The U.S. Can’t Build AI Without These Materials

Critical minerals fundamentally underpin everything that we do every day. They’re in your phone, your AirPods, your screens, and your laptops. Everything that we use—and how those minerals are refined and mined—happens in the background. This is the intersection of geopolitical urgency and tech.

We need a lot of aluminum. We need an insane amount of copper. We need more iron and more zinc. Now we have technology that can actually go and disrupt this: hard-tech companies working in dirty spaces and willing to go out in the field. There aren’t that many massive markets left that have been largely untapped by technology, and mining screams one of the largest markets in the world. So now is the time to build this company.

So, Turner, you’re coming out of stealth with $85 million raised. Why don’t we get into what critical minerals are and why they matter?

Critical minerals fundamentally underpin everything that we do every day, and that’s why we’re personally really excited about them. But it’s not just aerospace, energy, renewable energy, battery energy storage systems, and the massive growth in AI that’s happened in the last 12, 18, 24 months. It’s also defense, obviously, but it’s also everything that we use every day. You have rare earths in your phone, rare earths in your AirPods, your screens, and your laptops. It really does cross everything that we use.

Where they’re produced, how they’re refined, and how they’re mined all happens in the background. It’s something that really does need to be brought to the foreground—something that we need to support more and more. It’s a long chain to go from digging something up all the way through to something that can actually be deployed in an end product, and we’re excited to talk about that.

Why don’t we get into how we turn rocks into batteries or magnets, and why is that so important?

It starts with mining. Actually, it starts with exploration.

You’ve got to find the rocks.

That’s right. You’ve got to find the rocks in the first place, which is hard to do, and there are a lot of awesome companies that are working on trying to condense that timeline.

But once you do find them, you have to get that asset or resource permitted for extraction. You develop a mining plan, and you have to mine it. When those rocks come to the surface, you have to separate ore from waste, which is something that is not as trivial as people might expect.

Then you go through a concentration step. The ores will come to the surface at less than 1% concentration—definitely less than 5%—unless you have a world-class deposit. You’ll typically go through a concentrating step. That can be mechanical, thermal, or chemical, and it gives you an intermediate product. Those intermediate products move all over the world and typically go to refining assets.

The refining operation effectively goes from anything that is like a 10% concentrate to a 50% intermediate product and turns into a high-purity metal. Then you go into a specialty chemical. This is the intermediate product where you go through another chemical process to either make a metal sulfate or a metal hydroxide salt.

Then you convert that into an engineered material, which is the next step. In electrochemical systems and batteries, you’ll have cathode materials and anode materials, and the morphology and electrochemical performance in the system are really important. Then you’re ready to deploy into a battery cell, and then you’ll go into a module, a pack, a car, or a stationary storage product.

On the magnet side of things, similarly, you’ll get to a refined rare-earth product. There’s a long list of rare earths. They often get bundled into one group, but it’s important to break them out. The common way of making magnets—there are a few flowsheets—is that you’ll slurry the material, get the right blend of the different rare earths that you’re trying to put in, cast it, sinter it, and then go through a fairly intricate, high-precision machining process to get the geometry and tolerances that you need before you can deploy that into magnets and eventually into motors.

How specific is it for a given site, given the concentration and other waste products? How dynamic is it? Is one rare-earth mine going to have a process similar to another, or is there going to be a very bespoke setup?

It’s very bespoke. It’s actually part of the problem and what makes the minerals industry so complicated. The flowsheet, which is ultimately how you go from the ore all the way through to the refined metal, is designed for that specific asset.

You’ll have concentrations of impurities that you have to manage. The concentration of the target metal is obviously different, and there’s a library of metallurgical unit operations that are all stitched together to build a refining or processing operation. But how those are stitched together is bespoke for the individual unit operation and tied to the chemical metallurgist or process engineer who designed the circuit in the first place. There’s a lot of human input into what that flowsheet ultimately looks like.

I imagine it’s very hard to change as the nature of the ore changes while you mine a site.

That’s right. Part of what we’re working on—and what we’ll talk about a little bit later, I’m sure—is how to design circuits that have a little more flexibility to process ore as it changes over time as you mine through the ore body. One mine does not actually have consistent ore coming out of it. The Earth is heterogeneous. The ore grades are changing, the impurity concentrations are changing, and there are different ore zones that have different properties, including how they are floated or concentrated and how they perform in a leaching circuit. All those things are custom-built for a specific asset.

One more question on this: What are the types of job titles and backgrounds of people working in this space? I imagine that, across the supply chain you just described, there are very different types of people with very different backgrounds, but they all ultimately have to work together. Can you talk a little bit about that?

That’s one of the big hard parts. You have geologists, geophysicists, mining engineers, geotechnical engineers, process engineers, chemical engineers, chemists, metallurgists, mechanical engineers, structural engineers, and civil engineers. It’s the whole gamut.

Plus the long tail of workers on-site who are moving things from point A to point B.

That’s right, and they also have a super-diverse skill set because you need everything from the mining engineers, chemical engineers, and geologists who sit around to operate the asset, in addition to the folks who have to manage the back office. That’s something that often gets overlooked when we’re thinking about successfully building and operating a complex circuit.

Did you always love rocks? Did you always know that you were going to start a mining company?

It’s a funny story. The day that I graduated from college, the urge that I had was to just move to Australia and try to find a job in a mine. I did not act on that urge. Instead, I started at Tesla roughly 10 years ago.

I started out working on factory design and factory construction, and over that 9-plus-year period, I slowly worked my way upstream in the value chain. I worked on factory design and construction, then started working on battery-cell manufacturing. I spent a lot of time in Japan with Panasonic, our primary battery-cell manufacturing partner, working on incremental improvements to their legacy battery-cell manufacturing systems.

The pull has always been big things for me—large-scale infrastructure that has a large impact on the world. If you want to have a big impact on the world, you have to build things at scale. That’s how you get to impact.

I was working on battery-cell manufacturing, and because I was spending a lot of time in Asia, I started to explore the supply chain. I was building some of the early technoeconomic models of how cathode materials are made, how anode materials are made, and the balance of the components that go into a battery cell.

Ultimately, this was just following cost. When I was working on factory design and construction, the most expensive thing was actually the equipment that goes inside the factory. Then, when we started working on cell manufacturing, I realized that the expensive part of making cells is the stuff that goes inside the cells.

As you start getting further and further upstream, you realize that the primary driver of cost is the metals that are going into the engineered materials, which then go into the cells and eventually into the battery. I started digging much deeper into why metals are so expensive. What you run into is that there’s an interesting incentive misalignment that exists between the customer and the producer of metals.

In the mining industry, more demand—this is totally different from manufacturing—means that if you have a higher volume, the expectation is that the price goes up because it’s a constrained supply. The pricing dynamics are totally different from manufacturing companies, where higher volume means lower cost. As you scale up and capture economies of scale, you’re starting to drive down the cost that you can then transfer to your customers. That expectation doesn’t really exist in the mining industry: if you want more of it, it’s going to cost more.

That incentive misalignment was a big one that jumped out, and it only really starts to come to the surface when you start to engage with the mining companies. But what fascinated me about the mining industry is that you are effectively solving problems at micron scale to start. You have to figure out how to extract it—you have an ore that has 1% of a target metal—and get it to 100% purity. That fundamentally starts at the atoms. Then you have to take a process that you develop at micron scale and deploy kilometer-scale infrastructure. Going back to the thing that was exciting, which is scale, that opportunity to work across those scales is exciting.

Something I found interesting about Tesla, especially in the early days working with Panasonic, is that a lot of the know-how—the process knowledge—came from Asia. You spent a lot of time in Asia. I’m just curious, culturally, and just looking at scale as they built out a lot of the early battery ecosystem and then farther upstream, what did you see there? Is it just more chemical engineers? Is it more support for that industry?

Yeah, I think in the battery cell world, the precision at which you need to manufacture the product—the tolerances on the final battery cell—requires a level of rigor and attention to detail that does have a cultural aspect to it.

Like semiconductors.

Yeah, exactly. But also, it’s a long-term investment. In Japanese, it’s Kaizen, right? You’re gradually improving over time. In those super-high-precision, super-high-throughput industries, taking big swings where you make a radical change to 1 unit operation didn’t really happen in those companies. It was much more of an iterative improvement to the systems that eventually enabled you to get cost down. But yeah, the labor pool is a big piece, and I think some of it is definitely cultural.

Your time at Tesla—Tesla famously vertically integrated very early. You’re building a vertically integrated mining company, which we’ll get into in more detail later, but what did working at Tesla teach you about vertical integration and why it matters?

As we started to scope more and more vertical integration, even outside of going further upstream in the supply chain, the thing that’s really interesting about vertical integration is that you fundamentally are thinking about the incentive structure that exists between yourself and a partner. Tesla had to vertically integrate early because people just weren’t making the parts that were needed. That was do or die. There was no incentive; there was no market for people to build the subcomponents that were required. Ultimately, Tesla had to vertically integrate from day 1.

The things that pushed increasing amounts of vertical integration ultimately were the incentive-structure misalignment where suppliers and partners weren’t incentivized to invest and scale at the rate that we wanted them to invest and scale. They weren’t incentivized to innovate at the pace that we wanted them to innovate, and so you end up insourcing a lot of that development that enables you to get to the product specs or the component specs that you want. Then it takes a lot of guts, because at the end of the day, when you vertically integrate, you are transferring the risk profile of your partner into your new, expanded risk profile. You need to be really confident—or at least believe—that you’re better positioned to take on and manage that risk profile.

When you were at Tesla and looking at and developing relationships with all these global mining companies as Tesla was scaling and looking for suppliers, what were some of the key issues that you noticed from a market perspective, and then ultimately what sort of led to Tesla pursuing further vertical integration?

The incentive-misalignment piece between the industries—between a commodity industry, or the mining industry, and the manufacturing industry—was the big misalignment that existed between the industries. The degree to which automation was absent was pretty interesting. There’s a long period of time where mines make no money, right? A lot of capital goes into exploration, and it goes into development.

What a lot of people don’t realize is that when you start the mine, there are sometimes years where you’re just getting through the waste to get to ore, or drilling a shaft to get to the ore deposit. Oftentimes, the capital starts to get tired, and that additional automation equipment falls by the wayside. If you have people there who can drive the trucks or drive the excavators or run the drill rigs, you’ll take that.

We’re at this interesting inflection point now where those people are less and less available. The mining industry has been taking it on the head for a long time. It’s not been a sexy industry that everyone wants to go into. The labor pool is contracting; it’s shrinking, and it’s both the trades and the engineering skill sets. The first thing that mining companies say now is that the labor pool is one of the biggest challenges that they’re trying to solve for.

That was apparent. On top of that, these mines are not exactly in downtown Manhattan. They are in remote locations.

Where have you gone outside before?

Yeah, of course. Indonesia, Australia.

What is it actually like? People see pictures, but what is actually going on?

It’s a lot calmer than you might expect. You’re usually working a couple of faces, and you have excavators or front-end loaders that are picking up dirt. They’re taking it to the unit operation, and it’s not as rambunctious and crazy as you might expect a mine to be.

And that’s in Australia and Canada?

It’s not this buzzing atmosphere. In developing countries, it’s a little different. There’s definitely a different degree of automation. Obviously, we’ve come a long way from people with picks and shovels, and so when I say automation is absent, it doesn’t mean there isn’t heavy-haul machinery driving around the sites. But there’s a lot more people activity when you go to operations that are in Indonesia or in Africa. I’ve been to most continents.

Why did you leave Tesla to build Mariana?

I spent a long time building businesses within Tesla. I was early on the cell manufacturing side, early on the cathode manufacturing side, and early on the refining side of things. What really excited me was always pushing further and further upstream.

I do think that we are at this critical inflection point where not only is the labor pool going in the opposite direction of demand in the mining industry, but AI and ML are getting to the point where—and it’s not that it hasn’t been that long—the AlphaGo moment was in 2015 or 2016. People were using reinforcement learning. There’s a paper from 2008 on reinforcement learning for helicopter control.

But we’re at this point where the compute and machine learning, reinforcement learning, do really enable you to go to no humans in the loop in how a lot of these plants are controlled. As you build more of that large-scale infrastructure and see the problems that humans have to solve on a daily basis, it becomes pretty obvious that these are problems that humans aren’t best positioned to solve. These are large, multivariable optimization problems that RL is perfectly poised to solve.

Then on the construction side, that is an entirely different story. I think construction has a lot of workflow automation opportunities, and there are also tons of menial tasks where people are fat-fingering data between databases. The data systems are completely disaggregated, and LLMs are presenting this opportunity where we can start to—and again, this is really a 2-year thing, and it’s just going to get better—the opportunity to build from scratch, with no legacy systems, and have some control over the destiny of the company that we’re building. That’s just an exciting opportunity.

What is just a level set? What is the status quo in the industry today? What do BHP and Rio Tinto do, for a lot of the stuff you're talking about, from a software perspective, internally?

They have digital innovation arms.

They do have digital innovation arms. I think Freeport-McMoRan, Rio Tinto, and BHP outsource a lot of that. They've gradually started to hire more and more people who can do more things internally, but you have McKinsey and Palantir, which basically act as consultants. They'll look at the large data sets and provide recommendations. Some percentage of those recommendations, oftentimes below 50%, are taken.

What's interesting about what you want from the ML models or the RL models is that you actually want them to tell you to do things that are counterintuitive, because humans are naturally going to find a locally optimal operating condition. It's very risky to take shots outside of something that is currently working when there's a billion dollars on the line. That culture of trusting the counterintuitive recommendation from the model is one that we want to try to build, and it's hard to build that within large companies.

I think on the construction side, when they build new projects, they're building $5–10 billion projects. They're always bringing in an EPC and throwing that over the fence to the EPC. In the mining industry, it is a bespoke plant. It's custom, and you need to think about the refinery, the mine, and the processing facility as the product.

When you outsource that, you lose a lot of control over what you're eventually going to inherit and operate. EPCs have shifted away a little bit from the turnkey model—“We'll deliver you a project.” It's moved toward selling hours and man-hours, and selling reports, especially in the mining industry. They'll do a prefeasibility study, and they'll do a feasibility study. There's a lot of throwing things over the fence and outsourcing that happens in the large companies.

Some of that comes from the talent-acquisition challenge that the mining industry has faced over the last 20 years. Again, the mining industry has not been a magnet for talent. What that means is that even if they're able to hire the best-in-class machine-learning engineers and the best-in-class software engineers, they're not going to stay. They're going to run into the wall of bureaucracy, and they're going to have higher-paying opportunities in SaaS, fintech, and ad tech, and they're going to go chase those.

It's not just attracting talent; it's retaining talent. That has been a big challenge in the mining industry, and it forces an outsourcing of a lot of the core things that should be core today.

We've seen this from the investor side. There are a lot of really exciting new technologies being developed for mining, and a lot of incredibly impressive startups building for various pieces of the mining life cycle, whether it's autonomous vehicles, drilling, or other software and hardware tools for mining. The challenge seems to be, how do you get these calcified large incumbents, which operate in a very decentralized way, have very low risk appetite, and don't have a strong internal culture or affinity for tech, to adopt them quickly? If you're a young startup, you're sort of at the beck and call of this behemoth, and you have very little control over your own destiny. I think that's made it really hard for tech to penetrate this market up until now. That's at least what we've observed on the VC side.

“Calcified” is a good word. I think the way that this works in construction companies, mining companies, and really a lot of big companies is that fixing the status quo or making a step-change improvement in the status quo requires doing a thousand things. But you'll evaluate risk on each individual thing of those thousand things. The downside of each individual thing is that the plant goes down, which is a multimillion-dollar event. So you're really not incentivized to change things.

Even small changes could result in millions of dollars of loss. You need to approach it as, how do I do the thousand things all at once so that I'm not stacking incremental returns on innovation with the same risk every single time? That's where point technical solutions are challenging to sell into the mining industry.

They'll do pilots. They'll definitely do a pilot. There's no skin off their back to do a pilot, but you'll end up doing a lot of pilots. They don't build enough plants sequentially. They'll build one big mine every 5 years, if that.

There just aren't a lot of opportunities to get into a commercial-scale application. If you don't time it perfectly—if your pilot plant was 5 years before the commercial-scale plant was planned—you're not going to be in that one. You'll be in the next one, which is 5 years later. The pace at which the industry moves in terms of deploying commercial-scale infrastructure means that there just isn't a lot of opportunity to get new tech into commercial-scale applications.

A lot of people are doing SaaS products, too, which is the lowest-cost way to generate uplift in a mining project or a minerals refinery. The barrier there is ultimately: How do you get the operators to trust the recommendations from this SaaS tool, from this small company that's trying to tell you how to run a plant? The culture is typically, “Don't touch my things. Don't touch my cash register. What do you all know about running a mine?”

It does stack up, and it makes sense. I've been calling it a death spiral for a lot of the folks trying to sell into the mining industry, because it's hard.

Mhm. Just to step back a little bit on the geopolitical context, the stuff you're describing seems very obviously true of a lot of Western companies. At the same time, a lot of Chinese companies that have sprouted over the last 20 or 30 years have grown rapidly. Why do you think that is?

I think there's a lot of top-down and early recognition that critical minerals were going to be critical and needed to be supported. Everything around policy, and everything around supporting companies to deploy both infrastructure domestically and infrastructure internationally to secure critical minerals and build infrastructure that secures a position—that has definitely happened.

But I think what people often don't talk about enough is that the talent pool is insane. It's not just a large talent pool; it's a large, skilled, experienced talent pool.

I was in Indonesia in February and was visiting one of the recent Chinese nickel-refining operations. They buy ore, and they also have some mining operations. They had 13,000 people on site during construction and commissioning. If we were building a refinery in the US—which we did—it's hard to mobilize a tenth of that, realistically.

It's not just about the number of people. It's about being able to iterate on every individual work front as fast as humanly possible. We just don't have that labor.

15 years ago or 20 years ago, would the same companies that are big now have been big then? Where's the evolution of the space?

I think there's been a clear splintering around who does the exploration and who does the development. Right now, the industry is set up where junior mining companies, which don't mine, explore. They sometimes get a resource from a major mining company that's been held in its portfolio for a long time, but it's a different risk-reward profile from what the mining majors are ultimately looking for.

You have this junior-mining ecosystem that sometimes is well-funded and sometimes is competing for capital with the cannabis industry in Canada. They're taking shots in the dark, basically, and there's a lot of work going into making that exploration activity more intelligent: streamlining it, drilling fewer exploration holes, while still being able to interpolate or extrapolate what is in between those drill holes.

You're going out in the middle of nowhere. Either it's really far north, in the Arctic Circle or the Yukon, or it's overseas in Africa, Southeast Asia, or South America, and you're doing exploration.

Those folks have one job: to define a resource and pump up its value sufficiently to flip it to a major. A lot of companies aren't able to discover a resource that's either large enough, because the big mining companies want to deploy large amounts of capital. They're talking about multibillion-dollar investments, and so they won't really look at projects that don't have the scale that enables them to underwrite their own inefficiency. They want to build really large infrastructure that enables them to capture economies of scale.

There's actually a really long tail of mining projects that don't have the scale that would justify getting acquired at a major premium. They'll go into this kind of orphan period, as it's called in the industry, and it's hard for them to break out of that orphan period. That's kind of where we see our ability to step in as a more efficient builder and operator: take these—what the industry calls subscale assets, but we see metal there—and bring those into production as we're building the platform, and then eventually scale into the same scale that the big mining companies are operating at.

Your thesis is that you can get the metal out and process it into a product that you can sell more efficiently than the major, such that it's economically viable to offset the scale advantage.

Yeah.

Well, maybe this is a good opportunity to talk a little bit more about what Mariana's product is. Like you said a few minutes ago, you're not a SaaS product. What does it mean to be a diversified metals and minerals company, a technology-enabled mining company? Take us in a little more detail.

Yeah. We're a vertically integrated, software-first minerals project developer and operator. We focus on the back end of the minerals value chain, which is doing the detailed engineering, getting through the permitting, building the asset, commissioning the asset, and then operating the asset.

Going back to some of what we were talking about around the labor pool, those labor-pool shortages exist in construction and they exist in mining. They're felt very intensely. Our fundamental thesis is that, with a contracting labor pool, you have to start with an awesome team. The table stakes is that you build an awesome team, but how do you enable 200 people to do what 10,000 people are needed to do today, at least on the parent-co side of things?

That comes from leveraging recent advances in LLMs to automate workflows on the construction side, the engineering side, and the procurement side. These take an insane amount of time. You make a lot of lists and fat-finger a lot of data between databases, and that is all about reducing churn in construction. I think there's currently churn and latency.

Latency is one thing that I think people sometimes don't appreciate from status quo construction, like large-scale mega-projects. There's generally a 3-week lag for really large construction projects, where you're trying to aggregate data from all the different contractors and all the different parts of the facility into a consolidated, integrated schedule, which you can then make decisions off of: How do I prioritize what I'm doing today?

In between those 3 weeks, people stand in circles every morning and say, “What are you doing today? What are you doing today? What are you doing today?” Then they go off and do the thing. They'll send a very brief progress report back, and it takes a long time to take those progress reports and actually measure progress, so that you can reevaluate priorities and understand how the project is trending.

We're really trying to accelerate and democratize access to data, fundamentally, and run construction projects like manufacturing facilities. It starts there. The reason construction and mining are so integrated—and some people might disagree with me—is that a mining project is a big civil construction project. It just never ends. Or it's a deconstruction project.

Yeah, that's fair. You do that when you're constructing piles.

But there's actually a lot of similarities in just moving the dirt for site prep. The same kind of software stack that is enabling you to get feedback from the field live is the same thing that the mining industry struggles with. Mining companies will lose equipment, especially in underground mines that are these deep mazes, and the industry is getting better at having actual location sensing for where the equipment is. But losing equipment in the mine used to be a super common thing.

We start with construction, and then we start to get into the second core software stack, which we're calling Plant OS. The construction stack is Capital Project OS, and Plant OS is really aimed at removing humans from the loop in deciding how the chemical-processing and refining operations work.

Big refineries are effectively big robots. You have the sensing and telemetry; you have the actuators to control how the plant operates. Imagine T-800 humanoid robots operating forever. That is what the refining industry and the processing industry has been.

There are obviously PID control loops that maintain set points, so you can automatically maintain temperature and automatically maintain pH. But the thing that really matters is that the feed material to the processing facilities is constantly changing because the ore body is changing over time.

The way that the industry manages that today is that they blend the feedstock to minimize variability going into the processing facilities. That enables them to minimize the amount of change that has to happen in a processing facility. We're trying to flip that and say, “Okay, if we build a hyperdynamic and highly flexible refining circuit, ideally without adding a whole bunch of cost, what does that do to optimizing the global operation from the mine to the refinery?” It's first aimed at reducing reagent consumption and reducing energy consumption.

Google kind of proved this. They bought DeepMind in 2016 or 2017, and one of the first things they did was throw the DeepMind team at automating and optimizing the data-center thermal systems: air handlers, chillers, and cooling towers. That's not a super complex system. You have weather, which is a factor, and you have loads within the building, which are a factor. But you ultimately have nine control variables, including airflow rate, supply temperature, and the cooling-water temperatures and flow rates, both in the chiller system and in the cooling-tower system. In that relatively simple system, they were able to reduce energy consumption by 30% to 40%.

It happened relatively quickly. That's the opportunity when you remove humans from making the decisions on how these process systems operate. That's the opportunity. When we look at refining and processing facilities, that's like 1,000 control variables.

It's no longer single-pass, because what's really interesting about minerals refining is that you never want to lose the metal, right? Every piece of metal—or every atom—that you lose in the processing facility is another atom that you have to mine. Recovery in the refinery is actually the biggest lever when it comes to cost.

Relatively simple.

Yeah, right. The upstream operations obviously impact the downstream operations, because if you're changing the process conditions in the upstream operation, that changes what the downstream operation is seeing. But the downstream operations will recycle the reject stream back into the upstream operations.

It's this big interconnected web, and it's a high-latency web, also. If you make a change in one part of the circuit, you may not see that change cascade for another 24 or 48 hours. When we're commissioning refineries, that latency ends up being a major driver of the time it takes to bring a refining operation to spec and eventually ramp it to throughput.

So how long does it take to commission a refinery today?

There are some refineries that were built recently that are still not commissioned. They were built 3 or 4 years ago. Chinese companies are doing it in about 6 months, and for a lot of Western companies, it takes 2 to 4 years.

And that stacks up when we need to build an insane number of mines and refineries. If you're taking 4 or 5 times longer every time you build a refinery—

Every step of the process.

Yeah. And so we're trying to bring down the time it takes to bring the refinery to spec—basically, reaching the required throughput and hitting the output requirements of the product you're making. Ultimately, you start this historically very long haul of gradually bringing down the cost over time. That's something we think reinforcement learning is going to do quickly, much, much faster, in line with what Google demonstrated with thermal systems in data centers: achieving globally optimal operating conditions on an order-of-magnitude-faster timescale.

So how do you think about building a company that mines and refines a product? There's a lot of technology that you can inject at essentially every step of that process. How are you deciding what to build, where to partner, what you're developing in-house versus what you're going to market?

Yeah, I think at the beginning we're focused on: How do we take commercially demonstrated unit operations and be a better integrator and a better operator of that integrated plant? So, we focus on the software systems that enable you to control the plant more optimally. That's generally what project-level financing parties want to see. Also, it's hard to get project finance on a first-of-a-kind facility where you're demonstrating a new unit operation for the first time.

As we're entering the market, we think the right place to start is to take commercially demonstrated individual unit operations that operate globally and go after the uplift available just by being a better integrated operator. There's a whole bunch of bottlenecks in building these facilities that we will need to solve. The industrial supply base just for manufacturing tanks is broken.

That's a new one.

It just means that things we take for granted take a really long time if you don't want to go to China to source that equipment. That has a big impact on the operating side too, where the supply chain for a new pump in Australia could take 30 weeks. Getting that exact same pump from a manufacturer in China, it shows up in a week or 3 days.

That entire industrial equipment supply base, we're going to have to look at at some point. That's obviously a much bigger bite to go after, like commodity equipment manufacturing.

You're not going to vertically integrate to be a mining equipment manufacturing company. I don't think so. I hope not. You'll let me know.

Yeah, that's right. Well, this is the question: What is the incentive structure of the partners and the suppliers, and is it required or not? I think there are a whole bunch of companies working on awesome, novel process technologies that have not quite gotten over the hump of trying to sell to the big mining companies. We want to be the customer and the partner that helps accelerate commercial deployment.

One of the big issues that comes up when you're deploying new processing technologies is that part of the reason it takes a long time to get to the point where it's commercially viable—other than all the headwinds from the industry being conservative and process-driven and all those things—is that humans have actually never operated that process chemistry at scale before. You learn a bunch of things at pilot scale, but a pilot doesn't really tell you what's happening at scale.

Train people.

You have to train the people to operate it. There are new environmental things that might come up, depending on the chemical you're using. That scale jump is actually something we think reinforcement learning will enable with a pretty meaningful pace adjustment. You don't need the humans to fine-tune the process conditions around a new process chemistry because Plant OS is doing it.

Ryan and Erin, how did we approach this industry? Is this a space that we spend a lot of time thinking about, or thinking about opportunities in the space? How did we approach it?

Yeah, we've wanted to do a mining investment for a long time. When you think about venture capital, we care about massive markets, and there aren't that many massive markets left that have been largely untapped by technology. Mining sort of screams as one of the largest markets in the world. There's very little adoption of technology.

Over many cycles, we've gone out and spent a lot of time meeting companies. The challenge is: How do you sell a point solution or a piece of technology into this industry, which has very little incentive to adopt it and also has a very complicated geopolitical dynamic, where you have a very large global player with its hand on the scale?

We put out a piece a couple weeks ago around our thesis in mining and why we think a vertical mining company is the answer. We actually do believe you have to control every single piece of the entire journey—the entire life cycle of an atom of metal, end to end—to actually be able to build a tech company here. This is not about a point solution for one particular part of the process. To capture the gains in efficiency and build a feasible business, you really have to own the entire process end to end.

The only thing I'd add there is that this is the intersection of geopolitical urgency and tech. As Turner Caldwell has been talking about, now we have technology that can actually go and disrupt this. But there's also a talent base: people coming from companies like Tesla, SpaceX, Anduril, and other hard-tech companies working in dirty spaces, willing to go out in the field, roll up their sleeves, go out in the middle of the desert, and work on this stuff. So now is the time to build this company.

And the political tailwinds are there. Even my conservationist mother, who I think, if we'd had this conversation 5 years ago, would have clutched her pearls—she doesn't wear pearls, but she would have clutched her pearls—at the idea of domestic U.S. onshore mining.

Broadly speaking, the American public, and certainly the government, has come around to the idea that metals are in every single thing we use as consumers. Our supply chains are highly reliant on China. It's a huge problem. We have to figure out how to address it. That means investing in mining in the U.S.

We talked a little about rare earths, lithium, and things like that, but there are many different critical minerals. You talked a little about this at the very beginning, but specifically, what are the interesting ones for you? How does that map to what people see in the headlines, and what are the business opportunities?

When we look at what needs to happen in the next 10 years, forecasted demand will only materialize if the supply is there. So we'll see if that forecasted demand materializes. The metals that need to grow the most by mass flow rate are the big metals. We need a lot of aluminum. We need an insane amount of copper. We need more iron. We need more zinc.

What are some of the things that these metals go into?

For sure. Iron goes into everything that is infrastructure.

We got iron. We're good with iron.

Zinc is one that people sleep on because you actually have to galvanize a lot of that steel. Zinc often pops up every once in a while as being something that we really do need to continue to focus on.

Copper is the workhorse of this push to electrify everything and grow the grid. To be able to supply AI, enable accelerated renewable penetration, and allow EV penetration to happen, you're going to need a lot of copper. Aluminum is one I think is underestimated. People underestimate its importance. It's actually the number 1 most-consumed metal in defense applications.

The grid is something people talk a lot about in terms of copper, but there are a lot of aluminum conductors in the transmission lines that are critical to growing grid capacity. In automotive, obviously, aluminum is big. Magnesium has a whole bunch of defense applications and could get more into automotive applications for lightweight metals.

Lithium needs to 4× in terms of production capacity over the next 10 years, roughly, in order for the batteries we want to build to be built.

Well, we're all about batteries.

Right. Nickel is a big one.

I think what has happened in nickel over the last 5 years is that Indonesian production capacity has scaled to the point where something like 70% of global nickel now comes out of Indonesia. A lot of that was on the back of meaningful investment from China to expand production capacity in Indonesia and do more of the downstream processing there.

Nickel goes into everything that involves specialty alloys—anything that needs high-temperature or corrosion resistance. It’s also the unsung hero of high-energy batteries, where these lithium transition-metal oxides are high-nickel. Manganese is important, too: it goes into a lot of alloys and also into batteries. Uranium is going to be needed if fission continues to grow and we continue to deploy more nuclear capacity in the United States.

It’s a long list. Rare earths are important, obviously; they’re omnipresent in almost everything we use, but they show up as a relatively small amount on a volume basis when you look at the stack of metals we need to mine. We definitely need a ton of process innovation in how rare earths are refined. Solvent-extraction circuits are the status quo, but the chemical intensity is high, recoveries are relatively low, and the know-how is highly concentrated in China.

It’s a little bit of a frothy market right now, so being a diversified minerals company enables us to pick our spots in areas where it makes sense. These things still move on commodity cycles, and you actually want to build infrastructure at the bottom of commodity cycles, not at the top. It’s the Warren Buffett quote: invest when there’s blood in the water. You want to come into metals when they’re at this trough, when no one is investing in them, even though they still have a macro, long-term critical importance.

That’s why we’re focused on lithium. Copper has a macro trend that’s pretty hard to ignore. We’re going to need an insane amount of copper, and copper grades are going down globally, which means our ability to extract copper from those ores is going to get harder and harder. On the plant side of things, we have a high degree of confidence that we’ll be able to step in and optimize the refining circuits to extract copper from these lower-grade ores without seeing meaningful cost increases.

Everyone knows—or people hear—that it takes forever to get a mine started. I don’t know how many new greenfield mines we’ve developed in the United States in the last decade.

Not many.

I know Australia and Canada have been able to do this faster, which is interesting. You don’t think of Canada as moving quickly. What are some of the bottlenecks there? What does America need to do to accelerate this, as one of these companies trying to not only mine but also refine in the United States? What needs to be done?

I think one thing that folks don’t always see is the permitting requirements for exploration. If you’re exploring on federal land, and you’re exploring over more than a 5-acre parcel, you have to submit a plan of record or plan of operations that needs to be approved by the BLM before you can expand and explore over a larger piece of land.

Bringing down the permitting thresholds and the permitting burden associated with exploration is important. That’s why we have such a small rare-earth resource. It’s not because the United States doesn’t have tons of natural resources; the USGS estimate for the US reserve of rare earths is tied to a lack of exploration activity, not necessarily a fundamental lack of geological presence.

We haven’t either looked for it, or—

It’s hard to find in high concentrations that are mineable, which is what we’re trying to address by dropping the grade requirement that makes something economical. But there’s also a lot of permitting burden involved in deploying drill rigs to actually explore.

The government is currently doing a good job of highlighting the importance of the minerals industry. You’re definitely seeing a tone shift over the last 20 years that is much more supportive. There are way more tailwinds when it comes to making mining be viewed in a more positive light, and in a critical light. That will help solve some of the talent-pool problem.

People who are awesome and want to go build things don’t want to work on a project that sits around for 5 years and maybe gets permitted and maybe doesn’t. They want to work on hard problems where they can see the impact of the work they’re doing. If we’re getting in the way of enabling projects to get built, that’s a major deterrent for talent because people won’t actually see the output of their work.

I think the permitting requirements for going from a discovery to an operating asset should broadly focus on efficiency in reviewing environmental permits. There should be a big focus on streamlining those workflows and the back-and-forth between field offices and state offices at the BLM, focusing just on the federal side of things.

The way projects get permitted right now is that you submit your environmental assessment, and then they divvy it up among a whole bunch of experts or consultants they bring in to review the permit. They’ll get back to you eventually, at some point, but there isn’t a lot of visibility into how they’re progressing with the permit applications.

Discussions are getting more bilateral, and there’s definitely been a change with the new administration, with a little more accountability for the permitting offices. But there’s tons of room to make those reviews more efficient. LLMs will make it more efficient; we just need to penetrate that side of the federal bureaucracy and enable people to review things faster.

Aside from permitting efficiency, what are other things that, if you could send a list of recommendations to the government for how they should support the US mining industry, would be your top 3?

Supporting the demand side is probably the biggest lever. If you want to mobilize private capital into the sector, having some level of support on the demand side is major. That means offtake agreements with floor pricing, and they did this just now with MP Materials. Ideally, that provides some stability on the revenue side so that investors can participate.

There are trillions of dollars of dry powder just sitting around waiting to be deployed. That capital has historically avoided the mining industry because of market-price uncertainty. It’s a commodity cycle: what if you’re building at the wrong time? Infrastructure funds aren’t here to play the commodity-price cycle intelligently; they’re looking for annuity-type returns.

Those investors would mobilize if there were more demand-side support from the government, either through price floors or fixed pricing for critical minerals where you’re trying to incentivize more production in the United States. Participating in the capital stack is important, too. Lowering the hoops—the extra burden that comes with receiving government funds—is important, and some government agencies probably have more leeway to do that.

The Department of Defense obviously just did this big deal with MP Materials and went all the way to participating in the cap table as an equity holder. But when you receive federal funds from the Department of Energy, or federal funds on the debt side from EXIM, it sometimes comes with additional burdens. If you’re building on state land and just need a state permit, then bring in federal funds, you now bump your permitting requirement up to a federal-level permit. That’s the NEPA process, which wouldn’t be as burdensome if there were more efficiency on the permitting side.

Mineral deposits—specifically, high-grade mineral deposits—don’t obey borders. Is there a broader international strategy here? I would love to think we can mine and refine everything in the United States, but obviously there’s a lot in Australia, Canada, and Latin America. I’m curious about Africa, underwater resources, and the seafloor. What’s the overall strategy in your mind?

We’re starting in the United States because it’s closer to home, and we’re focused on developing a platform that we can scale off of.

But at no point have we told ourselves that the U.S. is the only, the sole focus. You have to be able to bolster the company to operate internationally if you want to scale beyond the resource base that the U.S. has available today. More exploration is going to happen in the U.S. We’ll probably discover more resources, and that pool of projects that we can build in the U.S. will grow over time.

But yes, we are absolutely going to expand overseas—and underwater, maybe.

When we look back a decade from now, what’s the single clearest indicator that Mariana has achieved what it set out to do?

We won’t be as worried about our ability to secure the critical minerals that we want to secure, because we will have rebuilt and established an entity, ideally, that is able to go across borders, to your point, and build these projects cost-effectively, time-effectively, and responsibly, ultimately. The reason that we are so panicked about it right now is that we have fundamentally lost the ability to build large-scale infrastructure, and we have lost the ability to operate complex mineral plants. That’s what we have lost, and we need to build that back.

We want to build 10 projects in 10 years. Those projects will be of increasing scale over time, but the work will not be done in 10 years. What I think will demonstrate that the 10-year mission will have been accomplished, other than building those 10 plants, is that we will no longer be as worried about our fundamental capability to go and build this complex infrastructure. We will have unlocked it.