3 Stocks You Need To Own For 2027

Hello and welcome to being exponential. So the AI trade continues to evolve and today we're going to be looking at three names positioned at different layers of the buildout. That's going to be Marll, ARM, and Nebius. So Luke, we're keeping this episode a bit simpler today. So let's start with Marll. Yeah. So I think the inspiration for this is in this podcast, right? We talk about a lot of different stocks. I mean, we talk about like a lot of stocks because the AI infrastructure buildout is so expansive and so robust that it does impact a lot of different stocks. It's a rising tide that lifts a bunch of different boats. And there are a lot of different ways to play this multi- trillion dollar buildout, the biggest buildout in the history of the American economy. But in the spirit of simplifying things to really help people understand where we think the best opportunities are, I wanted to take this podcast to go over what I think are my favorite AI stocks, my top AI stocks for the next 12 months. And those are as as you said Marll and Nebius. And the connective tissue between those three stocks is that each are very well positioned to benefit from the biggest shift happening within AI. Right now, which is the shift towards agentic AI. We are shifting from querybased AI and chat bots where you just talk with chat GPT or talk with Claude or talk with Groth towards Meta Muse towards OpenAI dots towards claude co-work and the difference is that these are always on always running aentic AIS that are inferencing constantly. So, example, if I'm working with Claude or Chat GBT and I want to recap the um Nvidia earnings report, I would go to Claude or Chat GBT and type into the little chat box, hey, Nvidia's reported earnings. Uh, please recap the report for me. What did Jensen say on the call, etc., etc. And then it would start, the computing would start for the AI at that moment in time. And then it would end once it is finished answering the question. Then maybe I ask another question and then it would restart and then run until it answers that question. The point being it's start stop start stop start stop with an agent. What I would do with an agent is I would tell an openi dot or metamuse or claude co-work hey constantly monitor Nvidia for all daily news developments that could impact its earnings. When earnings happen, make sure to recap that for me as well. provide for me a daily briefing on Nvidia that gives me an update on what's happening with the company and the stock and Jensen Wong and any insider buying and then I let it go run and what it then does is it runs constantly every day it's fetching stuff every day it's looking for things every day it's completing when I tell it to complete so it's not start stop it's recurring constant and what that does is it creates a lot larger bandwidth and compute requirements for AI infrastructure. And so then who's positioned to benefit strongly from that? Well, one company that is really well positioned is is Marll. You see, Marll basically has two businesses, right? They have their custom silicon business and then they have their um networking business or switches business. Now, on the customs side, uh as agents perform more tasks, the cost of completing each task becomes increasingly consequential. Right? If we're doing all of these things at sufficient scale, um, hyperscalers have an incentive to design accelerators and chips around their workloads, their memory requirements, their power constraints, their software. Uh, Marll helps customers implement those designs. They have wins with Amazon for Tranium. They have wins with Meta. They have wins with Microsoft. They have wins uh, with Alphabet and the TPU program. Marll is very much embedded in that custom silicon world, probably second only to to Broadcom. And some may argue they're actually first above Broadcom in terms of custom silicon. But the point here is that as we get into more inferencing and as we get into more bandwidth requirements and there's just a lot of agentic calls happening, it's going to get pretty expensive to run all that on Nvidia GPUs alone. And so you're going to want to run it on custom silicon. And so there's going to be the shift to custom silicon and Marbell is definitely going to benefit from that. Now the second one is on the um the connectivity side right inference systems they need to move model data and intermediate results among processors uh and memory while applications also access external data and services and Marll supplies the interconnect and switching technologies across those racks and those clusters. They have optical connectivity. They have support for multiple emerging scale up architectures. And basically what that is in simple terms is the highways for the data centers. If we indeed have these agentics, these agents that are just going doo constantly, you're going to have a lot of traffic on the highways between the GPU clusters and data centers. And so you're going to need to build more connectivity. You're going to need to upgrade existing connectivity to go from copper to uh fiber optics. And that is where Marll is a massive massive player. So Marll to me is one of my favorites for the next 12 months because when you talk about the shift Aentic AI, they have two businesses that are both going to benefit massively uh from that shift. And then if you look at the chart, the chart is just absolutely gorgeous. They just updated their their guidance and it was way above estimates and the TAM the long-term TAM was I guess 4x bigger than what they previously expected. Stock is surging higher here. You can see it. It's going absolutely bonkers heading up to 300. 300's kind of that top. That's where that shelf was before. We're going to break right through that. There's a lot of momentum on this one. So, I think this is a stock that is both technically in the short term very [clears throat] attractive and also very well positioned longterm over the next 12 months for some pretty big compounded gains. So, that's my little elevator pitch on on Marll. So, uh question here on Marll. So they they reported 8 billion in revenue roughly for 2026 and then in 2031 they themselves reported a essentially a 10x increase in that that revenue. So how much of that is kind of already priced into the stock and then what what has to happen for that to be considered conservative? >> Yeah see I mean let's look at the numbers right we got you're right 8 billion 8.2 two basically in 2026 fiscal 26 which is calendar 25 more or less uh fiscal 27 increase of 47% fiscal 28 increase of 51% fiscal 29 increase of 46%. So you're looking at 45% 46% plus revenue growth in each of the next three years like that's a lot of growth going from 8 billion to 26.6 6 million on the top line. Gross margins stable around 58%. Ebidon margins going from I mean they were 10% last year all the way up to 44% by fiscal 2029. EPS going from a buck 50 last year to 10.69 by 2029. So this is a major growth story. To your question, how much is priced in? I would argue not a lot of it is priced in. When you look at the valuation on the stock, we're trading at just 48 49 times Ford earnings, 39* 40 ET, huh? 41* 40 EBIT. Those are big multiples. But again, I just showed you 46% revenue growth next year, 51% after that, 45% after that. I just showed you EB margin going from 10% to 44%. I just showed you EPS going from a buck 50 to 12 bucks. So when you have that type of growth, this type of multiple is fairly reasonable and actually pretty cheap. And then you look at the estimates on the stock and they just continue to fly higher. However, 12-month EPS estimates on the stock keep inching higher. So that you know buck 50 to 12 ramp that probably is buck 50 to 1350 1450 those estimates are going to keep moving higher. So everything is moving in the right direction for the stock and there's not a lot to not like about Marvel at this point in time. Yeah, definitely. I mean, even like you said, they the estimates for their TAM were higher than they expected. So, it's it's really hard to argue with cases like this. Uh, so next up, uh, let's let's talk about ARM holdings, >> right? So, ARM to me is the CPU beneficiary, right? That I think that ARM is they are um the AI CPU company, right? A they have their uh technology which they license out to other companies for CPU technology and then they also are launching their new AGI CPU uh which was introduced in March with Meta as a lead partner. Um and they're going to start to really sell that in in 2027 and Rampid in 28 and 29. So they're launching this CPU, their first ever actual in-house CPU, not just licensing out the technology at the exact same time that CPU demand inside a data center should absolutely soar because of agents. So when you have agentic when you have just um query based AI if you will, query based AI is really all about the model. It's really all about how well was the model trained and can it answer the questions you want to answer. When you have agents, it's less about training and more about inferencing. And when you have inferencing, that means the communication between the clusters is super important. And so you need an orchestration layer. And that orchestration layer is CPUs. So CPUs were kind of forgotten about in the A infrastructure build out in the first three years when we were focused on just building better and more capable models. It was all about the GPU. It wasn't about the CPU. But now the CPU is having a massive resurgence because now it's all about servicing the models and e economically and efficiently servicing the models. And in order to economically and efficiently service the models, you need like a hand if the orchestra if you will. You need the CPU to orchestrate all of these clusters in a data center. You need the CPU to direct the traffic amongst the the clusters to show where the data should go. That is what the CPU does and that is why the CPU is becoming a more and more valuable component of the AI data center buildout and will increasingly become so because of the shift to AENTIC AI. And this is all happening at the same time that um our good friends over at ARM are launching their first ever uh AGI CPU. And that's why if you look at the estimates for ARM, you know, sometimes it just pays to launch the right product at the right time. Timing is everything. If you look at the estimates for ARM, the revenue growth trajectory here is expected to meaningfully accelerate. You're looking at 24% growth this year, which is decent, but then you're at 36% next year, 30% the year after that, 64% the year after that, and 57% the year after that. This is a story that unlike a lot of the other, you know, you pull up the financial models for a lot of the other a infrastruure stocks out there and it's decelerating growth into 2030. Like that is the story. You have big growth now and it's supposed to decelerate into the mid- teens or low teens or high single digits over the next 5 years. Not ARM. ARM's trajectory is expected to accelerate. It is an accelerating growth story. And that is really really rare in the AI infrastructure buildout. And it's one of the reasons I really do like the stock. Now the bare thesis here is that because they are launching their own CPU and not just licensing, they're not going to be running at these 97 98% gross margins that they have historically run at. They have to actually build the CPU. So now the cost of goods go up and the gross margins go down. So gross margins will drop from high 90s to mid to low7s. I I get that. But when you're accelerating growth in the manner that they are, you're still ripping off 30 40 50% EPS growth per year. And so this is a company that I think is just really well positioned, right product, right time, really well positioned to win because of the increasing importance of CPUs in the era of Aentic AI. And the chart to me is I mean it's very attractive like Marll's chart. You can see it's really starting to pull up in a pretty big way. Here's ARM. So ARM is, you know, we had the the big rally and then we had the big sell off and then we kind of just started to reinsert the uptrend and now we're retaking all the MAS. You know, obviously held the 200 the whole time. Reto the 50, retook the 100, just turned the 100 into support, bouncing off that, pushing higher. This is a to me looks like a classic technical turnaround. And so much like Marll, I think the short-term technical picture here is very attractive and the long-term 12-month fundamental picture is even more attractive. So, this is also one of my top three AI stocks for the next 12 months. >> Excellent. Now, we talked about Marll and they have a pretty long-term outlook. So, for ARM specifically, they have some pretty stiff competition in AMD and Intel. Do you think this is also a stock that could, you know, outlast or are we are we looking at hyperscalers kind of starting to integrate their own personal CPUs? >> Um, well, I think there's multiple lenses to look at that through one. I think that there's enough demand growth for CPU, so there'd be multiple winners. I like Intel for its CPU exposure, too. I like AMD for its Intel exposure, too. Or uh I like AMD for its CPU exposure, too. I like Intel for its CPU exposure, too. And I like ARM for its CPU exposure, too. So, I think that this growing Pi, much like the GPU Pi, right, like Nvidia 1 and AMD, um so like there the pie is growing so quickly that I think it's going to be okay. There's going to be a lot of people that can still eat. And so I think that that is one lens to look at it through. The other lens to look at is I think ARM has the best technology in the game. That ARM CPU uh technology has been that that core licensing technology has been the backbone for so many different chips out there. Um and Intel doesn't have that. AMD doesn't have that. So I think when it comes to CPUs, I think ARM is technologically superior. And so I think even though the pie is growing and there's enough for everybody to eat, ARM will have the biggest share and will be the biggest beneficiary. Uh, and then they're also the ones that are launching a brand new CPU. So, you get that growth infusion that you're not going to see at at Intel or AMD. So, that's why I like ARM. Better than both of those right now. Though, I am still a fan, as we've talked about on this podcast previously. I am a fan of AMD and I am a fan of Intel. Right. Of course. Okay. And then on to the final stock here. I think one of our one of the channel favorites, Nebius. Like you said the with the agentic AI market we're going to see a lot more handshakes a lot more pinging and and AI cloud neo cloud is probably one of the biggest benefactors right >> yeah I mean the whole thing here is just the increase in compute consump and and compute consumption >> right if agents do become a part of everyday business operations customers will need infrastructure to run those workloads reliably and repeatedly like I said there's querybased AI where it's stop start stop And there's aentic AI where it's recurring and constant. If we are all doing agentic AI, consumers like you and me, our businesses, if everybody's doing aic AI, the compute consumption growth is going to be absolutely enormous. Um and that's where the Neoclouds will benefit because the whole Neocloud bull thesis is we are compute short and so long as we remain compute short they can be the suppliers of marginal additional compute that the market very desperately needs. They can supply that at very high prices and high margins and win in a very very very big way and they can be companies that grow very quickly on the top line and bottom line and have stocks that absolutely soar. That's the whole bull thesis on Neoclouds. Now, Agentic AI, the shift to Agentic AI dramatically reinforces that bold thesis. I mean, in a massive massive way again because the compute consumption is significantly an order of magnitude larger with agents than it is with just queries. And so if if I am right about agents kind of taking over the world, if I am right about agentic traffic far surpassing human traffic on the internet, if I'm right about the compute consumption growth that comes with this, then Neoclouds look really, really good for 2027. And in the space of Neoclouds, there is one that reigns supreme and that is Nebius because the growth profile is absolutely absurd. Again, there is no company that is growing like Nebius. If you pull up the numbers on Nebius, this is a company that is projected to grow revenues at a 100% compounded annual growth rate into 2030. I have never in my life before seen a company that is expected to grow revenues at a 100% compounded annual growth at 100% keer over the next 5 years. Revenues are expected to go from 530 million last year to 50 billion by 2030. That is a massive massive increase 500 million to 50 billion. EBID dot is expected to go from 250 million to 35.5 billion. These estimates are absolutely absurd. The risk in this stock is the estimates are wrong and I think that that risk is meaningfully offset or meaningfully uh decreased eased by the shift to agentic AI. I think agentic AI shift that shift will allow for Nebius to grow at this at these projections and if it does grow at these projections then this stock is the cheapest stock you could possibly imagine. Remember 100% compounded annual growth rate on the top line and it is trading at a whopping 13 times forward. IBITa It's trading at a I mean 13 times what I do. You would expect the growth profile of that stock to be like 5 to 8% on the top line, 10 to 12% on the bottom line. But instead, we're trading at 13 times. And what we have here is 100% compounded on the top line for 5 years and then IBIT going from 250 million to 35 billion. It it doesn't it doesn't it's inongruent. So either the estimates are wrong or the multiple is wrong. And I think with the shifts in the AI market right now, specifically the shift towards Agentic AI, the multiple is wrong. And if the multiple is wrong, this is a stock that could benefit from not just, you know, PE times EPS equals stock price. You can get massive P multiple expansion here with this massive EPS growth and boom, you're looking at enormous stock price growth. So I think that this is I mean it's a $70 billion market cap right now today on this spike but again if we're going to 35 billion in IBIT by 2030 why would this not be a 10x multiple at least on that that's 350 billion right you put a 20x multiple talking about a $700 billion company so very conservative multiples pave the way for very enormous upside in Nebia stock and that is why I very much like this as one of my top plays for the next 12 months. >> Yeah. Excellent. Yeah, I I know you're uh on a trip right now, Luke. I don't have a lot of time, but it great wrap up. You know, great great stocks here. You know, Meta's leading the way with MW, so we're only going to see more of the Agentic AI shift going forward. Um, you know, before I let you go, do you have any uh additional closing thoughts uh before we wrap this episode? >> No, no. I mean, I I just really wanted to simplify things. I think that there's there's a lot going on. There are a lot of stocks that will win. We own a lot of stocks in our model portfolios, and I just wanted to simplify things cuz people they like top picks. say like top one, top two, top three, etc. And so that's what I want to do this week based on everything that I've gathered in the last, you know, several years and where I think the market's going in the next 12 months. I think um there are higher torque stocks out there, higher upside, higher octane stocks in terms of like balancing risk and reward profiles. Um I think Marll, Nebas, and ARM are at the top of of the list in terms of best stocks to buy for the next 12 months in the AI booth. >> Yeah, Nebius is carrying my portfolio personally. So, agreed. Um, but that's it for this episode everybody. Uh, thanks for tuning in. Appreciate you watching. Take care. [music] [music]