Stock Talk Podcast Episode 337

Speaker A: Foreign.

Speaker B: You are listening to Keystone stock Talk show episode 337. I kick off the week with our star cyber security giant Fortinet Inc. Symbol FTNT on the NASDAQ which has jumped over 60% in the last month after reporting better than expected Q1 numbers and increasing its guidance for the full year of 2026. Brett takes a look at Neo Clouds and NEO Cloud companies. Specifically the astonishing gains this year in names like coreweave, Nevius and Irin. What do Neo Clouds actually do? Effectively they are like traditional cloud providers, aws, Microsoft Azure or Google Cloud, but specifically for AI training and inference. The compute used to actually run AI models on a day to day basis or business for a business. Uh, Brett will let you know if these gains in these companies are at all sustainable. All right, let's get to the show. This week we got a quick one. I welcome one half of the killer bees. Brett Hill's on assignment. He will be back. And Brennan is doing something today that we're not going to reveal. It's very special I'm sure in his life. Um, we'll reveal that next week if he is comfortable and if he's not, we'll just reveal it anyway.

Speaker A: That's one way to go about it. But yeah, I, I think I getta be the new guy this week technically. Um.

Speaker B: Oh yes, yeah, but you're the old guy now.

Speaker A: Yeah, exactly.

Speaker B: Don't worry. So do we have a poll question?

Speaker A: Yeah, we do have a poll. No, it's actually um, quite ah, I would say how people really view stocks and investing in general can really shift this. Obviously we're limited to four answers with Google but really what kills a stock idea facets for you. So 70% of reviewers had weak or shrinking margins, 17% said high insider selling, 31% said too much debt and 35% said expense evaluation. So for me it's really that too much debt which would rise above the other ones because it is something which is more structural to the business overall. So if you have let's say uh, a 20 billion dollar debt and you have a billion dollar market cap, even if you are cash flow positive before your dad, you have an EBITDA margin but you don't have gap earnings that is hard to overcome. You can't raise the capital. You're kind of just in a spot where you can't build out of it. Weaker shrinking margins. I'd probably put a second. But it is something that can turn around. Maybe it was a quarter, a year, a period was there commodity Price fluctuations. An expensive valuation. You cannot, you can support an expensive valuation if you continue to grow.

Speaker B: Right.

Speaker A: A good.

Speaker B: We talked about asinine valuation but we didn't really said. We said, you know, expensive valuation. Yeah, it can be. You get into a situation with too much data to catch UH22, you try to pay it off with equity. You got to issue too much equity or, or you just then can no longer uh, no longer issue any more debt. And we see many companies that grow by acquisition do it with that strategy. They issued too much debt over time and then they end up selling off assets that they've purchased to pay down the debt and then your revenues come down, your earnings come down. Uh, it just that situation is, ends up in. We've seen many companies go into what we call a death spiral and that's what you see in those businesses where they uh, really cannot recover from too much debt. So it's something we really do pay attention to and it's something that I would not now if we had astronomical valuations then that and too much. Those are things that are. But, but we said expensive. There are companies that are expensive and are high quality businesses that grow over time. You got to keep doing that. But it can be done. Certainly the, the death or debt spiral that a company can uh, head into really can eliminate every, all the equity. And that's.

Speaker A: And even if it's not like a proper death spiral to where they are going bankrupt, you, you get it a lot of times zombie companies, so ones that all they're doing is they are paying off their interest rate, they need to refinance but you're never going to have any sort of equity return. The debt may be able to consist for effectively in perpetuity until they need to refinance it. And then sometimes they are able to refinance it, sometimes not. And you're never going to have an upside position though in that scenario. And even on expensive valuation you can always even argue what is the correct valuation. Like you know, at least making more of an argument for it.

Speaker B: Yeah, too much debt. There's not.

Speaker A: It's hard.

Speaker B: Too much of an argument at that point. Yeah. Okay, let's uh, get right into the show. We're going to get into our star of the week. I am going to share my screen Fortinet. It's been in a company that's been in coverage since May of 20 uh 18 in our US large cap and dividend stock research. Uh then it traded at $12.24 today it trades at over a hundred and forty five dollars. It's over a hundred billion dollar market cap one hundred and seven to be exact. Um, this is, you know it's been a tremendous gain over 1100% over that time. Ah, they are leading provider of network security and unified threat management solutions that offer broad integrated and high performance protection against a wide range of security security threats while simplifying the IT security infrastructure for its customers. The company has a flagship family of uh, flagship family of 40 gate UTM appliances provides a variety of network security functionality including firewall virtual private networks or VPNs antivirus intrusion prevention systems, web filtering, anti spam on a wide area network extension acceleration. Uh these solutions are built around Fortinets proprietary Forta ASIC processing chips and 40 OS operating system which are key differentiators and kind of the secret sauce for the company versus its other hardware or software solutions based competitors. Um, by combining multiple security and network functionality into a single application or appliance. Sorry, uh, based on its proprietary platform fortnet is able to provide exceptional performance at ah, reduced costs versus traditional best of breed solutions. So let's look at the company's share price in the last month Like I said, up uh, now 69%. Uh, driving the growth was the 5-6-2026 this year's quarterly numbers for Q1 2026 and they raised their guidance for the next quarter and the full year. Let's look at those Q1 numbers. Strong revenue growth up 20% which beat expectations. Product revenue UH grew 41% billings were up 31%. UH GAAP earnings per share grew 29% non GAAP was up 41% a record free cash flow. The company does kick off very strong free cash flow. It's one of the reasons we were drawn to this business. I'll give you some inside baseball on this company. Uh about eight years ago when we were looking at the cybersecurity industry, uh we wanted some participation in that segment for our clients. We looked in the Canadian market initially found about five names. None of them even met our profitability criteria. Uh we found about 65 names in the US market. Fortinet was our pick of the litter and it has done uh very well over that period of time. In terms of the Q1 results uh we were impressed with the billings to revenue growth of 31%. Uh the consensus was 20%. Uh and you can see here it outperformed off a number of metrics. We look at billings. Revenue outperformance was the most since Q4 2023 and Q1, 2021 respectively, Billings and revenue outperformance. So uh, great to see the valuations or the company has increased its guidance though. However we can see here uh these, this is the increase now that we see for the next quarter. Strong increases across the board. The valuations are what I really wanted to get to. Uh, these are premium numbers relative to historical. As we can see here. Non GAAP right now is in the range of around 47. Uh, that's based on 2026 expectations. GAAP is around 54 and the price to free cash flow which uh, it could be had earlier this year in the about the 2020 range. Um, it is now 40 and a half. So ah, significant premium that you're paying for the stock right now. My conclusion here is AI looks to be accelerating tailwind for the business and expanding that attack surface and increasing the performance requirements. Uh, this is positive. We also like the positive views the company uh, had versus its current demand. Ah, as the. It's kind of different from the environment Covid where there was a one time gain. Here it looks to be tailwinds that are more sustainable as AI data centers are built out and customers deploy AI internally. Um, again we're seeing that as more of a sustainable trend versus the jump that this company saw in demand uh, during the COVID period. All of this makes Fortinet our star of the week. Clients will be updated shortly with our current recommendation and here are the disclosures. That was my quick segment on fortnet. Um, again very, very strong results from the company and it's really the forward guidance that was the most positive that we've seen um, for at least 18 months on the business. So they are seeing tailwinds from uh, from AI. There had been fears that they were going to be seeing headwinds from AI. So we may start seeing this in some of the businesses that uh, have taken a hit perhaps in the software segment. Uh, are we going to see some tailwinds, uh, picking up from demand in AI data centers that are helping these businesses versus headwinds which were expected. Now some companies are certainly more vulnerable than others but uh, Fortinet in this case the market was absolutely heartened by those results and this stock price jump, it's like almost 70% in a month. That is a tremendous re rating and repricing of this business following those numbers and the guidance from management that they're seeing. Again I keep saying this tailwinds rather than the headwinds from AI.

Speaker A: Yeah, no, and it's kind of the opposite of what we saw in Q2 last year for them is um, like summer. You can see it on even the stock price graph you showed. There was, it was something along the commentary of uh, this is the end of the sales cycle for this period. Now then, now we'll have to wait. But the market took that very, very naively. The market was expecting. Okay, we're still going to continue I think for about another six months a year before it starts to slow down again. Now you have the reverse.

Speaker B: Yeah, I mean they were, the market, it was pricing it like it was uh, you know, three to four year slowdown or you know like a significant slowdown. And this is a complete reversal of that. And you can see some suddenly the share price recovered and then surged higher. Um, it wasn't expected clearly, uh, or else the price would have adjusted already. Uh, but now you've got a scenario where valuations are certainly like we look at historical valuations and even when this company was getting premiums, I mean this is getting to your top level of that premium range. So uh, we'll adjust our rating going forward.

Speaker A: No, and uh, just with the poll, that was exactly what we're talking about. It's an expensive valuation but it's a good company and that's that trade off you need to analyze.

Speaker B: Yeah, for sure. Is it uh, too expensive relative to historical or is it more in line? Are you a hold, are you a sell? Are you buying more? Because you buy into the growth going forward. And that's what we debate internally and then come up with that recommendation to our clients. All right, let's move forward. And uh, you're going to talk about Neo clouds, correct?

Speaker A: Right in the same space here we're talking AI, uh this entire show and I think about a tire year, maybe decade now at this point it seems like at times neoclouds, the builders of modern AI which have had just very strong returns since the AI boom. Core wave CRVWV up 212% since its IPO in March last year. So we'll use that just our time period. NEB NBIs have a whopping uh, 1100% iron iren, 980% in the same time period. So with a surging industry or business structure the questions are is this sustainable? Can it continue to grow? Can it even be profitable or cash flowing or is it just sales on the surface and nothing more? So what do NEO clouds actually do? Effectively they are like traditional cloud providers, AWS, Microsoft Azure, Google Cloud, but specifically for AI training and inference. They use a lot of GPUs. The compute used to actually run AI models on a day to day basis as well. To build AI, uh, you need massive amounts of GPU which is why Nvidia, now the world's largest company and NeoClouds are renting them out effectively you need to raise capital, equity or debt, buy up uh, likely tens of thousands of GPUs and either build buy or co locate the GPUs and data centers and then rent them out whether it's by an hourly basis. You can google renting a GPU now which really wasn't a thing even a few years ago. You could do it for crypto mining but really nothing else besides a few edge cases was not a big business. Or you can have a contract term committing to purchase X amount of capacity for Y period normally at a discount compared to the on demand rate. The key for these businesses is a high utilization so it takes it makes sense for them to sell you effectively their capacity if they can lock it in. But the playbook is extremely similar to crypto mining. Specifically I'm going to use Ethereum mining as it does use GPUs or did use GPUs pre merge when they did shift to staking. I'm not going to get into that effectively just don't use GPUs anymore. But prior to that very very similar. You have very high fixed capital input, the GPUs and surrounding servers and databases, the variable electrical cost and any maintenance with the variable selling price which is extremely variable. I want to put emphasis on that to market conditions. So whether the GPU enterprise for Neo clouds or the price of Ethereum for the prior crypto miners it that output is that variation. So it's a very like to like business structure despite seemingly very different industries. But it's no surprise given that that's using the same hardware reality that the prior crypto miners like Iron or core Weaver, they were crypto miners shifted to AI Neo clouds because it's operationally very very similar. Even if you're not thinking AI crypto very different obviously are as an end product but what's running them. It was GPUs for crypto previously for Ethereum mining and now it's AI models are getting ran by GPUs. In both cases you are selling compute and in many cases you don't even need to own a data center you can co locate with the current prices you'll be looking at a uh, really 12 to 24 month payback period and that's payback period for the GPU. So. So that's before your additional cost on a per GPU basis at 100% utilization. It depends obviously on market price, how much you paid for the gpu, the specific GPU and who you're selling to and all that. But it is still you're looking 12 to 24 months. But the price you receive at the market rate will fluctuate during that time, during shortages. So you'll see, oh, GPUs are going up to $10 per for an H100 per hour. But then you can find it's now $2 or $1 month or a few weeks later. This is to me very, very deja vu to Ethereum miners. You see Ethereum go way up. Your price. You would receive similar to COMPUTE for neo. Clouds go way up, supply gets met, or in Ethereum, the price goes down. Then you go and see the reverse. You see it drop right back down. But the industry for now, it's been surging. You can see that in revenues and backlog. Looking at Nebias for example, the stock's up 18% keep just today increasing by $10 billion in market cap. After Jensen Wong, the CEO of Nvidia, commented that Computex effectively used to be about computers. Now it's about AI that they are working with Nebius and making comments around that, making the market cap hit about 69 billion. Revenues are up 684% over the past year to 399 million. Massive. But you have even on an adjusted basis a loss of 100 million. You may say, oh, there's positive cash flow though, 2.3 billion in operating cash flow. But there's capex of 2.5 billion. So negative free cash flow of about 215 million for the quarter. As they are building out, debt is starting to begin to mount. It's in a net debt position of 197 million, including leases. But even that, that doesn't seem that bad. But that doesn't tell the full story as the company has $4.8 billion in deferred revenues which are effectively receiving the cash up front to build that compute from your hyperscalers, which adds to cash, but you still have that liability, that unearned revenue which you have to meet the services. So M. Nevia still needs to provide the services down the line and that's why it's a liability. And so even though that cash value right now is really inflated because of that, your services haven't, um, met, that's not even A bad thing but it is a quirk of the financials. It does make it look better than what it is on top of. When we're looking at capital funding to fund this massive capex build out you have activity raises like $2 billion in pre funded wardens sold to Nvidia to buy Nvidia GPUs of course so you can see a bit of that uh, circular finance which is constantly talked about. But the backlog, what the future growth expectations are really pinned on is at a massive top. Of course it's highlighted at the top of the top of the financial letter. Contracted is over 4 gigawatts that's up 33% from February's end. Ferry four times last August is 1 gigawatt contract signing just a as well they cite a growing pipeline of three and a half times just quarter over quarter. So you have this massive interest in the space, massive contracts. So and I want to highlight this, this is just Nebius but when you look through similar the other NEO clouds you see similar financials, similar deal structures, similar Nvidia deals, similar high capex and ah right now pretty good operating cash flow. But it's once you look a bit deeper, once you see that it starts to become a bit of concern. While the debt does concern me, the bigger issue is this dislocation of capital costs and income. Normally you see this being reconciled in accounting via depreciation amortization which of course when you're looking at accounting earnings if you see that in NEO clouds right now they're at a loss. But the issue with depreciation and this is in general past NEO clouds it's, it's an estimate of useful life and with all estimates that life can be justifiably changed or not, it, it is likely going to be wrong compared to the real economic reality of the company. At year end though this and this is something we've seen throughout the uh industry. Nebby has changed the expected lifespan of GPU from four to five years effectively adding a year of expected lifetime which is year of expected revenue generation and perhaps more key lowering the depreciation cost from 25% of the GPUs uh capital expenditure per year to 20%. Iron has five years, Core Reef has six. And we've seen as well, like I was saying, hyperscalers over the past few years really shift longer which does lower the appearance of expenses but it doesn't really change the economic reality. Just the graph I have up there you have, you can see it, your cost would be shifted out but in reality it probably more would closely follow that uh, red dash line where you see a high initial cost but then the GPUs rapidly depreciate and it kind of settles out over time. That would I would say be a more illustrative reality of the company. But it's something which is very hard to judge and that's why it's a lot of times you'll see companies just say look at our ebitda. But that's not something you should do. You should not ignore the depreciation. In fact it's a really key input which is when I see these massive EBITDA margins which are excluding the really the two key expenses for these companies depreciation. So your GPU costs as well as interest to fund the massive capex. That tail value of the GPU I have up here is in my simplified graph is the profit which you are going to be making from an individual GPU is going to rely on the tail value of the gpu. So when you're extending that lifetime of the GPU or estimated lifetime that increases makes the business make sense. But I know this is going to be shocking to some people. Technology progresses. Nvidia will release new GPUs Google will uh, have iterations of its TPUs. Sarah Brass will launch updates of its massive wafer scale engine. So in reality the price of both rental rate of used or new market of GPU will normally decline quite rapidly after each iteration or competitor enters. Lots of times you'll see more of a step function even which I getting to probably uh, niche in this, this case when you are looking at GPU price once a new one comes out the old ones might get a haircut of 30% versus this more uh linear curve graph I have up. But the point being so whether it's a four or six year reality can vastly differ. Neocloud really they are partially a bet on hardware stagnation. The longer the same tech is in place, the longer you can still run your H100 the better they are. And right now we're really in the early innings of this play. The rental prices are quite high. They are at times they're going up quite rapidly and therefore the expected lifespan are rising. But this can change very very quick and you've seen and you can look at the prior crypto obviously much more of a niche market at that time. But it was still Nvidia GPUs and on a dime. Suddenly GPU prices are cut in half. They're cut to a Third of what they were for the inflated prices during crypto booms. And you could see the same thing here if we see a massive overbuild of NEO clouds or GPU data centers in general. Another facet of concern here, which I'm sure this is an image that many, many people have seen at this point is the uh really circular financing the concentrated customer base. So whether it's nebbyous with Meta or Coreweave with Microsoft, NeoCloud's really have a reliance on the key players who have historically built and ran their own data centers. But with AI needing a massive amount of compute to ramp in such a short period of time they looked elsewhere which caused NEO clouds to boom. Remember these were not that long ago really crypto miners or private companies. So once compute demand really becomes saturated which it uh will at a point that question is when more than anything the hyperscalers will likely stop funding the NEO clouds or look for better terms. Because why was it wouldn't you. There's a commodity product at that point and that commodity is just compute power. So is this like the 2000s the dot com boom of Dart Fiber? Massive build out on the assumptions of continuously rising demand but you need the long term growth to follow through for the economics to make sense. This seems very very familiar to me and even if you look at valuations at this time obviously trailing basis and we'll get obviously massive caveat there but price to say all valuations are obviously absurd. Nevia 79 times I run 31 times and that was my core week we've looked cheap at only 11 times. Still all those for a uh hardware heavy capex company are absurd. So let's look at price to cash flows as they they they're more reasonable. Neus 244 times iron 24.4 times iron at 60 times core weave at 11.5 times but wait let's add the capex in because this is missing the major cost of GPUs. So if you look at free cash flow you can't since it's negative and this is still including the cash up frontage which I stated before for some of the contracts which NeoClouds are receiving while the expenses are whether it's a non cash depreciation expense or real cash cost electricity costs will occur down the line. So even cash flows which are inflated once accounting for the cash outflow for growth is negative. That's not great financials at this time. You can obviously argue this is going to grow out the uh, GPU Costs are going to go way higher. But that's not the structure I am seeing here. Given that expected advancement of any technology, you aren't sitting on the same GPU for 20 years. You need to make the money now, not 5, 10 years from now. So overall neoclouds are really to me an opportunistic business. They are filling in this gap where hyperscalers couldn't or won't or don't want to be able to scale as aggressively as what they needed for this last few years during this AI training boom. They're now in an inference boom. And this entire business model to me is really deja vu to Ethereum miners which every downturn during the crypto market, I don't think AI is crypto for the record, not, not trying to say that, but very similar business models for the Neo cloud. It's many of them did go out of business every single time there's a downturn every time that variable cost you are or price you are selling at goes down far enough. If you're loaded up on debt, you're reliant on capital markets. That is that issue with the financial model. And you'll have many people quote, oh their ebitda, uh, margins are massive but it's not including the main cost for them at this time. And then you have the timing features to cash flow, inflating operating cash flow but negative free cash flow. The current financials are not great. So when you're looking at the surface they may seem better than what they are obviously past looking. And people will say that growth is real as far as the backlogs go, but it's concentrated and some of that is prepaid. You've already received really the positive benefit and now you have to deal with the negative down the line. And this is all at current really massive valuations. And even if you just ignore uh, what you're dividing by and just look at the market caps of those three companies. $162 billion. You need long term growth to support that industry. And this is just three, there's many, many more. And it doesn't actually have the highest barrier to entries once you reach a certain scale, obviously an individual cast our Neo cloud, but it's not something unique. You are selling a commodity and it's a commodity that can really be placed anywhere in the world. It's not a resource where you're limited to this location. This finite is something that can be built out once enough GPUs are produced. And lastly to me this really seems to echo that dark fiber of the dot com boom, the assumptions of massive continuous growth, which if you are assuming massive continuous growth, that's a red flag. It's not a buying signal. And just to close off, we could be referencing dark clouds in the next 20, 25 years over dark fiber.

Speaker B: Yeah, I mean it's interesting like the thought of the high EBITDA margin is uh, I mean particularly in this segment where hardware like it never stands still and you continually have advancing and then pushing out the length that this hardware can. The life cycle of the hardware. Five, four, five, six years out now. Um, yeah, I mean it does raise red flags to me.

Speaker A: And just for reference on that, hyperscalers before AI were looking more at three to four years.

Speaker B: Yeah, yeah.

Speaker A: So we're already seeing it expand nearly double in some cases.

Speaker B: I think that puts it in perspective. Right. So like if you're, if you're talking about going from maybe three to four to six, I mean it makes the financials look better. Is it realistic? I mean like you said, the demand has boomed. There are massive competitors there. The hyperscalers I would think would move into this, um, once things settle. And I mean, and they have the capital to do it. And so, um, yeah. Is this an opportunistic window of you know, three to five years or something? Or maybe it's a little longer and then, you know, and the, the commoditization of this just renting that space.

Speaker A: Yeah.

Speaker B: And to me, like the price to sales that you're seeing right now, obviously.

Speaker A: Yeah, it just doesn't click. And even then you saw with. I'm, um, going to compare to crypto again because I think it's just such an apt comparison is for Bitcoin and you're going back about 15 years now is there was a shift to ASICs, so applicating specific integrated circuits which effectively they replace GPUs. And you're starting to see like Google TPUs, these switches in technologies. Sarah Brass does effectively big wafers. And you're seeing these shifts. So if there is a proper shift to whether a different set of hardware, those GPUs can just suddenly fall all off a cliff. I'm not saying that will. I personally don't even expect that to. But you do have these very real risks which we've seen historically within the same space that can realistically occur.

Speaker B: It does seem like history may be repeating itself, at least to some degree. So. And that's what we'd be wary of. Um, are these companies really worth 800 more than they were, you know, six, eight months ago. Um, yeah, I'm. I'm uncertain, I'm not very certain on that. And, and I do think there is like there is some bubbles forming and it gets us.

Speaker A: Yeah, especially when you're seeing like that circular financing graph I add up with the, the Nvidia pre funded warrants and you're seeing that throughout the industry really especially with Nvidia, they effectively funding purchase of their own hardware. I don't even think Nvidia is a bad company. It's just the reality of the space. And I always gotta add a million caveats for these more negative ones because

Speaker B: some of them say you're an idiot. Nvidia is the best company.

Speaker A: Yeah, exactly.

Speaker B: And just say it wasn't a good.

Speaker A: We don't have any positions. I didn't have a disclosure site. No positions, no shorts, nothing.

Speaker B: No on those companies. Okay, well that'll close out our show for this week. Next week we'll have a little more regular show I believe. And um, as always, ah, keep your questions coming into our your stock rtake segment. We'll endeavor to answer those on a weekly basis. Uh, if you're watching this right now on YouTube, smash that subscribe button. If you are listening to this wherever you consume your podcast rate and review us. And as always, I'd like to wish you profitable investing. Thank you.

Speaker A: Thank you.

Speaker B: Mhm.

Speaker A: Sam.