Critical Minerals Explained: What Powers AI | AI Fundamentals

So in the last video in this mini-series about the fundamentals of AI, we asked a pretty simple question. What is the input to AI, and the primary input to AI, or going one layer upstream, is power? And so we talked through how AI is reshaping the global energy or power markets. And today we're going one layer deeper in that stack and asking the question of, well, what are the inputs to energy or to power? So that's going to be this video. We're going to hear a word from the sponsor of today's video, and then walk through at a high level what those inputs to power are, and how AI is reshaping that market, so you can be more informed about the entire full stack of AI from the model and the consumer application layer all the way through what we're talking about today. That's going to be the video. It's going to be a fun one. Let's go. So we're going to talk about critical minerals, and one of those matters more to our everyday lives than almost any other, and that's lithium. Lithium sits inside every EV, every grid battery, a whole list of systems like this, and even things like your phone and your computer. And right now, the US produces less than 1% of global lithium supply. And because most of the refining happens here, for Washington, lithium is a matter of national security. And for that reason, in October of 2025, the DOE took an equity stake in Thacker Pass, and that's the big lithium project in northern Nevada. Now that brings me to the sponsor of this video, Surge Battery Metals, ticker NILI, on the TSX Venture and NILIF on the OTCQX here in the US. Now I mentioned Nevada, and Surge runs the Nevada North Lithium Project in Elko County, Nevada. That is that same northern Nevada battery belt as Thacker Pass. And just this week, Surge dropped some big news. And before a mine gets built, it goes through the series of studies. And on October 2nd, Surge released its second one for this location, the Pre-Feasibility Study. And the study puts the project's after-tax value at $9.8 billion. That's up from $9.2 billion after the last study. So build and operating costs both came down in this study, and it pays back its build costs in about four years. That is very good for context. And so that's important. And part of the lithium in that location now counts as official reserves. And that's the industry term for lithium that engineers have shown can be mined for a profit. And there's enough there for 42 years. So Surge owns 67.5% of this project that's valued at just under $10 billion. Well, the whole company trades at a market cap of around $250 million, and they have $75 million in cash. Now a study value is not the same thing as a stock price. These guys would still need about $2.8 billion to build the first phase. But this is one of the highest-grade clay deposits in the entire United States, with about 10.5 million tons of lithium carbonate equivalent. Test work has turned that clay into battery-grade lithium carbonate. The partner who owns the other one-third of the mine is a major player in this space. And the chairman founded a lithium company which sold for $490 million previously. So next up for Surge is the final, most detailed study. Now of course, this is still a pre-production company. Permits and financing take time, lithium prices swing. So these are study estimates, so always do your own research and diligence. And that said, if you want to dig in and see if Surge is a fit for your investment thesis, the link in the description will take you to the investor site. And once again, that is ticker N-I-L-I on the T-S-X-V and N-I-L-I-F on the O-T-C. Thanks again to Surge Battery Metals for sponsoring this part of the video, and let's get back to it. So the input to power is a set of very specific physical materials, and specifically what bottlenecks that layer. So the most critical resources in the set of physical materials required to create power are a set of critical minerals, which are like the bottleneck of the bottleneck. You have power in many ways, bottlenecking compute, which is really what AI is. And then you have a set of critical minerals, bottlenecking in some ways, energy or power production. And the core reason for that is exactly the same that we talked about in the previous video, which is AI is constrained by the time it takes to do things in the physical world, right? Maybe your model is getting exponentially better, but maybe it takes two years or five years or whatever it is to build physically in the world and get the permit for the regulatory approval, your data center, which powers the increased capabilities and usage of that AI. But if we go a layer below the data center, a mine might take 10, 20, 30 years to go through the full process of siting, going through funding, regulatory approval, building the mine, and then scaling up production over time, refining, exporting, etc. So if we look at an AI data center, what are some of the specific minerals or materials that are required to make that happen? Well, on one hand, you have copper. Copper is very good at carrying power. So the wiring of things like transformers and your cooling loops, that's in a lot of cases, in most cases, going to be copper or used copper of some sort. Then for your chips and your fiber optic cables, so the actual GPUs and what connects them, you need things like gallium and germanium. And then beyond that, in a data center, your fans, your motors, your pumps, things like that use rare earth magnets. So these very small, very strong magnets, which require what we kind of arbitrarily call rare earth materials. And then finally, you have a set of resources used for battery systems. That's things like your lithium, your cobalt, your nickel, your graphite. And so those are just some of the minerals that are called critical. And specifically, there is a list that's updated over time, which determines or just denotes what is critical or what is not critical. And in part due to AI and exploding demand for power, you see things like copper and silver being added to this list in the last year or two. Now, across all of those materials, the process to make them useful is roughly the same. First, you dig them up from the earth. And digging these resources up is not easy. That's a very expensive, very long and difficult process in of itself. But what's really important and really, really hard to do is the refining of these materials or taking what you dig up and turning it into something that is commercially useful. And so at least geopolitically, who mines it matters, but who refines it really matters more. That is the critical junction of that whole process of using and having access to these minerals. And then the reason we get all this geopolitical tie-in to these rare earth minerals and critical resources is because China controls about 90% of global refining of these materials. Only about 60% of the actual digging them out of the ground, the mining, but again, over 90% of refining. And I guess even crazier than that with some of this, like if my numbers are correct here, something like 94% of the finished magnet production that is also done in China. So you might be digging the rock out of California, but then you still ship it somewhere else to get it refined. And that's something the Chinese are great at, which is good for some people and for other people and government bodies. That's not such a great thing. Another angle on that, for example, the IAEA tracks about 20 minerals that they consider strategic. And China is the top refiner by far in something like 19 of them. So long story short, if we consider this layer of the stack, it is absolutely dominated by this one country. And there's a lot of implications there, right? That means, okay, great for that country, big industry. What it also means is because of that mismatch, there are going to be a lot of investments and attention put into developing that kind of capacity elsewhere. That means tens of billions, hundreds of billions plus of dollars put into startups, put into established companies trying to get into these spaces. They're also going to see geopolitical tension because of course that refining layer is going to be used as a bargaining chip. And so you can start to understand a lot of what's going on with, for example, mining companies or a certain geopolitical case, if you understand this grander context of AI, of power demand, of electrification as a whole, of defense technologies, which require all of these materials and so on. It's because of that stack in that bottleneck of refinement. And so I want to go even deeper into just one of these materials to give you an example of exactly how important it is for something like AI. And let's start with copper, which I've made a bunch of videos about before. And this is a crazy stat. For every megawatt of capacity in an AI data center, somewhere between 30 and 47 tons of copper is used. And people are building 100 megawatt, 500 megawatt, 5 gigawatt campuses these days. So imagine 30 tons times 500 megawatts times 100 plus projects out there like that, right? That is a massive amount of copper just for AI, ignoring all the other, again, defense, general electrification that copper is used for. And that's just inside the data center, right? Then you're talking about your transmission lines, your substation, your grid infrastructure. A gigawatt data center that might need 40,000 tons of copper also needs a massive amount of grid infrastructure, in many cases built out around it. That is even more copper. And meanwhile, if you look at the supply side of the market, so how many copper mines there are, it's barely growing. These mines take a very, very long time to scope out, to test, to develop. And so you have this crazy deficit already, but what is going to become an even larger deficit over time, where there just won't be enough copper. And it's this bottleneck in this whole chain. And just to give you ballpark numbers, on average, you might take 15 years globally to stand up and actually get production out of something like a copper mine. In the US, that's closer to something like 25 or 30. So again, if we just think about this, right? Okay, a data center might take two or three years if you have the material to build and turn on or scale up. But if you have mines that take 15, 20, 30 years, of course that's going to become a problem given enough time and enough demand. So that's just one example. Again, copper is one of something like 60 materials on the list now in the US. And like I said, again, there are a lot of implications there. Number one, refining capacity is going to be used as a geopolitical bargaining chip. Number two, there's going to be more investment or you would at least predict more investment into mineral mining and development. Number three, where that investment comes from is going to change. For example, in the US, you have the case of MP Materials. It's kind of a public-private partnership. Four, you're going to get reserves for this kind of thing. And then five, the technology is going to change, right? The R&D in this space is going to try and diversify away from these kind of supply chain bottlenecks and materials, which are just hard to get over time. That may take a very long time, but that is one of the dynamics that will play out. And so if I had to summarize the supply-demand side of this, it's that demand moves at the speed of software while supply moves at the speed of mines. So at this point, we're getting pretty deep into the stack. We have the consumer AI platforms, so what you engage with. Under that, we have the AI models. Under that, we have the data centers, so your GPU, your CPU, whatever, the training, the inference, et cetera, is run on. And then as those components, right, we have this set of minerals. That's a pretty high level stack of AI and what goes into that. And so you should keep watching the series because we have a lot more to get into about AI from this very high level bird's eye perspective. But anyway, that was the video for today. I hope you find it useful. I appreciate you being here and I'll see you in the next one.