Talking AI With Alger

Jeff Buchbinder [0:14] Hello, everyone, and welcome to another edition of LPL Market Signals. Jeff Buchbinder here, your host for this week, and we have a special guest. It is Monday, October 12th, 2026. As we're recording this, we have Brad Neuman, Director of Market Strategy at Alger. And Brad, well, first of all, thanks so much for joining today.

We greatly appreciate your time and the partnership with Alger.

Brad Neuman [0:41] A pleasure to be here. Thanks for having me.

Jeff Buchbinder [0:44] Sure thing. Sure thing. So you and I have met a few times over the years, and I've always found your insights around technology development to be particularly interesting. Of course, when we had those meetings, I didn't know what AI was really [laughs] nor had ChatGPT been rolled out. So I'm really interested in getting your sort of up-to-date thoughts on AI and technology, not just the build out, but how companies are using it.

So let's get into it. I want to focus first, Brad, on monetization. So can you, can you give us some examples of, you know, rollout adoption of AI that's really impressed you to give you confidence that all this investment's actually going to pay off?

Brad Neuman [1:33] Yeah. Well, first, let me say that the investments that you know, the people who are making these investments are mostly the hyperscalers, that's to say the very large cloud service providers who are providing cloud as a service. So those are your Amazon, Googles of the world, Microsofts plus some of the Neo Clouds you know, like Nebius and others. Those companies are doing most of the spending. You know, we can, we can see that from some of the company reporting.

And those companies have recently given some very bullish indications about what the return on investment is. First off, just in aggregate, what we've done some of our analysis on our own. And if you go back you know, you mentioned the introduction of ChatGPT, which was November 2022. If you look at the return on capital from those handful of companies, those hyperscalers back then to now, one would think that their return on invested capital, one might think that their return has dropped because they've done so many hundreds of billion dollars of investments, right?

Brad Neuman [2:36] In actuality, their return on invested capital has risen since the introduction of ChatGPT, which I think is extremely counterintuitive. And so how could that be the case that they're in this huge investment cycle, yet their return on invested capital has risen? The answer is because the individual investments, these new data centers that they're building with all of these new GPU chips in them and all the fancy equipment that cools them and connects them is providing a tremendous return on investment. We got some interesting data points in the last earnings season from Amazon, SpaceX, and others that said that these chips clusters were providing break even payoffs in two years or less, which is interesting because we believe that they last for more than five years. So in two years or less, they break even, which gives you three years of generating cashflow return.

Brad Neuman [3:30] So it's a very strong return on capital that these new data centers that are being deployed are generating. Then of course, we can get in, and I'm sure you want to who's actually you know, using the AI services as an end customer.

Jeff Buchbinder [3:47] Yeah, absolutely. The, the adoption rates have been, at least based on RAMP, been, you know, up and to the right. Clearly more companies, including us at LPL are using AI to become more efficient. But I know you have better examples than I can come with, up with, Brad. So yeah, what specific companies or specific industries would you say are really doing some powerful things that can drive productivity, expand margins because that's really what seems to be embedded into this market right now, at least high level.

Brad Neuman [4:27] So first, let me just say, you're, you're absolutely right. The adoption has been up and to the right, and it's important that it's been much faster than the internet, because I know a lot of your listeners are probably thinking, "Well, I lived through the dot-com bubble and you know making comparisons and et cetera towards, to that cycle." That took about 14 years to go to 50% penetration of U.S. households from the introduction of you know, the worldwide web to 50% penetration of U.S. households using the internet. Today in generative AI, we've achieved that same feat in three years.

So the adoption rate has been much quicker. In, in my view, AI is much more useful than if you think about, you know, 1999, you know, using the internet, which you could send an email or load a webpage, but, you know, if you loaded a webpage, you'd have to walk away and come back to it for it to load.

Brad Neuman [5:22] So the adoption rate has been much quicker. You know, we divide AI investments into AI enablers and AI adopters. So AI enablers are the companies that are building the infrastructure for artificial intelligence. Those are the companies supplying components to the data center or power to the data center, et cetera. You're asking, I think, more about AI adopters who is using AI and where is that really taking off?

So first off, it's, it's taking off the most by far in software development. So software development this year is up about tenfold over the prior year. That's not a mistake on my part. That's what the COO of GitHub has said, which is a repository for software. There's been a tremendous explosion in software.

For example, I didn't know how to write software before this year, and I still don't really understand it, but I've written hundreds of thousands of lines of code with Claude Code.

Brad Neuman [6:22] I've built many apps and websites, et cetera. So many people are doing what I've been doing. They've been generating a lot of software you know maybe as much as half of the software of those large tech companies is being written by AI. The you know, Claude Code is an agent that helps people write software. And the creator of Claude Code, the head of it has said that he hasn't written a line of code for Claude Code in many months because it writes itself, it improves itself.

It writes its own code. So software is writing software. So nowhere is the AI revolution taking off more than in software development where you're seeing huge amounts of productivity. Some, some studies showed it like 50% plus productivity improvement. So that's the first place.

Second place would probably be online advertising.

Brad Neuman [7:16] A company like Meta uses AI to place ads the right ad at the right time for the right you know, viewer. Sometimes those ads are even AI generated, but mostly it's about matching the correct ad. They said recently this year that model, that ranking has produced a 16% uplift in conversions on Facebook, a conversion being an action that the advertiser's paying for, like a purchase or a signup. So it's having a tremendous effect on online advertising. Now, your listeners probably also want to hear, you know some stuff that's non-tech related, right?

They have too, probably hear too much about tech. So let's talk about, we can talk about farming, public safety, healthcare. I really like the agriculture example because it's so low tech to see AI helping in an agricultural setting I think is really cool.

Brad Neuman [8:08] So I recently spoke to Deere and heard them say some of the following uses that they're using Deere being the you know, the tractor company, farming agriculture company. One application they're using AI for is something they call see and spray. So this is using AI for herbicides, you know, so farmers have to, you know, spray herbicide all over you know, the crops. But instead of just spraying, you know, every yard of their, you know, huge farmland, what Deere has developed was this giant pole or boom, 120-foot wide boom with 36 cameras attached to it that detects weeds. So it sprays where the herbicide needs to be sprayed and not where it doesn't need to be sprayed.

That and their test over five million acres has saved 50% of herbicide. So that's a tremendous example of AI providing a cost savings.

Brad Neuman [9:02] They also have something called JD for John Deere an AI assistant that connects all their machines and allows the farmer to figure out the optimal time to harvest based on historical trends allows the farmer to ask, you know, how did my fuel use during this tillage season compare with the last three years and helps identify best practices. And then of course much like Tesla and others, they have a driverless tractor that, you know, allows the tilling operation to be done in a much more efficient and cost-saving manner to get, to get ready for the winter or planting season. So I could go on and on, happy to, but those are some of the examples.

Jeff Buchbinder [9:45] Wow, those are great. Yeah, so the industrial sector has certainly been kind of in the middle of this AI build out. You know, people talk about Caterpillar and the construction of the data centers and the cooling that you mentioned, but the farming angle, I didn't, I didn't think about. So that's, that's really a really interesting example. And I think it, you know, it shows all of our listeners that you know, this is more than just writing code.

But the writing code thing is really interesting because, you know, tech companies obviously benefit from cheap code, right? But every company can benefit from cheap code and not having to hire so many software engineers.

Brad Neuman [10:26] I think there's two fascinating things about it. One, first of all, the speed, you know, we've been trained that progress happens at a certain speed, but this is really, you know, the first time in history where the tool that we're inventing used tools itself. You know, normally we, you know create an axe that cuts better or we create an engine that takes us around faster. But here we created AI and AI is using AI tools. So it's software writing software, and that really gets the growth going exponentially.

And so I think that is going to be really powerful. And then I think the other thing is that I mentioned the explosion in software. I think we're all going to have really more customized software. You know, we you know, if you're, you know, if in the kind of age range that we are, you know, you think about buying software, you know, you remember the days when you bought software literally off the shelf and it was literally in a package and you put the CD in your computer.

Brad Neuman [11:21] Now, obviously everyone downloads it, but, you know, it was basically, like, one size fits all. I think, you know, in the future, everyone's going to have customized software. In fact, I think, you know, AI, your own personal AI agent will probably make your own software. And, like, I, like, I developed AI software or with AI, I developed software for my family that just does all our scheduling in the house and tells me, you know, where I have to drive my kids and drop them off and all those things and integrates all our calendars. So I think there's going to be a lot more people operating with software that helps them live their lives better and more efficiently.

Jeff Buchbinder [11:55] Yeah, and certainly one example of the demand for these agents among consumers is the fastest adoption of Muse AI agent, certainly. So yeah, the demand is there. It's, it's more than just selling cloud services, right? It's more than justifying the investment in these data centers just to rent compute.

Brad Neuman [12:18] Well, yes. And, and you hit, so ultimately, you know, we, the way I think about this whole thing is supply and demand of a very important economic unit. You know, we all, we all watch TV and read the newspapers, and we're all talking about barrels of oil because of what's happening in the Middle East, of course. But the, I believe the most important economic unit is not a barrel of oil but a token of artificial intelligence. A token is like a word or a fraction of a word.

And when we talk about this build out, we're talking about the supply and demand of these tokens. So you mentioned agents, you know, Muse being, you know, one of the first mass adopt public agent. Of course, Claude Code is another example of an agent that's been used more by enterprises.

Brad Neuman [13:02] But these agents are very token hungry. Like, you know, if you type in, into an AI query, you know, what's the capital of France, you get back an answer, it's just a sentence, that use about 10 tokens. If you ask AI to, you know, write software like I was talking about, that's using hundreds of thousands of tokens. And so the reason why this is such a big investment opportunity, and there's going to be so much infrastructure, trillions of dollars over the next year, is because as that token demand takes off, we need to build token factories. And those token factories are these data centers with all of that infrastructure in them.

So you're absolutely right that AI agents are driving an acceleration in token demand, and that's why there's such an investment opportunity in the token supplier infrastructure.

Jeff Buchbinder [13:47] So how about, I mean, you mentioned some, what might be considered investment opportunities, right, maybe where investors are underestimating just how powerful the benefits are from AI. Like, when you think about enablers and adopters, or maybe the various industries within those buckets, are there certain areas that you think are really particularly attractive where maybe the market's just missing the value here? And at the same time, maybe the opposite? Maybe the market's getting too excited about something, and if we have open source, you know, cheap Chinese models, maybe there's no you know, there's no reason to put such high values on certain companies. What, what comes to mind?

Brad Neuman [14:26] Well, so I think, I think we've been so successful over the past few years for really two big insights. One is that the AI total addressable market is bigger than people think, which we continue to believe. And I think we've been right about that for a little while now. Second thing is understanding how the data center will evolve. In other words, like, what is the new this increasing accelerating demand for AI tokens, what does that mean?

How does that mean the data center has to change? So in the beginning, you know, which is now very obvious to most folks, is that the CPU had to be swapped out for GPUs because GPUs are what process in parallel and they're necessary for AI to for generative AI to function. You know, and then those GPUs were a lot hotter than the CPUs, and so air cooling had to turn to liquid cooling.

Brad Neuman [15:19] Now there's a lot of transformation in the connections from copper to fiber optics. The power inside the data center is changing to a much higher voltage so there can be more power in these chips. So those are some big themes that we're interested in right now and on stocks on the, in supplying the power and converting the power inside the data center and the transmissions the fiber optic cables and those are, and that optical equipment. And then I think a big area is probably the biggest bottleneck is power to the data center. And power has historically, well, hasn't, hasn't grown that much prior to ChatGPT in the past, you know, few decades.

And so growing at all is a big challenge for utilities.

Brad Neuman [16:14] In fact, it takes a long time for utilities to add power generation equipment, and it takes several years for new data centers to be connected to the public utility. So one area of big investment for us is power behind the meter, which means when a power generation equipment is not purchased by the utility, but purchased outside of the utility by the data center and works next to the data center. And that could be a gas turbine like a utility would purchase, or it could be solar, or it could be nuclear, or it could be geothermal even, we believe or something called solid oxide fuel cells. So those are all things we're investing in, the components of them. There's even really innovative ways that the economy is attacking this because we are so power starved.

Brad Neuman [17:12] There's a company that we've invested in that used to or still does refurbish jet engines and lease them out. And now they're taking those jet engines and actually using them to produce power behind the meter next to the data center. So I think that's a big area. And then, and then a kind of an interesting area that's a combination between you know, the adopters and the enablers, I think is healthcare. Healthcare is a sector that we were underweight, say, six months ago, and now we're much more positive on because of the convergence of AI in healthcare.

And particularly having AI now be able to design new DNA and proteins to create new drugs. We think that's going to lead to a huge increase in the amount of experiments that are being run.

Brad Neuman [18:03] So the companies that provide those lab experiments, that generate that DNA those consumables the filtration equipment, all of those things we think are going to be in high demand.

Jeff Buchbinder [18:15] Yeah. Not only is there personalized compute or personalized agents, there's also personalized medicine.

Brad Neuman [18:22] Yeah. I, one of my favorite trials and examples of, in medicine in recent history has been this Moderna customized vaccine for melanoma. They recently had a trial that was successful. And you know some cancers, many of the they have very common mutations. Because melanoma is caused by the sun, many people's mutations in their individual cancer differ.

Like, you know, what happened because of the sun get melanoma. And so it presents a problem. And what Moderna was able to do using AI was able to customize vaccine for each of those different mutations. And it was able to do it at scale and worked across, you know, a broad set of different mutations. So that's the power of AI, which we're, we're seeing work a lot in healthcare, but that's one great example.

Jeff Buchbinder [19:17] Yeah. We, we at LPL have been very interested in that piece. We've been at least positively biased toward healthcare, not fully overweight at this point, but definitely have owned some biotech and some pharma and have been impressed with what's, what's going on there. So really good stuff, Brad. The are there risks here?

Like, certainly a lot of people point to the overbuilding risk like we had in the late '90s with dot-com. How would you compare this to then? Is overbuilding the biggest risk? Is there some other risks that you think people should be paying attention to, even if they're optimistic like I am about the promise of AI?

Brad Neuman [20:00] So I normally divide up the risks into demand and supply. You know, some investors I find are worried about demand. That is, like, is the return good enough? You know, is this really going to be useful? You know, it's cool, but is there really a return from the actual implementation of it?

That I'm not so concerned about, at least at the moment, because I see it accelerating due to agents and everything you and I talked about. On the supply side, I'm more concerned. I think there are more risks, not on the overbuilding, which I'll address in a second, but on the ability to have enough supply. I think that the issues that we're dealing with in terms of regulations and heading into the midterms there's a lot of backlash, a lot of NIMBY, not in my backyard type of arguments against, you know, the electricity costs the noise, the water, et cetera usage of data centers.

Brad Neuman [20:54] Those things worry me, but at the same time, they elongate the cycle. So it's both good and bad. I could see a certain amount of data centers, or I believe that a certain amount of data centers will be canceled because of these issues and not come online in time. And I think that could be a speed bump for the market. But ultimately, like I said, that keeps supply and demand tight, and we see supply and demand is very tight because we watch, among other things, the price of compute, which is how much it costs to rent out these processors, and they are increasing very rapidly over any time period.

You know, just in the past few months, there's been significant steps up. But no, I don't think it's, I don't think it's really anything like the dot-com bubble, and I know that a lot of people want to draw parallels.

Brad Neuman [21:44] The biggest one I can give you is that back then, the amount of fiber laid was really, you know, if you build it, they will come. It was built in anticipation of demand that didn't materialize on time. And there was only, you know, single digit capacity utilization of that fiber at that time, meaning most of it was never lit up, it was unused. Today according to reports, we have 99% capacity utilization at data centers, so very different than the single digit capacity utilization we had of fiber during the dot-com era. And I think that I think that AI is just much more useful today.

And the companies that are building it, they have a backlog, just those several hyperscalers that I mentioned, their backlog is $1.7 trillion, and it's actually increased faster than their CapEx has increased over the past year.

Brad Neuman [22:39] So they're not building if customers will come, they're building slower than customers are coming. So I think it's a very different scenario than what it was back in the dot-com era. Yeah,

Jeff Buchbinder [22:50] That, that's a, that's a great way to think about it. I know that hyperscaler cloud computing revenue accelerated last quarter, and I think analysts are expecting it to accelerate again this quarter. That clearly is cause for optimism, if that plays out here, in terms of these hyperscalers actually getting the payoffs. And then as you know, the, some of these, you know, contracts for memory are just chips, GPUs, CPUs, power, whatever. A lot of these contracts are basically telling you that we're booked up through 2028.

Yeah. I'm assuming you're seeing the same thing. Like, if you know the demand is there through 2028, or maybe even beyond because of the bottlenecks in power or regulatory pushback or anything else, and we can talk about safety maybe briefly too here. It seems to me like the risk of overbuilding and having waste is pretty low through 2028.

Jeff Buchbinder [23:53] I assume you agree with that.

Brad Neuman [23:55] Yeah. Yeah, we do agree. You know, the hyperscalers have all commented that their demand exceeds supply for certainly 27. They're, they're, you know, hand to mouth, not even. 28, they see the same thing.

I could give you an example of several companies who are booked out in terms of components through 29. We have examples of companies in certain parts of the system that are booked out through 2030. So yeah, they're, it's, it's obviously very tight in most of the supply chain.

Jeff Buchbinder [24:24] Yeah. And there's really no sort of big wave of pets.com type businesses here that are just trying to figure out how they can get value in the marketplace from eyeballs.

Brad Neuman [24:36] And the multiples are much more attractive than they were in the dot-com era. You know, I won't, - Great point. You know, you know, the bellwether of, you know, the build out back in the dot-com era with Cisco Systems, I think it traded 125 times earnings at the peak. You know, Nvidia is, you know, a small fraction of that. You know, it's extremely reasonable multiple on consensus.

And we think consensus is probably too conservative. You know, you look at the memory stocks, they are trading at single digit multiples of earnings. These are very low multiples and nothing like, you know, the unprofitable companies that we had in the dot-com era. So I think it's really different today in that sense. That's

Jeff Buchbinder [25:17] A great way to look at it in terms of both the capacity utilization back then and now, and then the multiples, which are much cheaper for the folks building this out. And these are the best cashflow-generating companies ever. Now now obviously, their cash flow is depressed, but these are the best cashflow-generating businesses that the world has ever seen coming into this. So that certainly gives us confidence at LPL that we're not going to have another dot-com bust before the end of the decade. At some point, there probably will be some waste, and it is tough to see these turns before they come, but if that several years out, the, then - Yeah.

Investors probably won't worry too much about it in Q3 earnings season. So so if you want to comment on AI safety, feel free, but I want to hear what you're looking for during earnings season.

Jeff Buchbinder [26:08] Anything that investors should be watching here before we wrap up?

Brad Neuman [26:14] Well, you know, this earnings growth this year is historical. You know, it's, it's well in excess of 30% growth for the year, which is, you know, basically the highest we've ever seen, excluding coming out of a recession. So, you know, we're it's, it's historically essentially unprecedented. In fact, going into this earnings season, positive pre-announcements are the highest they've ever been. So that's very reassuring.

I mentioned to you some of these data points on demand accelerating. We just got Taiwan semiconductors revenue numbers publicly reported for September. It accelerated to 55%. We're seeing as I mentioned compute prices increase. So, you know, the things that we're watching are the things that we were have been have had our eye on, which is demand supply, like we've been talking about.

Brad Neuman [27:07] The price of compute, I think, is a big one. You know, the ability to bring on capacity. And, and if, you know, we're, if there's cancellations like I was saying, which I think is the, you know, the bigger risk and the ability for that supply to come on. So those are all things that we're watching.

Jeff Buchbinder [27:26] Great. That, that's a great list. So in terms of safety, so maybe the skeptics would say something could go wrong in terms of safety that could slow down the build out and the adoption. Is, is that a risk that people should be thinking about here? It's certainly gotten a lot of PR.

Brad Neuman [27:44] It, it's obviously a risk. I think it's, in some ways, it's maybe less of a risk than the market had reacted to initially because there doesn't seem to be an appetite from the federal government to restrict or put on significant regulations. And as long as that, as long as they leave it to the market, then the market will continue to run. Now, obviously, there could be an accident and you know, I'm sure at some point there will be certain type of accident. I don't, you know, no one knows how severe that will be.

Hopefully it's not you know, life-threatening and hopefully it's, it's somewhat you know, as minor as it could be. But it there will be some kind of issue. You know, I think ultimately, though, where we're going to be is we're going to be in a situation where only AI can police AI.

Brad Neuman [28:35] And so accidents, while they may have you know, a negative impact in the stock market on that day or that week or month or quarter, I think ultimately, it may actually be positive for AI because, you know, there's going to be good and bad AI actors. I don't know that it's so dissimilar from nuclear weapons. Companies, just like countries, will want, and countries will want their own AI just like they want their own defense systems. So there will need to be a duplicative build out, which will accelerate the spending on AI. And, companies and people will all have their own AI that will do their best to keep them safe, to figure out what's real and what's not real, and you know cut through all of the you know, fraud and and evildoers out there.

Brad Neuman [29:24] But, you know, they're always going to have some good and bad. But I think AI is both the culprit and the solution, which means, you know, we're going to be living in an AI world.

Jeff Buchbinder [29:32] And certainly you can look at the charts of some of the leading cybersecurity stocks, and that tells you how much people are valuing cyber defense in this environment. So I agree, that's going to be the solution, and hopefully this build out and adoption ramp up can happen safely, or safely as can reasonably expected when these agents in sandboxes are supposed to stay off the internet and they find a way on or whatever they're doing. I understand why that's making folks nervous, but I share your confidence. We'll get through it. Yeah.

So I think we'll go and wrap there. So, Brad, thank you so much for sharing your insights. This is really helpful. A lot of very important topics that I think our many listeners are very interested in.

Jeff Buchbinder [30:26] So thank you for taking the time. Thanks so much to Alger for the partnership. And yeah, I'm sure as this evolution or revolution continues we'll, we'd love to have you back on to talk about what you're seeing maybe in AI's next leg. Yeah.

Brad Neuman [30:45] AI's moving so fast. We, we have to do it again soon. Thanks for having me.

Jeff Buchbinder [30:48] Sounds great. So thanks to everybody out there for listening to another edition of LPL Market Signals. We greatly appreciate it. Have a great week. Take care.

We'll see you next time.