From Memory to Software: Threats AI Presents to Tech Sector

back. It's now time for the Watch List panel. We are talking about the state of the AI trade. So joining us today, John Freeman, co-founder, Senior analyst, Ravenswood's partners. And Kai Wu, founder, Chief Investment Officer at Sparkling Capital. Thank you to both of you for joining me today. John, let's start things with you. The SaaS apocalypse, as we describe it, under reaction or overreaction, do you think? So I actually think it's an overreaction. And I've my you know, I've spent my most of my career looking for good opportunities in the software space specifically, you know, I've covered all of tech, but software has always been a great, you know, business over and over again. Right. And I think now, so it took me a while to come to the realization that these auto coding tools, maybe they can't do it today, but if you give them five years, ten years, they're going to be able to replicate all of, you know, most of enterprise software really. So, you know, that that I think puts a long, you know, ten, ten, 15 years out right there. You start to look at a terminal value for the business as a possibility, right? Whereas that was never the case before. So that's kind of how I view software. There are some exceptions. I think particularly in the next five years, I mean, we're talking about, you know, still that that narrative is, you know, can build and be dangerous. But there are companies like ServiceNow that I think are doing stuff, you know, like managing AI agents that, you know, would be a significant revenue offset to anything they lose over time. So that's my that's my take on software and AI. Understood. I mean, many have said it's overblown. So it's nice to obviously see that balance and that debate in this market. I mean, what would you say? I mean, you know, particularly because we've started to see investors wrap their arms a lot more around the software space lately, particularly as they sort of sift through the winners and the losers. And as John mentioned, ServiceNow is a name that we're going to be hearing from this week. You just wonder what they've got to say in all of this. Yeah. Look, I've spent a lot of time studying not just this cycle, but but historical disruptive periods. So think about like brick and mortar retail, other periods where technology has gone in and disrupted the moats of existing incumbents. One thing you find consistently is, you know, obviously the headwinds, but more importantly, increased dispersion. And I think, you know, this is something that we are seeing now where, yes, many of these stocks are down, you know, 60, 70%. You mentioned ServiceNow Adobe Salesforce Intuit, many firms, which up until today were kind of darlings of the growth stock investor class, are now being challenged. That being said, you know, when you study the historical disruptions, what you find is that there are really interesting examples of companies that do indeed survive and thrive through disruptions. Examples, for example, would be like New York Times in like the newspaper decimation or Walmart in the retail apocalypse. And so I think you have to step back and ask the question, which is in which for which of these companies was the moat just the code, right? If you're only advantage is that you can you have technical ability to code, then yeah, you're a little bit in trouble now because these AI coding tools are able to now make what was once a moat into a, you know, abundant commodity. However, to the extent your moat was not the code, but instead was, say, customer relationships, brand equity, network effects or other intangible assets that go outside the technical capabilities of AI coding agents. I think, you know, many of these companies, their stocks are down big, yet they, I think, are potentially positioned to survive as we move forward. Yeah, you raise an interesting point because we do have a number of disruptors coming on our program. And I always ask them, what is it that the incumbents cannot do themselves that makes your product so much better? And, you know, it's a really interesting conversation. So you're right in saying, obviously, those who are able to do what they can with artificial intelligence, because they already have, of course, the loyalty there, it's going to be very interesting to see how this all plays out. Guys, I do want to get your thoughts right now on what is happening on memory. So I'll kick things off with you, John, because I wanted to get your take on what we heard from SK Group, obviously, the company behind SK Hynix, which is all the rage right now. The chairman has reportedly warned that memory prices are abnormally high. We know that. And that supply must be increased to prevent chip flation. How are you thinking about all of this? Well, we definitely have inflation. Prices have definitely, you know, increased substantially. Right. There's there's no doubt about that. Like, you know, and now you have memory companies who normally do 30%, 40% gross margin in a good year, 50% gross margin in a great year, they're doing 80% gross margin plus, right? And they're growing very rapidly. And yet they are priced like cyclical companies. This is the same kind of cycle that I saw with semi cap equipment back in around 2010. And you had stalwarts that really broke out of that cyclicality, Lam Research and ASML, which are still, by the way, very good stocks I think. But you know, so so that's, that's what people look at memory and they and they only see the cyclical business and they, you know, they bet on a downturn. But we're talking about AI CapEx. And so right now, at least for the foreseeable future, we're talking it's gobbling up massive amounts of memory. If we're talking, you know, have large spatial models that start training video content on, on video content rather than text. That's another, you know, five x ten x sort of increase in memory requirement. It's the single thing of networking. You need a lot more compute. You need a lot more network throughput. You need a lot more memory. But memory is the one, I think, particularly at these stocks trading at six and seven times and so forth earnings, I think they're still very attractive. And Kai going into this reporting season obviously we'll start to hear from some hyperscalers this week. Does it change what we've heard from the likes of the Chinese companies more recently around some of the models that they've been developing and coming out with how these companies here in the US are actually managing costs. How closely do you think all of this is going to be scrutinized in light of what, say, for instance, the Chinese are doing right now? Yeah. Look, I mean, since the Kimi release and a lot of the news of improved models out of China that are open source and more importantly, open weight, and we have started to see a kind of rotation out of the, you know, big trade that's powered the market for the past year, right. Which is, you know, semiconductors memory hyperscalers, anything that is really the AI CapEx. And I think the concern is it's, it's mainly driven by the kind of obvious, the obvious observation that, look, if we need less compute to run an equivalent model, then, you know, perhaps, you know, the, the, the build out may be overdone. Right? We saw this in the, in the internet boom with the telecoms. We saw this in the railroads 100 years before. So I think there's really there's there's very real risk. I think the countervailing argument would be that, look, there's this kind of this is the Jevons paradox and point that, look, if the price of a commodity goes down, then demand may go up. And if it's elastic enough, you may end up in a situation where actually, you know, so many other use cases for AI that were previously too expensive are now available. And then the, the demand and quantity of consumption of this, you know, goes up. So I think like, you know, the, the chip stocks, which, you know, we just discussed, you know, that this affects them less because whether or not it's powered by an open source or closed source model, they stand to benefit. You know, it's it is concerning, though, for the open eyes of the world and the closed source labs up until this point, you know, they have yet to IPO. But, you know, trillion dollar valuations are being thrown around. We're seeing increased competition competition not just for from Chinese open source, but even other AI, US based AI companies like the Meta's and Google's. So I think this has become a very crowded space. And, you know, there is a legitimate state of the world whereby AI becomes a game changing technology that, you know, completely revolutionizes how society functions. Yet the actual production of intelligence, these models end up becoming utilities end up being commoditized, in which state of the world the value may flow to other parts of the stack outside of just the labs themselves. Yeah. And as yet, my previous guest said only 2% of us are actually paying for a lot of these models. So it's really interesting, guys, we've got to leave it there. Thank you so much for joining us today. John Freeman Co-Founder, Senior