and industrialization of the back end. And you look at the Mag 7 and what's, you know, the opportunity for Amazon and Microsoft and Google to build that back end And AI is an incredible opportunity for AWS.

01:05:42
Speaker 2: Which is already the biggest profit center for Amazon.

01:05:45
Speaker 1: Yes. And so, you know, which is not the, if you say Amazon, everyone thinks about e-commerce and they don't first think about AWS. But AWS is the more important part of the company.

01:05:55
Speaker 2: Yeah, to say the least. What are you streaming these days? I know you're on a plane pretty regularly. What are you listening to or watching to keep yourself entertained?

01:06:06
Speaker 1: Well, entertain, I mean, sure. X is entertaining. You know, all the debate around our industry is pretty fascinating. Friends and neighbors is a guilty pleasure. So, you know, that's something I'm streaming regularly.

01:06:22
Speaker 2: Anything Jon Hamm is always worth watching. Final two questions. What sort of advice would you give to a recent college grad interested in a career in either investing or technology?

01:06:34
Speaker 1: The number one thing that I tell younger folks is you have all the tools today to make an impact. And so if you want to get in the investment business, one, of course, start investing. But two, do your research, go online, and then publish your research. Put it on X. Interact with people like you, people like me. And if you can uncover... the story behind a stock and make some great recommendations, you're trying out for the world in real time. And if you have the courage to do that and you do it well, it's a no-brainer to hire that person.

01:07:21
Speaker 2: Our final question, what do you know about the world of investing and technology today? Might have been useful back in 1993 when you were first getting started.

01:07:31
Speaker 1: Yeah, I think early on and for investors coming to our market, there's this perception that things happen very fast and no doubt they're changing rapidly. But at the same time, there's this reality that things do take time. And we talked about the emergence of the smartphone. The first smartphone-like device that came out was the Newton. And it didn't really work that well. And then General Magic had a solution that also didn't really work that well. And then Palm created the first thing that actually got some adoption, but it didn't do any email or messaging. And it certainly wasn't a phone. And then Palm created the Trio, right? And then I'd say in some ways, RIM was the real first, you know, RIM BlackBerry was the first real working smartphone. but it was somewhat clunky. And some people love that clunkiness, right? And love that keyboard. But then ultimately got to Apple. And so while things happen fast, it also takes years for things to really develop. And so I think if we apply that today, AI can do some incredible things, but it's going to do way more in a few years. And there are some obstacles other than the ones we mentioned to adoption, right? Data security and comfort of your coworkers and your superiors in terms of giving access to data to an AI agent. So it will take time in order to see, you know, ultimately what it can deliver. And so I think we're just scratching the surface, even though, as I mentioned, there's a lot of battles between open source, for instance, and the closed frontier models. But there's way more to go here.

01:09:42
Speaker 2: Glenn, thank you for being so generous with your time. This has been absolutely fascinating. We have been speaking with Glenn Kacher. He is the founder and chief investment officer of Light Street Capital. If you enjoy this conversation, check out any of the 651 previous discussions we've done over the past 12 years. You can find those at Bloomberg, iTunes, Spotify, YouTube, or wherever you get your favorite podcasts. I would be remiss if I didn't thank the crack team that helps put these conversations together each week. Elizabeth Sedrin is my video producer. Anna Luke is my podcast producer. Sean Russo is my researcher. I'm Barry Ritholtz. You've been listening to Masters in Business on Bloomberg Radio.