John Tinsman: AI Stocks Are a Generational Trade
Transcript — TFTC
Article: https://www.tftc.io/john-tinsman-ai-stocks-generational-trade
Transcript page: https://www.tftc.io/john-tinsman-ai-stocks-generational-trade-transcript
Published: 2026-08-05
Machine transcription, lightly cleaned; may contain errors.
========================================================================

[0:01] Marty Bent: John Tinsman, welcome to the show.

[0:04] John Tinsman: It's a pleasure to be here, Marty. Thank you.

[0:06] Marty Bent: Pleasure to have you on such short notice, too. I think we threw this together less than 12 hours ago. I'm excited to have this conversation because we've we've had analysts covering the AI space on quite a few times over the last year. Jordy Visser, probably the most prominent. I've really liked his coverage and was telling you before we hit record, I've I've found some clips of you talking about your thesis and what you're doing. And said I had to reach out to John, get him on the show. And for anybody who's unaware, John is the founder and portfolio manager of AOT Invest, which trades on NASDAQ.

[1:19] John Tinsman: Yeah, so my first ETF I launched in 2022 is AOTG. That's the ticker, the AOT Growth and Innovation ETF. so you know, I came from the world of market making, which was all about high speed and algorithm and h algorithmic trading and statistical probability. So I tried to take that statistical probability to long-term investing. And so when you invest, you know, 10 years from now, you can almost guarantee that the stocks that have gone up the most. Have had the best earnings growth. So, like in 10 years, if there's a company that's earning 30x more, that is probably have gone up quite a bit. and So I think like when you invest for the future, you think, well, what companies give me the best probability of earning the most in the future?

[3:27] John Tinsman: reasonable valuation metrics possible. So if there's something like super overpriced like Palantir, it won't be in my fund. And you know, even if it is decent growth, like why would you even bother wasting your ammunition on it if you can buy something at one-tenth the valuation metrics and three times the growth? So I take that approach with AOTG. And then number two, the second highest thing I found correlated with long-term stock price performance in my research was low marginal cost. And I think this one's really interesting because nobody else was focusing on it. And that's the second criteria for AOTG, high growth and low marginal cost.

[5:17] John Tinsman: It was hard for me to wrap my head around that being you know, wanting to grow and do better than the year before every year. so and so I take that investment approach, that I've learned from personal, you know, managing family businesses and took it to an ETF and focus on that high growth, low marginal cost within profitable businesses. And so it's been out for, you know, three and a half or four years and it's up, you know, it launched at 24 on the NASDAQ and it's trading, I think it traded yesterday in the 63. So in in that short amount of time, it's up over 160%. so it's been great to do well for shareholders, of which I'm a big one, of course. It's like, I have all my money in. Yeah.

[5:59] Marty Bent: Well, I mean jumping r like deeper into the framework and particularly I think your your hottest thesis right now, which is being all over this XAI anthropic deal, which was announced I believe last week or the week before, where Anthropic, I believe they're paying what, one point nine, one point six to one point nine billion a month in compute from one of the Colossus sites.

[6:17] John Tinsman: Yeah. Yeah, yeah, and I tweeted about it because I hadn't seen anybody talk about it. I'd seen people talk about, they're leasing it, but nobody had talked about the profitability metrics. They were open up the, they opened up, it's so crazy. Marty, they opened up the S1 and they're like, ha ha ha, look at how much money XAI is losing. And in that, they basically had $15 billion a year coming online like next month. So like in like three days. And they somehow didn't realize that it's gonna be making money. Like, like not only is the XAI revenue gonna jump 300%, maybe a lot more, but year over year, but also they're gonna be making a ton of money.

[8:23] John Tinsman: The hyperscalers are spending a trillion dollars now this year building them. And you know, did you know like Microsoft and Google historically their return on capital invested is over 35%? And why are they spending so much more now? Is the ROI trending upwards to a hundred percent? And if you look at Google Cloud, if you look at Microsoft Azure, if you look at Amazon Web Services, Oracle, it seems like it quite possibly is. Like The revenue growth is accelerating, the profit margins are getting fatter and fatter. And I think there's a chance that what we're doing is we're building data centers in the US and we're leasing it out to the rest of the world at like 90% profit margins.

[10:18] Marty Bent: Can't be dooming. Can't be dooming, especially not right now. And it's funny, today in our newsletter we actually covered it. We've been sluicing the banking reports. So one came out of Morgan Stanley yesterday where they're they're citing these these CapEx numbers and they're warning they're saying, has AI made the US economy inelastic? Hyperscaler CapEx estimates nearly doubled to eight hundred and five billion for twenty twenty six, headed towards one point one trillion in twenty twenty seven. US GDP has been revised up to plus two point three percent SP earnings growth.

[10:43] John Tinsman: Mm-hmm.

[10:47] Marty Bent: Revised up plus twenty three percent. Their thesis is that Google, Microsoft, Meta, and Amazon aren't cutting CapEx in response to higher rates because AI spending is too strategic, too essential, or too well funded to care. if the biggest spenders in the economy are rate insensitive, the Fed's transmission mechanism is weaker than at any point in history. So th they're making a a comment on this CapEx spending as juxtaposed to to high rates and and what the Fed may want to do moving forward. I think to your point, what people haven't really grocked is the ROI on these CapEx investments, particularly if you're you own the compute and the energy infrastructure is so high it's hard to pass up.

[11:27] John Tinsman: Yeah, I mean I think there's two things to think there. One, like if the Fed raises rates one or two percent and your ROI is a hundred percent, or actually like an XAI's case, like a thousand percent, do you even care? And of course the answer is of course not. The second thing is you have to remember all these hyperscalers were sitting on like more cash together than the US Treasury had. So they were sitting on huge amounts of cash. And somehow we're supposed to be sad that they're spending it, right? And they're spending it with like the most absurd ROI we've ever heard of.

[13:43] Marty Bent: Yeah, and I mean and to this point too, what are the barriers to entry to this this compute leasing market? is it the constraints of the energy grid and access to to power and power infrastructure? Like how how how long can these ROIs persist? Is it is it because of the the supply of the infrastructure is still constrained and

[14:03] John Tinsman: Well, we have a lot of natural gas in the US. And so like in Iowa, what they do, we have natural gas pipelines that run up from like the shale drillers. And what in Iowa what they do is they build their own generators onto the data centers. And then they try to get like wind energy too and things like that. But like generally they're building they're bringing their own power. So is it a concern if we use more natural gas when we have plenty of it and we're exporting it to the rest of the world? Like, I don't think so because it's pretty stupid. To put natural gas on a natural gas tanker and send it to Europe to power a data center over there and raise the cost of natural gas like 20 bucks.

[15:37] Marty Bent: Yeah, that Memphis build out. I was telling you before we had recorded, I was been in the Bitcoin mining industry for for almost ten years now, and watching how they did Colossus One by daisy chaining the gen sets and putting the the battery powers, the battery walls inside it he he was acting like a Bitcoin miner, 'cause a Bitcoin mining to point, the the whole name of the game is is basically shortening the the distance between the power production and the compute and it seems like AI is adopting that quickly as well. And I mean to that point it looks like SpaceX is going to build another another large data center classes too. It looks like it may be a seventeen billion dollar outlay.

[16:19] John Tinsman: Yeah. I think. Yeah. For two gigawatts. So like if you can lease out three hundred megawatts for forty five billion, how much can you lease out eight times more compute for? Like if they only spend seventeen billion, it sounds like at the current rate they could lease that out for over a hundred billion. I'm sorry. They could lease that out for over I was I was way low on that. But they could lease that out for over three hundred billion. So Like, I mean, if you do a 10x ROI on it, like they're getting on whatever, I mean, it's it's a lot. And how's that gonna be for SpaceX shareholders?

[18:34] Marty Bent: Well, I mean, that begs another question too. Like how many winners could there be? I could see there being a lot of winners in this just 'cause of the insatiable demand for for tokens. And it doesn't seem like to your point of I forget which bank was was saying it, but the the growth and token usage just in the last year. And if you think about robotics and things that haven't even come to market at scale yet that are gonna demand probably exponential amounts of of compute compared to the nation agentic economy that exists now.

[18:38] John Tinsman: Yeah. Yeah.

[19:03] Marty Bent: You could see a ton of winners and not only and I I think that begs another question I I mentioned, I sit on the board of Cathedra. we're in the process of merging with a company called Sphere 3D, but we operate on a lower scale, like twenty to twenty five to fifty megawatt modular data centers. And excuse me?

[19:18] John Tinsman: Nice. Well I hope you have stock based compensation. I said I hope you have stock based compensation. Some stock options in case it spikes.

[19:24] Marty Bent: Well that's I mean so go going back to going back to I mean, how many winners could there be? Could you could you see a scenario where there's a ton of winners that are that are across the board different scales? Yeah.

[19:36] John Tinsman: of course. I think that's where we already are. The question so we've already seen a ton of winners in the semiconductors, because first they have to order the stuff, right? So we've seen all this like we've seen Sandisk and Micron. I mean Micron was up like 20% the other day, like so many back-to-back five, 10% days. It's crazy. And And I think the thing is when you have like a 200% ROI and you really want this stuff because you want to get this data center built right now, how much pricing power does it give the semiconductors? And I think they're figuring that out. AMD, you know, Nvidia, they can continue to raise their prices because they these these semiconductors that are crucial to to the compute are generating such a high profit for the people that have the data center.

[21:47] John Tinsman: I mean, I'm pretty sure they're up like five hundred percent in the last year. So there's other people that are, you know, that are moving before us, I think, on this one.

[21:55] Marty Bent: Yeah, and and to that point, when when do you think it'll the we'll know for sure that that this is happening? Like you you don't wanna wait for those those earning reports. Is this Q three, Q four this year, Q one next year? When do you think all these numbers will hit?

[22:10] John Tinsman: Well, I think like a month ago, people s people in the industry started to talk about it. And one of the people that posted it was like David Sachs. And he's like, Hey guys, like you can build a fifty billion dollar set data center and lease it out for thirty billion a year. And then like we just learned with what Elon Musk did that his numbers were way low. Like way low. You can build a thirty fifty billion dollar data center and maybe lease it out for, you know, a hundred and fifty billion a year. So like he was off by a magnitude of five.

[24:01] Marty Bent: What what would slow this down? I mean, we've seen the headlines in the last week. Uber has run through their their compute budget for the year. They ran through it, I I believe, mid April. You have many other companies saying, hey, the the use of of these models isn't helping us add add bottom line revenue to to the business. Like is what if like what could slow this down? Is it the inability for companies to actually integrate

[24:20] John Tinsman: Yeah.

[24:28] Marty Bent: these models in a way that that makes them more productive, more efficient, and most importantly more profitable. what could pop a hole in this this gross story right now?

[24:38] John Tinsman: So I think the interesting thing is anybody buying these tokens in the first place has probably already really done the math on it because you'd be pretty stupid to just buy tons of them and make it more expensive than human labor. Okay, so like imagine this, you're sitting in a data call center and your job is literally just to answer the phone and and take customer service calls. You know, so they're paying everybody twenty-five bucks an hour, you have t thousands of employees sitting in this call center. Like you could have a chat bot that's like, okay, like we've seen that you, you know, got a fee on this credit card or this card, this payment was fraudulent, like we're gonna go ahead and reverse it and send you a new card.

[26:44] John Tinsman: Iowa and everything, they're gonna be buying those tokens from these companies, and the tokens are gonna be made in the United States. They're not gonna be made in Europe where natural gas is 50 times higher, they're not gonna be made in Africa, they're not gonna be made in China. So, like, we really are the answer to the whole world here. And you cannot like, do you remember like Google, like hundred billion dollar market cap, and people are like. it's so big, it can't grow anymore. You know what I mean? And they just forget that Google can go and sell their ad words to the entire rest of the world at 100% profit margins.

[28:02] Marty Bent: Yeah.

[28:12] John Tinsman: You know, like isn't that better that we have like the place where we can break the red type and put up faster than anybody else anywhere in the world? So like that that just gives us a bigger moat and it makes our it makes investing in safer.

[28:24] Marty Bent: Yeah, I mean I and two points to add there. I think when it comes to implementation of these AI workflows or agentic workflows, i it really comes down to being able to implement I th I think Jack Dorsey and Block are a great example. I've watched their implementation of AI into their company over the last year, and I think they're doing it the right way. And I think Jack's Interview with the Sequoia partners a couple of months ago really highlighted how they're viewing it. And I think the approach to how you actually implement these AI systems is is very important. There's gonna be a ton of companies who don't think strategically about that those implement d implementation details that that mess it up and they wind up wasting a bunch of money. And then on the other hand, like I think for small teams like my humble media company that we're running here, we've been running an open claw agent since January and it is it is

[29:15] John Tinsman: No way. Nice.

[29:18] Marty Bent: expanded our ability to do what we do here by orders of magnitude. And I think for small teams it's going to be particularly powerful. And then when you think about the profit margins that can be gained by small teams that implement this from zero, the the the profitability of of the smaller team is going to way exceed the larger companies that have a ton of bloat. And and I think that's going to be actually an interesting

[29:30] John Tinsman: Yeah.

[29:44] Marty Bent: theme to watch moving forward is the the incumbent behemoths that were using headcount as a KPI back in the easy money days of the two thousand tens and early twenty twenties, trying to unwind the mistake of that KPI.

[29:59] John Tinsman: Yeah, I mean, how amazing is it that like Elon Musk went in and like cut eighty percent to X and it worked just the same. You know what I mean? And I do think there's a thing of like When people throw crazy money out of s at things, you there's like it's kind of infamous for getting ripped off. And so what companies are run so well that they can build them the leanest? And I think they'll have a huge competitive advantage, right? To build them the fastest and the leanest. so you probably I honestly it sounds like maybe you know more about some of these like data center plus plays, and it's amazing that you're in this space on the board of a company.

[30:52] Marty Bent: Yeah. And I I mean that's the thing. It's not right now it's the other thing you have to consider too is these models are progressing so quickly. So you'll build something one week and then a new model will be released. It's like, okay, I have to take this out, put this in and r rejigger things a certain way and there it's a lot of learning on the go. And so to your point about like the early movers and not trying to blow smoke up my own ass, but it it is like an involved process and if you're not staying up to date with the the latest developments in the space, it's it's gonna be hard to keep up, but there will be a point, I think, where the model, the return on the model, sort of the the model progression sort of plateaus.

[32:06] John Tinsman: Yeah. I mean and if demand grows by twenty four X or eighty X, we don't have anywhere near the amount of compute. To i even what we're building today, it's it's nowhere close. So it honestly seems like it could be like an eight or ten year build out to catch up, which means we'd be in the very early, I mean, like in the first like, you know, year or two of a very of like a decade-long build out. So it could be one of the longest like CapEx cycles that's ever been in the history of the world. I mean, it it can really only be compared to when they like needed to build railroads across like the whole world because like they realized that was better than taking stuff on a donkey.

[32:44] Marty Bent: Yeah, well, and I and I think many people are comparing it to the internet and the broadband build out too and to to your earlier comments about people worrying about this being a bubble and it being too too hard to believe and they're pointing at the dot com bubble. I think that's one thing that many people are missing is that the demand for tokens is there. It's like the the dot com bubble, the infrastructure was built out and we didn't have the means to really lean into the digital economy.

[32:49] John Tinsman: Yeah.

[33:11] Marty Bent: That exists today in the nineties and it didn't you weren't able to monetize websites the way you are today. There's a ton of sort of software development kit infrastructure and payments infrastructure and logistics infrastructure that need to be built out and Amazon and Google and many others did that in the late nineties, early two thousands, and then that caught up. But I I think with AI, it's different because the the models they work. If you know how to use them, they work well for you and you want to use them more, but the tokens simply aren't there yet. The the the ability to to ping the models as much as you want isn't isn't there. Whereas in the broadband era the the fiber was there. It was laid, but the the the supply of people bringing businesses that that actually worked on the internet wasn't there.

[33:58] John Tinsman: Yeah. I mean, and I think the thing is it's like it's kind of like the railroad, like Wells Fargo started as a carriage delivery, like a wagon delivery service. And so like they'd go down the Oregon Trail from whatever in New York all the way down to Oregon. And it's like the railroad basically put that out of business overnight and it became a bank. But it's like if you can do something for one one thousandth the cost. You know, and software has always been limited by extremely expensive developers sitting in seats. If you can or like lower things by one one thousandth of a cost, do you really ever go back to the way things were? Do you r is there really any slowing down the adoption? Because it's savings are so big, it only accelerates. And and I think that's I think that's probably where we sit today.

[34:41] Marty Bent: Yeah. What outside of compute infrastructure specifically, again, bringing back to energy, 'cause that's that's one thing I've identified is maybe the the biggest constraint right now is energy infrastructure. I know you talked about natural gas pipelines, but I think we really need to expand generation and great interconnectivity by orders of magnitude in the next decade as well. And I think that's another part of the market that provides an incredible opportunity. And it's it's just good.

[35:07] John Tinsman: Yeah. Well, the data centers aren't so I think if the data centers are bringing their own energy and they barely connect to the grid, then one is that true. And two, and two, like there's people like Bloom Energy making like like obviously they make electric generators, right? so that you can like bring your own power to a to a data center. But like the question to that is is that like That's really expensive to build each d data center. So or b to b build each generator, I mean. So the question is like that's not low marginal cost, right? Like you can make an additional memory chip for not a lot of cost, but it's really expensive to build a big generator.

[37:22] John Tinsman: It was always limited by expensive manpower and now software demand is totally unleashed because AI agents need to need the tools to get the work done. And so I think that could be like really interesting in that in that the demand is actually gonna be higher. Additional sales come at 100% profit margins, and you have you know five X demand or something as the tools just become, you know, because the the most expensive thing to Adobe was never was never the a subscription cost of 300 bucks a year or a thousand. It was the person that was you were paying to do the work in it.

[39:03] John Tinsman: Like when you have those things and you come like sorry, I hit my mic. When you combine those all together, those are the makings of the best trades. And so we'll see what the future holds. You know, anything can happen, but I am excited for it.

[39:16] Marty Bent: Yeah, I think we saw an example of this a couple of weeks ago when Figma announced their Q one earnings and they surprised massively to the upside. I think many people were thinking Figma UI designer tool was gonna get blown out by by people just building it themselves, but they actually implemented some AI functionality and I think to your point what they experienced is people actually using them more because they had these tools to to go in there and and leverage their their software tools.

[39:45] John Tinsman: Yeah, well here's the most interesting thing is a lot of these companies have like a five-10% profit margin. So imagine you're a company like Spotify and you can look at the earnings today and their year over year operating expenses are falling. Meanwhile, they're still growing the revenues at 10%. So What that tells me is that their profit margin will go from 5% to all the all the additional 10% sales go straight to profit. So now their profit margin will be 15%. Well, that's 200% year over year gap earnings per share growth for a company that had a low PE ratio. So it I mean, usually when they have low PE ratios, they shouldn't get any lower.

[40:57] Marty Bent: Yeah. So you're not a doomer of this. You don't think AI leads to massive job disruption that completely throws the economy into a tizzy and destroys the demand for tokens that that existed in the first place? Do you think this is gonna be an accelerant for human productivity and creativity? Are you yeah.

[41:16] John Tinsman: Yeah. I mean it brings in a lot of wealth. Like I just look at what it's doing in Iowa and every it seems like people are doing better than ever. And so I mean I think in the 1900 they were building the Panama Canal and they built they came out with a steam shovel and there was actually a political debate at the time with like the presidents and stuff. Should we bring this steam shovel to Panama Canal because a bunch of people lose their jobs? Or should we just keep paying all these people to die of malaria in Panama and and like keep doing it because we want them to be employed?

[42:47] Marty Bent: Is there anything on this AI topic that we haven't covered yet that's top of mind that you think the audience should be aware of?

[42:54] John Tinsman: Yeah. We we've been we've been whacking all the moles on this one. So I'm trying to think, but you got me in the corner now. I'm trying to make sure I didn't leave anything out. yeah. You can't think of anything, man.

[43:10] Marty Bent: Well, on that note, moving to fertilizer since you since you're close to the the fertilizer fertilizer industry as well, I think there's been a lot of headlines about obviously the the straighter Hermuse messing up the supply chain for for sulfur and helium, which are key inputs to to fertilize. What are you seeing on on that end?

[43:14] John Tinsman: Yeah. Yeah. Yeah, it's not great. so now we're now we're taking a little worse of a step. But I think there's there's three things to keep in mind. I I won't hit on helium, but So you've got nitrogen. The way you make nitrogen is you use a ton of natural gas to superheat the air and cause a catalyst and you take the nitrogen out of the air. So it's like naturally occurring, but you take nitrogen out of the air, the atmosphere is 70% nitrogen, and then you turn it into like ammonia or a UAN, and that's what you use to put on corn and things like that.

[45:16] John Tinsman: You like try catalyzed sulfur into sulfuric acid, and then you use sulfuric acid to melt phosphate ore. So when you're mining phosphate, you melt the ore and that takes that makes a phosphoric acid. Well, phosphoric acid is the base of like phosphate fertilizers. and so Now the sulfur prices are so high that this they can't make phosphate profitably. So they're shutting down. Like Mosaic is a publicly listed company. They're shutting down production in their phosphate plans because they can't get enough sulfuric acid. And so phosphate prices are now going through the roof. and sulfur prices, which is another key ingredient. And so you've got sulfur, phosphate, and nitrogen that are all getting are spiking really high in prices.

[47:12] Marty Bent: Have you experienced an environment like this before? As your family?

[47:16] John Tinsman: the the not me, my it the farm the farm crisis in the eighties was bad and we're not near that. but that's when a lot of people are going bankrupt. But I would say that usually all of our customers use John Deere financing, which is, you know, kind of thought of as being very easy to get, because John Deere will like lend you know, they'll they'll lend and they'll finance. They actually make more money financing, I think, from John Deere than they do from their tractors. Nobody seems to know that. But but This year John Deere is saying, like, you guys are so stretched we won't lend to you.

[48:28] Marty Bent: Yeah. Very interesting times. You got the incredible optimism and in growth potential with AI and then some supply chain disruptions messing up messing up the food supply. but I think if we learn anything from twenty twenty, twenty twenty one specifically is the supply chain disruptions can be fixed in time, maybe

[48:50] John Tinsman: The Yeah, and part of the problem is that core land prices just had gotten so high, right? So people kept buying higher and higher land and you need a higher profit just to cover the cost of the land and then they rent it out for more. And so some of it is just that the asset prices had been so high, kind of like how you could see like, you know, home prices getting too high in a area and then they kinda stall out for a w while and fall and You know, I I think there's some of that going on too, which is that that this was like farming had been so safe, the land had been such a good investment, that it was so reliable that just more momentum, money just kept pouring into it and it overpriced the land. And that exacerbated, you know, a tough time because rents are high and you know, interest that you're paying on overpriced land is high and all that stuff.

[49:38] Marty Bent: Yeah. Maybe the agents can help us figure out a solution to these problems, you know.

[49:43] John Tinsman: That's good. That's good. I hadn't thought about that one yet.

[49:47] Marty Bent: Awesome. Well John, again, thank you for for hopping on and short notice. This was a fascinating, high dense, high highly dense conversation. And I I have a lot of people on here who are very very skeptical of the AI thing, but just using it myself over the last couple of years, but with the agents over the last five months, it's it's become pretty obvious to me that this there's a there there. It's real. If you know how to use it, it can make you a better operator and

[49:51] John Tinsman: Yeah.

[50:17] Marty Bent: It's just going to be there's just gonna be a a a lag period bet before the the broader market wakes up to that and begins implementing this. And if that happens the demand for these tokens is going exponential. We're we're at the cusp of of what we're gonna see in the next decade.

[50:34] John Tinsman: Yeah, well, you've got a great show. You provide enormous value to your viewers through your insights. So I'm just happy to be a part of it, and thanks for having me on.

[50:42] Marty Bent: All right. We'll we'll link to to John's ex account, what he's doing at AOG in the show notes and hopefully we can do this again soon.

[50:51] John Tinsman: Yeah, yeah. so thank you and yeah, check out AOTG ETF if you guys are interested.

[50:56] Marty Bent: Awesome. Peace and love freaks. Take.