Dan Dreyfus：AI 的下一个瓶颈是铜

He's with Fornight Capital.

We're going to be measuring human progress by how much electricity we consume.

The semiconductor industry, I view that as an industrial or infrastructure company. I mean, it's effectively a factory.

In the next 10 minutes, I am going to try to teach you about critical minerals, commodities, and our incredibly fragile infrastructure here in the US, which is going to require trillions and trillions of dollars of investment if we want to achieve our technological objectives, our reshoring and reindustrialization objectives, and our national security and military objectives.

But first, a little bit of history. We are at a very significant inflection point right now in US economic growth and what it is going to look like. From the early 2000s until just a few years ago, the US went through what I think was effectively an economic miracle, where we created so much growth, market cap, and value without really having to invest any capital at all.

Think of all the companies that were created with no capital. You had Google with the search engine. You had Meta with social media. They bought WhatsApp for $30 with 12 employees—no capital whatsoever. You had the streaming platforms and the food delivery platforms. You had Apple Computer, which was capital-light and created trillions of dollars in market cap. You had software as a service. Absolutely no capital was required to create all that value.

At the same time that we were creating these companies, we were literally tearing down all of our critical infrastructure and moving it overseas to China. We were really doubling down on that capital-light mentality.

But then it started to come back to bite us. We had COVID, the Russia-Ukraine conflict, the tariffs, and now we have the Iranian conflict. Every time we had one of these geopolitical flare-ups, inflation spiked like a rocket. You needed a telescope to see how high inflation went, and it never came down.

The reason for that is that we let our supply chains get far too fragile and weak, and there is no resiliency in them. Now we are at this inflection point where we want to reshore everything that we tore down and moved to China. We want to reindustrialize. We have this technological compute revolution that is infinitely more infrastructure-intensive than compute was in previous generations, and this is creating a really wild demand shock for infrastructure, critical minerals, and commodities at the same time that there is a supply shock because we just have not invested in this stuff for so long.

There are so many capital cycles going on at the same time. I have never seen this many going on simultaneously in my career. We have the aerospace cycle. Boeing and Airbus have a trillion dollars of backlog over the next 10 years. Now throw in the space economy, which is going to compete for the exact same materials and backlog that Boeing and Airbus are trying to source.

We have the grid. Anytime it gets a little bit cold in Texas, ERCOT—the Texas grid—is not connected to the rest of the US grid. Every time it gets a little bit cold, that grid shuts down and people are freezing in the dark.

Then we have Paradise, California. There was a power line that caught fire and killed 300 people. Did you know that that power line was over 106 years old? There are parts of the grid in this country that are over 106 years old.

Here in California, if half the people buy electric cars, or there are robotaxis, and we all go and plug them in at 6:00 p.m. after work and turn up the air conditioning, we are just going to kill the grid. We are all going to be sitting in the dark. The grid barely works for what we need it for right now, and we have not even started talking about the tsunami of electricity demand that AI is going to bring.

There is power generation, too. We have let China go and build multiples more power generation than we have here. This is a trillion-dollar-plus capital cycle that is probably going to be a trillion dollars every 10 years for the next 30 years.

Data centers are now a trillion dollars per year in infrastructure and commodities. Then there are semiconductor fabs. The CPU is making a huge resurgence. CPU intensity is going up like a rocket, and I bet this number is way too low: $750 billion. I bet that is going to be measured in the trillions.

Then there is defense. Taiwan has turned into a porky pie. Japan is raising its defense budget. Europe is raising its defense budget. The US is raising its defense budget. The similarity among all of these end markets is that none of them will work without critical minerals. None of it can happen.

Here is the problem: Last April, China announced that it was going to cut off exports of some critical materials to the US—samarium, gadolinium, terbium, dysprosium, lutetium, scandium, yttrium, erbium, and silver. It just cut them off.

We are close to a lot of big industrial supply chains, and with the cutoff of samarium-cobalt magnets, we learned that Ford Motor Company was literally days away from having its entire production line shut down—the whole Ford Motor Company. And same with McDonnell Douglas, too, by the way.

This put people in the Department of War and the Department of Energy into a panic. To their credit, they are doing something really aggressive and important. They are now going around to small resource owners across the US and into Canada, knocking on the doors of companies that were left for dead over the last 20 years.

They are saying, “Here are 3 pieces of paper. The first piece of paper is an equity check that we are investing into your company so that you can start converting your resource into a mine.”

Then the company says, “That is great. That is a shock. But the problem is that I have been waiting on my permit for the last 20 years. Nobody wants to give me a permit.”

They say, “Look at the second piece of paper. There is your permit. Go and start building right now.”

Then they show them a third piece of paper. The company says, “What is this?”

They say, “This is an offtake agreement—a take-or-pay agreement with a minimum floor price—that is going to guarantee you a very high internal rate of return on your project. You can keep all the upside above the minimum price, but here is a minimum price that will allow you to go out and raise a bunch of capital to get this thing fast-tracked and up and running.”

China has an absolute grip on all of these critical minerals. It is going to take at least 10 years, probably 20, to catch up, but we have to start somewhere. We just cannot have China leading over us and squeezing our testicles every time we do something they do not like, saying, “We are going to cut off your exports. We are going to cut off exports of critical minerals, and you are going to freeze in the dark.”

I give a lot of credit to the administration for doing this. I have been doing commodities for 25 years, and I have never seen anything like this happen before. It's truly what I call a déjà vu day moment, which is the overwhelming feeling that none of this has ever happened before.

Here is copper. This is the king of metals. This is just one example. We need copper for everything.

If we want clean energy, solar power per megawatt takes 5 times the amount of copper as a typical baseload CCGT gas-fired turbine. It is the same with wind: 7 times the copper.

For data centers, a 1-gigawatt AI factory now needs 50,000 tons of copper per gigawatt. We are going to start building 15 gigawatts of these things per year. So, 50,000 tons per gigawatt multiplied by 15 gigawatts is 750,000 tons of copper that we are going to need for these data centers.

Do you know what the copper supply was last year? It grew by only 500,000 tons, and that is just the data centers.

Then there are electric cars. If we are going to have robotaxis everywhere, an electric car consumes 5 or 6 times more copper than a traditional internal-combustion engine vehicle.

Then there is the military. In the Russia-Ukraine conflict, did you know that we used more explosives than in all of World War II? Do you know that? The artillery shells used for these explosives—guess what they are made of? One of them is called the Copperhead, very cleverly named after a poisonous snake. They are all made of copper.

Do you think we go onto the battlefield and recycle that copper? No. That copper is gone. We need these metals for everything that we do.

Now, where are we going to get it? Going back in human history to Mohenjo-daro, we have mined 700 million tons of copper over the past 10,000 years.

We could probably get 80% of that copper back if we wanted to. But what we would have to do is tear down this building, rip up the grid, and tear down the buildings in Europe, Japan, and China. We could get all that copper back, sure, but then we would be doing this conference in a tent.

So, how are we going to get it? Right now, copper demand is 30 million tons per year. About 4 million tons of that supply comes from recycled copper. The rest—26 million tons—is mined.

If we just grow in line with GDP, forgetting about data center upside and green-energy and solar upside, just growing at GDP like we used to—listen carefully—that means that over the next 18 years, we are going to need 700 million tons of copper.

Over the next 18 years, we're going to need as much copper as we mined in the last 10,000 years. That means we're going to need 5 world-class, mega, Tier 1 mines coming online every single year. You can go and Grok this or ChatGPT this. You can count on one hand and have some fingers left over the number of Tier 1 mines that are coming online between now and the end of the decade.

I don't know what they're going to do, because it takes 7 to 12 years to build a copper mine. The existing copper mines are dying. The big mines in Chile are over 100 years old, and the grades are depleting. This is going to be a major, major challenge and an upcoming bottleneck.

Today, all the rage is in memory, HBM, and NAND. Prices are going vertical because that's the bottleneck. If you want to look around the corner and see the next bottleneck coming, I strongly urge you to look at copper. Here we are: a supply shock meets a demand shock.

Commodity cycles typically last 15 years and have multiple hundreds of percent of upside. We're only a few years into this. This is just really getting started. I want to say one more thing. We spoke about demand—we're having this demand shock—and we spoke about supply. But what we haven't spoken about is how we're destroying the value of the US dollar.

Since COVID, we have absolutely destroyed the value of our fiat currencies. Today, we have $40 trillion of government debt that's growing by $2.5 trillion every year. On top of that, we have $100 trillion of discounted present value of future social liabilities: Medicare, Medicaid, Social Security, and pensions. That's also growing by $2.5 trillion a year.

So, you have $2.5 trillion of growth in the federal debt and $2.5 trillion of growth in the social liabilities. The US government only has $5.5 trillion of tax receipts every year. What's going to happen the next time we have a recession, when tax receipts go down and spending has to go up? We're going to print giga-dollars.

In the 1970s, we had this problem as well. The way we did it was we just debased the currency through some inflation and some growth, and the currency lost 70% of its purchasing power. Commodities, hard assets, and infrastructure will protect your purchasing power in that kind of environment. Go and look it up. In the 1970s, what was the best-performing asset class by a mile? That's your homework.

With that, thank you, and I look forward to chatting with you guys.

I think on the prediction show, you did—it was your call. I forget which category it was, but you definitely had copper as the top—

The best-performing asset was going to be copper, yeah.

Yeah. And that's before I talked to Dan Dreyfus.

That's right.

Which is saying something.

Well, I think the copper price is easily going to double from here. I've seen molybdenum go from $1 a pound to $33 a pound, so a double is no big deal.

Yeah.

Take a step back. You said something really interesting backstage, which is, if you look at everything that we're doing right now, we're barely going to keep up with just the natural energy demands of humanity. Just explain that thesis the way that you framed it in the back.

Here's the issue: we have not invested in upgrading, modernizing, and hardening the electric grid since post–World War II. We just let it go. The last 2, 3, 4 administrations were sleepwalking and haven't done anything to harden this infrastructure.

If we simply want to achieve our objectives to reindustrialize, reshore, and electrify—when I say electrify, that just means replacing your old gas boilers in these buildings with heat pumps, which every commercial building is doing—it means electric-car penetration going up. It means using your electronic devices more. Not even talking about AI. Not even talking about AI. We're going to have shortfalls just from that.

Just from living our lives. So, what happens? Blackouts, brownouts?

Blackouts, brownouts, and we're going to have to—

Prices.

Rising electricity prices. But you brought up a really great point on one of your shows, where you were talking about how the utilities are really goosing up the cost to do everything so that they can report to their regulator and earn that ROE on the higher capital base.

What's really interesting and underappreciated is that's where all the inflation is coming from. It's from the transmission and distribution from the utility. Power prices over the last 20 years, even after the rise we've just had, are still down—they're definitely down in real terms—but they haven't really gone up much in absolute terms. So, when you're talking about power going up—

Making it is still cheap.

But getting it is more expensive.

No, it's getting it to people.

Getting it to people, because the labor, by far and away, is the biggest bottleneck: craft labor, right? What do we tell all our kids to do? In the last 10 or 15 years, they go out and stand for you become a coder.

Northeast.

Yeah.

Right. Big mistake. So, I'm curious from the audience: in your homes, how many people have put up solar and/or Powerwalls? How many people have actually done that?

How many people, second question, are planning to do that in the next year or 2?

Okay, so that's another 20%. It's pretty obvious. This is obviously an affluent crowd. They are routing around the grid.

Is the solution to this energy independence in the home, in the business?

Great question.

Businesses are not waiting for the government. So, maybe the grid is going to be this weird, archaic infrastructure, and it's just going to be a ground-up solution.

Well, you're going to need the grid no matter what for industrial use. I mean, that's the foundation of industrial use. The scale of what we have to do just for industrial use—here's a good stat for you.

A 1-gigawatt AI factory, if you wanted to do it all with solar—and I'm a big solar bull, okay—solar's capacity factor is 20% because the sun doesn't shine all the time. With a capacity factor of 20%, a 1-gigawatt data center needs 5 GW of solar. Each GW of solar takes up 7,000 acres. So, at 5 GW, that's 35,000 acres. That's bigger than San Francisco.

Where are you going to find the people? That's the biggest bottleneck we have, by the way: craft labor.

Yeah. What about scarcity generally breeds innovation? I've seen some startups that are talking about new technology in mining to access what I think traditionally are rare earths. That's kind of the pitch. Everything we need is in the earth below us; it's just that we only mine the stuff that's on the surface. That's the general thesis.

Is there a set of innovations that you think are coming to market that are going to ultimately unleash more productivity than we see? We're still using the same technology we did 100 years ago to get this stuff out of the ground.

For some commodities, yes. You brought up rare earths. Coming out of the 14th century, there were these guys called alchemists. They said they could turn lead into gold.

Back then, the periodic table was just 4 elements: water, fire, air, and earth. Fire you could figure out what it was. The air was pretty pure. The water was pure. But every time they saw something in the earth they didn't know what it was, they called it a rare earth.

Rare earths are everywhere. The technology to extract rare earths is going to allow us to have a huge abundance of them. But the problem is processing them. That's the problem. The Chinese have all the technological know-how to convert what you take out of the ground into something that we can use.

There's always going to be some element of conversion that you're going to need. With something like copper, the market is so big that it's really difficult to find a technology that could solve that problem overnight.

If we're having this rivalry with China, and they are the provider, and that's the brittle part of the supply chain, we can solve the problem of job displacement—not apocalypse, displacement. People in America who want jobs: these are going to be incredibly high-paying jobs.

We can start bringing the fabs from Taiwan here, which we're doing. And we're going to bring both to North America and, I understand, South America, from a friend of mine who's got an automated mining system. We're going to be able to just create a large number of jobs here.

So, maybe you could talk a little bit about what impact this has. We keep talking about how behind America is, but what happens to China if we stop buying and start building?

What you said is very important for this whole jobs debate. The craft labor that we're going to need is going to be almost limitless for what we have to build, and there's really no other way around it.

In many ways, look what happened in the 2000s: we tore down all our factories and moved them to China. Who got killed by that? It was the blue-collar craft labor. It created all kinds of unintended consequences: fentanyl and wealth gaps.

Pennsylvania, Detroit.

The coasts were making all the money, and the heart of the country, the salt of the earth, was getting killed.

What’s ironic today is that that same part of the middle country—those people who got displaced—are now getting entry-level salaries. If you go to Quanta University and you’re at the top of your class, you’re starting out at $150,000 right out of high school. And the jobs they’re doing, ironically, are the jobs that may or may not be displacing some of the early, lower-level white-collar labor. So the tables have totally turned.

It’s an efficient market. The jobs are going to flow where the money is, and the money right now is really coming into this area.

Can we talk about a couple of other areas? What’s your take on other forms of energy—not gas, coal, nuclear, hydrocarbons? I mean, the demand pull seems like, based on this, maybe the most reductive takeaway is everything. But then how do you differentiate why you said, for example, you’re super bullish on solar? What are your thoughts on nuclear? How do you trade all these different sources of energy off?

We’re swimming in natural gas in this country. We can build solar; that’s not a bottleneck. And nuclear—we can’t really build it. We can’t even build the containment vessels in this country. The Koreans can do it, but we can’t do that here. So there are always going to be these big bottlenecks in the system.

Whether you’re talking about solar, natural gas, or uranium, we’re going to have the raw inputs, like the natural gas that we drill from the ground. But what we’re going to be short of is the critical minerals to build the nuclear power plants. We’re going to be short of silver, for example, to build these solar panels, especially if we start launching data centers in space. These are going to consume incredible amounts of silver.

Right now, the silver supply-demand dynamic is that we consume 1.2 billion ounces a year and supply 1 billion ounces a year. So there’s a 200 million ton deficit per year. And we only have 600 million ounces of above-ground inventory left. The clock’s ticking. We’ve got 3 years left, guys, before we stock out. Then the solar story is: where do you get the silver for the photovoltaic cells?

So for our kids and for the country, generation tool belt for us allocating, get some exposure to copper, silver, minerals, and then there’s a bunch of service providers in and around that area that we should be investigating over the next year.

Don’t forget the labor—the service providers. That’s a big one.

Okay. And how do you allocate capital? You’re at the front end of owning what? Mines and production, but then also the end-use cases? How do you decide where not to play? Because a lot of these things look like incredible end markets, but you can get run over. If you’re in the wrong part of the market, there are supply shocks, supply shaping by China, and price dumping. It can all look obvious, and you could lose a lot of money, too.

Yeah, look, you really have to understand supply chains. And I think, to a lot of people out there, supply chains are this sort of weird, mystical concept. I still think a lot of urban Americans think a ham sandwich comes from the refrigerator, and they don’t think about the 30 million pigs every month getting slaughtered outside of Chicago.

But you have to understand where the pinch points are in the supply chain, number 1. And number 2, you have to make sure that you’re not going to get technologically disrupted, where you can find something that’s going to replace that tightness in the supply chain.

Give it up for Dan. Well done. Very informative.

Thanks, bro.