99% of Investors Miss The AI Backbone (My Full Map)

$7 trillion is about to get poured into one specific area, and most investors aren't even looking at the whole picture. McKinsey states that data center infrastructure will grow between 14 and 23 percent a year over the next five years. This is the biggest physical build-out since probably the railroad was booming, and it's happening behind all the headlines. Of course, everyone is watching Nvidia's stock price, but almost nobody pays attention to the buildings that are going up behind it, and that's where the real story begins. People are asking where all of these Nvidia GPUs are actually going, and the answer isn't a mystery. They're all being packed into these massive data centers that are built to run AI models around the clock. But most people never see that part of the industry, so they have no idea why the build-out is exploding. Even Bill Gates said that AI is the biggest technological shift of his lifetime. And I've seen this firsthand, because I used to work on machine learning models for Amazon when I worked on their pricing models, and that was long before this was ever mainstream. My point is that AI is not speculative, but I do believe that it is completely misunderstood. It's already replacing entire chunks of work, and it's coordinating fleets of robots in logistics and manufacturing. Data centers aren't these little server rooms. These buildings stretch the size of football fields, and some use more electricity in a year than the entire state of Alaska. And inside, you've got hundreds of millions, possibly even billions of dollars worth of hardware that has to stay powered, cooled, and connected every second of the day, or the whole system just falls apart. Goldman Sachs expects data center power demand to jump 165% by 2030. And if power demand grows that fast, something has to build and fuel the grid behind it. That's why I covered nuclear stocks a few weeks ago, and of course several times over the past two years. Because AI doesn't run without massive, stable electricity. Now here's where that $7 trillion actually goes, and more importantly, who gets paid from it. And I'm going to state right now that I had created a massive spreadsheet of every public company that I could possibly find that fits within this space. And of course, I will have a link to all of those down in the description. But for this video, I'm going to be breaking out where the money is flowing, several companies that are tied to each of those areas, and how I'm going to be investing broadly into each tier. But first, here's why the buildout is hitting overdrive right now. People think that AI lives in the cloud, but every single prompt runs on physical hardware that's within these facilities. Elon's Colossus cluster in Memphis uses around 100,000 NVIDIA H100 GPUs for training. Meta is building the same kind of scale for LLAMA. At this point, these facilities need their own power substations just to stay online. NVIDIA's newest AI chips pull up to three times more power than the last generation, and all that energy turns into heat real fast. That's why these facilities need industrial-grade liquid cooling systems, dedicated power infrastructure, backup generators, battery banks, and of course, networking. Goldman Sachs says that the U.S. data center construction has tripled in the last three years, and it's still accelerating. Once again, AI isn't just about cloud-based software. It includes steel, concrete, electricity, and cooling, and that's where the real money is flowing right now. Now, before I jump into the breakout, if you're getting any value from my content and my spreadsheets and my free newsletter with my portfolio, then please consider pressing the like button so my content can continue to grow. McKinsey's report breaks that $7 trillion into two main buckets. 60% goes to compute, the servers, GPUs, chips, memory, and of course storage, basically the machines that are doing all the work. 40% goes to the facilities, the buildings, the power systems, the cooling, the real estate. It's the shell that keeps the whole thing alive. That's roughly about $4.2 trillion flowing into hardware and cloud platforms, and another $2.8 trillion into power, cooling, and the physical footprint that's behind it. Here's why this split matters for us investors. The two layers behave nothing alike. They have different growth, they have different risk, and they have different winners. The compute layer is the high growth side. AI chips, cloud platforms, server manufacturers. These are companies that can grow 20, 30, and even 50% a year because AI demand isn't slowing down, but it is a faster game. Technology moves quickly, and one new chip design can reshuffle the entire market share simply overnight. The facilities layer, then, is steadier. Power systems, cooling equipment, data center real estate. Think more like 10 to 15% annual growth. It's not flashy, but it is essential. And this side wins no matter which chip company comes out on top. Think of it this way. The compute layer is betting on who's going to win the AI race. But the facilities layer, well, they're betting that the race happens even at all. And someone, of course, has to build the track. That now leads us to our sponsor, Fundstrat, which was founded by Tom Lee, a trusted Wall Street voice and former chief equity strategist at JP Morgan that I often refer to in my videos. Tom Lee's FS Insight by Fundstrat is dedicated to democratizing Wall Street research. And their goal is simple, empower self-directed investors like us with the same evidence-based research that Wall Street uses to navigate the market, where I know that I personally look forward to their emails every day. We know Tom for his signature evidence-based research and famous calls like the V-shaped recovery after COVID. When you join, the multidisciplinary team sends you real-time market alerts called Flash Insights, daily video updates and actionable research across equities and crypto. This gives you the clarity to make better, more timely decisions and confidently control your portfolio. If you want access to the same research that banks and hedge funds use, now more than ever is the time to invest in knowledge. And right now, this is FS Insight's biggest sale of the year. You can choose from macro, crypto, or even their pro package. If you'd like to learn more, please feel free to check it out down in the description below. McKinsey says that 60% of the money is going to be going into compute and 40% into the facilities, but that doesn't necessarily mean that we have to invest 60-40. If you're like me and your goal is maximum upside, then I would lean in heavier into the compute, because that's where the fastest growth lives today. That and it has long-term demand, meaning once a data center is built, the construction companies don't really make any more money from it. When I look at the data, McKinsey's 60-40 split makes a lot of sense from a high level. Compute on one side, facilities on the other. But of course, there's a third group that's sitting above both layers. And they break every rule within this model. And of course, these are the hyperscalers. Think Amazon, Microsoft, and Google. These companies don't just buy data center capacity, they're actually building it. Entire regions at a time. They're negotiating multi-gigawatt power deals before the rest of the market even knows that demand is coming. In fact, they're even designing their own chips. Amazon's Tranium and Graviton, Google's TPUs, Microsoft's Maya. They're engineering silicon specifically for their own AI workloads, so they're not solely dependent on anyone else. And even now, they're inventing their own cooling system. Just last week, Amazon rolled out its own in-house liquid cooling system because the traditional suppliers are backlogged for years because they weren't willing to wait. They built their own system so they could deploy their high-density GPU racks right now, not in 2027. They're also running the clouds where every enterprise AI workload lands today. This is AWS, Azure, Google Cloud. And of course, they're not alone. Oracle is accelerating with their AI HPC leasing. Alibaba and Tencent run massive AI regions across Asia. And IBM is carving out their own niche within the regulated industry. But once again, those big three still sit in a category all their own. They're the only players that are touching every layer of the stack. And that's why hyperscalers, in my mind, get their own bucket. Let's go ahead and jump into the compute layer because this is where most of the growth is happening. And everything begins with the chips, where NVIDIA is still way out in front of everybody else. But AMD has real momentum with their MI300. And Broadcom and Marvell are showing up inside almost every major AI system. And Intel is still there, pushing hard to get back into the conversation. And the thing that most people overlook is memory. GPU demand is huge, but memory demand is exploding right alongside of it. This translates into Micron, Samsung and SK Hynix, which are all sold out in high bandwidth memory for years out. Once again, these systems can't run without massive amounts of memory. Then we've got the companies that are turning all that silicon into actual racks. Supermicro has been scaling almost faster than anyone else in this area. Then you have Dell and HPE, which anchor the enterprise market. Then there's Lovo, which is huge across Asia and has a big share of global server shipments. And it's not just the big companies anymore. There's a growing group of GPU cloud providers trying to keep up with all that demand. Think Applied Digital, Akamai, DigitalOcean, Iris Energy. They're all building out dedicated AI compute as fast as they can get hardware delivered. And then they're leasing it out to those big players also as quick as they can. And of course, behind all of this is the semiconductor supply chain. Taiwan Semiconductor manufactures almost every advanced AI chip that's out there. ASML is the choke point for the tools that everyone needs in creating these chips. Lam Research, KLA and Applied Materials, they handle the rest of the equipment that makes these high end chips even possible. Then once that hardware hits the racks, the networking becomes critical. Arista leads the cloud scale switching. Cisco drives a lot of that enterprise traffic. And companies like Sienna, Lumentum and Coherent move data across long distances between buildings, regions and entire countries. This whole layer is moving extremely fast. It's where most of the revenue growth is happening today, and it's the part of the stack that investors look at when they're aiming for a lot of that upside. Now, let's go ahead and move into the facilities layer. This is the part of the system that you never really see, but nothing works without it. We'll start with power and cooling because that's where most of the physical build out is happening, where Vertiv is tied directly to the rise in AI data centers. Eaton and Schneider Electric handle the electrical distribution and the switch gear. Then Johnson Controls, Trane and Daikin manage the thermal side. Then Modine and Invent are growing really fast, too, as more racks shift to high density cooling. Then, of course, you have the grid itself. AI is pushing power demand higher than the grid was ever designed for. So companies like Siemens, ABB and Quanta Services are all seeing real tailwinds, where you have Bloom Energy and Cummins helping with onsite generation and backup power when facilities need more stability than the grid could ever provide. From there, it's the companies that actually own the buildings. Equinix, Digital Realty and Iron Mountain build and lease these types of shells. They provide the space, the interconnects and the reliability that lets everyone else plug in and scale. And then you've got fiber and optical side. The long haul links between all these data centers. Infineera is a company that handles long distance optical systems. Then you have Fujitsu and ZTE, which are major suppliers in Asia. And these companies move the data between campuses, regions and entire countries. Finally, you have the software layer that keeps the whole environment stable. Think VMware, IBM, Nutanix and ServiceNow. They handle the orchestration, virtualization and the automation. The stuff that keeps workloads balanced and the hardware running efficiently. But unfortunately, this layer doesn't move quite as fast as Compute. But it does scale with every new facility, every new rack and every new watt that gets pulled onto the grid. And the best part is, it's steady. And it benefits no matter which chipset or cloud platform is winning. So now that we've laid out all the layers, the hyperscalers, the Compute names and the facilities, let's talk about how I actually invest in this. Because knowing the players is only step one. It isn't the same as knowing where to put the money to work the hardest. But before I break anything out, here's the simple truth. Not every part of this ecosystem grows at the same speed. Like I keep saying, Compute keeps moving the fastest. Facilities move a little bit slower, but they are very consistent. And hyperscalers sit right in the middle. They're big, they're steady and they're essential. For me, the goal is never just to own everything equally. My goal is to match the growth, the risk and the timing with what the data is telling us today. And I think that most of us can agree that the data is fairly clear. The money that's flowing into AI and data centers, it is not being split out evenly. Most of the upside is landing in Compute. And most of the stability is coming from facilities. And the hyperscalers, well, they capture pieces on both sides without a lot of volatility. And once again, I want to share that I have a massive spreadsheet down in the description with over 100 companies that I happen to be tracking in this particular space. Where I promise that I'm going to continue to drill down and find the undervalued and the highest return opportunities over time. Today, I've probably mentioned a lot of stocks that you're unfamiliar with. And I'll dive into those in future videos. But for now, here's how I'd put $100 to work across the data center stack today. $45 goes straight into Compute. This is the fastest growing part of data centers. Think chips, memory, servers, networking. And the demand is still running really hot. $30 then would go into hyperscalers. Amazon, Microsoft, Google. They're building the data centers. They're filling them. And they're running the cloud platforms that sit on top. And they'll generate revenue for the long term. And my last $25 would go into facilities. Power systems, cooling, electrical gear, real estate. These companies get paid every time a new data center flips the lights on, regardless of what chips are inside. Now, what does that $100 look like in five years? Using what I think is realistic growth rates for each of these groups, Compute would be growing in the low 20% range. Hyperscalers around 12% to 13%. And facilities around 8% to 10%. That simple $100 would then grow into roughly $200 to $235. There's really no guessing. There's no moonshots. It's just clean exposure across the parts of the data centers that are doing the real work. Now, I always try to give some added information to those of you that prefer ETFs. And two that align to data centers the most are the Global X Data Center and Digital Infrastructure ETF with the symbol DTCR and the iShares US Digital Infrastructure and Real Estate ETF, IDGT. Now, I have to admit that I have not dug into these much at all, but they do cover a fair amount of these basics. Honestly, I wish that I had the ability to make my own ETFs as it'd make them so much more efficient from what I see in the market. Some of the companies in these ETFs really make no sense to me. But hey, I digress. In summary, by 2030, analysts expect global data center power capacity to jump from 81 gigawatts to 222 gigawatts because AI needs far more horsepower than the grid was ever built to deliver. And that's the real investment story. The companies building the compute engines and the companies building the backbone that's right behind them. They're all stepping into a $7 trillion wave that's already in motion. In my opinion, the money isn't in the hype. It's in the hardware, the power, and the concrete that makes AI even possible. And that's where the next five years of returns gets decided. As always, thank you so much for watching.