离开 Tesla、重建美国的创始人 | a16z

The US is 50 years behind on critical mineral supply. We are too slow at designing, building, and ramping up new minerals capacity, even after we have a license to operate.

Even though there's so much innovation happening at the edge of the grid, on the other side of the wire, there's really been no change.

You both came out of Tesla. What does the Tesla model give you that a traditional industrial company doesn't have?

The belief that you can innovate on systems that are old and archaic. If the outcome is worth it, Tesla will fight through the challenges of getting to that outcome.

We're making a big bet on autonomy and refineries, where we use reinforcement learning to actually remove humans from the loop in determining how refineries operate.

The world's leading producer of silicon carbide, which is a key power semiconductor, is based here in the US, and so we should be leveraging the applications of that technology here first, manufacturing here at home.

Now, it's tempting to talk about the AI race as a competition of models and chips. But the truth is that AI dominance and reindustrialization more broadly are physical projects. They are energy projects. They are mining and refining projects. They are manufacturing projects. They are grid-scale projects.

Every breakthrough model, new factory, and autonomous system that we'll talk about here today has a real-world requirement underneath it. They require materials, energy, and the ability to move electricity where it's needed, when it's needed.

We increasingly hear concerns that AI will put an undue strain on an already faltering grid, will demand more energy than we can give, and will require more build-out than we can keep up with. In many ways, these are fair concerns. But rather than taking this at face value and putting our pencils down on progress, we see this as a call to action—an opportunity.

We can do great things in this country. We have rallied around national projects before, accomplished things few dreamed possible, and we can do so again. This is the next chapter of American dynamism.

If we want to rebuild the industrial backbone of the United States, we have to rethink the entire stack, from critical minerals to energy generation to transmission, as well as how we build and interconnect new infrastructure at the speed that it's needed.

This next conversation brings together 2 incredible entrepreneurs building across that stack to talk about what it will take to do just that. Please join me in welcoming co-founder and CEO of Mariana Minerals, Turner Caldwell, and founder and CEO of Heron Power, Drew Baglino.

We'll spend a lot more time in this room today talking about AI, but the constraint on America's AI future—and, as I mentioned, on reindustrialization more broadly—is, in many ways, atoms and not algorithms. You 2 are both building fundamental pieces of the physical infrastructure that the future AI economy can't live without.

Maybe, just to get the audience started, you could briefly explain what you both build and why these physical industries matter.

Mariana Minerals is a software-first minerals mining and refining company. When I say software-first, what that means is that about a quarter of the company consists of software engineers and machine-learning engineers who are developing 3 core operating systems to accelerate project delivery and increase the amount of autonomy that we see in minerals operations and refining operations.

Capital Project OS is basically a product lifecycle management tool, is the way to think about it, but that goes from process development and mine development all the way through engineering, construction, and procurement, with agentic workflow automation through that stack. Plant OS is how we use reinforcement learning to control refineries. Mine OS is, again, how we use reinforcement learning to do short-interval autonomous control of mining operations.

But we do not sell software. We are not a SaaS company. We develop, engineer, build, and operate minerals projects. We have a copper mine operating in Southeast Utah that's producing high-purity copper materials today. We're building a lithium refinery in Texas, with the goal of building 10 projects in the next 10 years.

How about you, Drew?

First, thank you for having me here today and representing the Heron Power team. At Heron Power, we build power electronics to accelerate the electricity sector.

Over the last 4 decades, in parallel with the improvements brought by Moore's Law in transistors and computing, there's been a similar improvement in power transistors. Over those decades, it's enabled more and more applications. We see them in how we charge our phones, in telecommunications, and in data centers.

But really, that improvement hasn't been brought to the grid itself. In a time of growing demand for electricity for so many different reasons—all of them positive—and given that electricity growth and energy growth are correlated with economic growth and prosperity, we need new solutions. Luckily, the power semiconductor space is ready to bring those solutions, and I'm excited to do that.

At Heron Power, we're focused on building solid-state transformers to use silicon and software to replace steel, oil, and copper in power conversion at data centers, large-scale energy installations like solar and battery projects, and others.

Amazing. The US government has made it a pretty clear and loud priority to reshore critical supply chains, from critical minerals to advanced manufacturing, and there's also been a lot of focus on the AI race against China.

In plain terms, where does it leave us if American companies like yours don't exist and win?

I can take that first. This power semiconductor capability that's enabling solid-state transformers is actually the outgrowth of many decades of partnership between the federal government, academia, and industry. Both the DOE and the Navy have focused a lot on advanced semiconductors.

It just makes sense that the place where this technology was first developed should be the place where all the benefits are commercialized. The world's leading producer of silicon carbide, which is a key power semiconductor, is based here in the US, and so we should be leveraging the applications of that technology here first, manufacturing here at home.

If we don't, we're basically losing all the benefits that accrue from that technology to other countries. I don't think we should do that.

Put plainly, the US is 50 years behind on critical mineral supply. If we're not innovating in the critical mineral space, we will be perpetually behind.

And when you say behind, do you mean specifically behind China?

Specifically behind China. But I would say also globally, we have a couple of decades of lag.

The things that we can do at the top level are accelerate permitting, and we can make project-level finance more available. But that doesn't actually solve the underlying problem, which is that we are too slow at designing, building, and ramping up new minerals capacity, even after we have received a license to operate.

Mariana's laser-focused on that phase of project development. You have to get things permitted, yes, but once you start building, it can take 5 years to get something built, and then it can take 3 to 5 years to get something actually operating at rate.

That's why we're laser-focused on that, so that even if we start to lower the burdens to play catch-up with China, we actually have to go faster than China does.

You both spent a long time at Tesla. Drew, you spent 18 years at Tesla. You're something of a deity among power electronics nerds, I would say.

Now you work on grid-scale power systems, which is different. When you looked at the grid, what convinced you to leave Tesla and tackle this seemingly unsexy problem in this way?

And, adding on to that, what does it actually take in terms of time, cost, and regulatory hurdles to do this in the US? Big questions there—2 questions. I apologize.

I had a front-row seat to an amazing set of impactful innovations at the grid's edge: EVs becoming more affordable, not just more affordable but more omnipresent around us; building the Supercharging infrastructure to support those electric vehicles; and then working on grid storage. I was responsible for the Megapack and scaling the energy business at Tesla.

All along the way, what I saw was that even though there's so much innovation happening at the edge of the grid, on the other side of the wire, there's really been no change. The systems underpinning the grid today are largely the same mechanical systems that were developed over 100 years ago.

You don't get control, you don't get monitoring, and you end up with an overbuilt system that is fragile. Also, there aren't a lot of suppliers providing that equipment, and most of them are actually headquartered overseas. That just doesn't seem like a secure position for such critical infrastructure for us to have here in the United States.

I think you can. I built the Megafactory with my team in Lathrop, California, in 11 months. It was a JCPenney warehouse; 11 months later, the first product came off the line.

Ultimately, what it comes down to is alignment.

You know, when you're working with your local jurisdiction, they can use the process for a code-compliant project to say no at every step, or they can say yes at every step. So how do we, as a collective, gain alignment that building and reindustrializing the U.S., building critical infrastructure, and supporting our critical supply chains here in the U.S. is a good thing? How do we identify ways to say yes at every step along the way and really accelerate these processes, versus saying no? When you do find that, it can be magical. That's been my experience.

And I know in particular you've talked—we've talked—about labor costs and labor shortages. Oftentimes, people point to that as the reason why they can't get things done in the U.S. But what's your experience?

Yeah, today's factories are really automated. If you're building a new factory today in China or the U.S., the labor differential is less than 10% of cost of goods sold. It might even be less than 5%. What is actually driving the competitiveness of the different locations, in my mind, comes down to supply chain. And how do we develop co-located critical supply chains in the United States where the logistics costs are much much shorter, and much much lower because the logistics time is much much shorter.

If you look at China, they are so thoughtful about building these industrial areas. Everything that you could possibly need to build a car, which has 7,000 parts in it, is within less than a 3-hour drive. Getting to that kind of co-location of the supply base in the United States would be a major unlock, along with automation, while still providing immense numbers of high-paying, important jobs. I think that's a vision that I'd like to advocate for.

Yeah, and when we're talking about jobs within factories, this isn't your grandfather's or great-grandfather's supply chain assembly-line factory floor. These are technical jobs. They may require training, but they have skill and the pay associated with that.

100%. Okay, Turner, the U.S. government ranks onshoring critical minerals as essential to economic and national security. It's been in the news a lot for the last year: rare earths, critical minerals. These things feel very bottlenecked. So much of the processing capacity for these materials sits overseas, especially in geopolitical rivals, namely China. Given that vulnerability, how does Mariana Minerals' work help the U.S. reclaim not just the extraction side—the actual mining—but processing and supply-chain sovereignty for critical minerals?

Yeah, we focus on the full chain, from mining all the way through refining. You have to focus on the full chain; that handoff in the middle leads to a lot of actual market inefficiencies. We're making a big bet on autonomy, fundamentally. We're making a big bet on the fact that we can build systems that enable us to engineer things faster using large language models, accelerate the procurement life cycle, and do autonomous short-interval control of construction operations, where you're really doing resource balancing between what materials you have on-site, what your list of tasks is, and what people you have on-site. That kind of optimization is all something that can be done algorithmically.

We're also making big bets on autonomy in refineries, where we use reinforcement learning to actually remove humans from the loop in determining how refineries operate. When you have a highly variable feedstock, because the earth is heterogeneous, you need to constantly tune the temperatures, flow rates, chemical addition rates, and residence times of a highly complex refining circuit. We don't have the labor pool here that has the embedded know-how to walk up to a refinery, quickly get it operating on-spec, and then manage that variability.

The same is true on the mining side of things. At those mine sites, we're making thousands of decisions a day. When you don't have a labor pool of folks who are able to make the right thousand decisions, that can cascade into low productivity, low availability of equipment, and low utilization of equipment.

The software angle is really not enough, and we talked about how we're vertically integrated. What sets the rate of software and technology penetration in these plants and mines ultimately is the operating teams. What is the tech stack that they're comfortable with? For the most part, it is pen and paper and maybe 150 spreadsheets scattered around an operation.

What you need to do in order to actually accelerate software uptake in this space is go down into that operating layer and understand the core problems that they're facing. You also have to really control the culture and make sure that the software tools themselves are designed for the folks who are going to have to interface with them. That's why we think that sitting the software engineers right next to the operating teams—but not in a forward-deployed-engineer-type way, where everyone has the same incentives—is what's going to yield the best results when it comes to trying to optimize these assets.

Now, you both, as I mentioned before, came out of Tesla. It's one of the companies that proved the template for American dynamism and success, building factories in America for the first time in a long time. Turner, you led Tesla's minerals and metals team. Drew, you ran powertrain and energy. What does the Tesla model give you that a traditional industrial company doesn't have? What is genuinely different, maybe, about building in your new respective sectors that you didn't expect relative to Tesla?

I'd say there are 3 big ones, and they might overlap a little bit with Drew's. I think the general techno-optimism and belief in what technology can do in these sectors is much higher at Tesla. The belief that you can innovate on systems that are old and archaic is at the core of the company.

The other is a general appetite for risk, which enables super-fast decision-making and enables the teams to move really quickly without being burdened by fear of making the wrong decision.

The last is a clear, firm commitment to not giving up on projects when the outcome is worth it. Tesla will fight through the challenges of getting to that outcome. What we see, at least in the minerals industry, is that folks will give it a shot for a year. People have tried to do autonomy in mining for a long time, and generally a lot of companies will fail, put it on the shelf, or isolate it into a small team that doesn't get tapped again. Tesla does a really good job of barreling through the challenges as long as the outcome is worth it.

I would add a couple more aspects to it. Many times in Tesla's history, the company's future success—whether or not the paycheck will clear—was bet on the team within the company executing well. That is a very focusing reality. It drives people to do their best work, and you end up needing to manifest that outcome. I hate to say “do or die,” but it's equivalent to that. That's something that exists uniquely within startups. Turner and I are bringing that to our own teams, but it wouldn't be in a legacy industrial company.

The other thing is that there was always a clear vision of the purpose of the company. That's a beacon for talent, right? People think, “I want to work on that. That sounds amazing.” So you get to pick from the best already. Then you're in this high-growth environment. Anybody who is excited about their career trajectory and having it trend in a good direction is also going to want to work there, and is also going to want to stay there and see it through, because their impact is real. They see the impact of their actions on the outcomes around them, and those outcomes result in their own growth. They move from one part of the company, like Turner did, to another—or like myself, given my career history.

I think those are in stark contrast to a multiproduct industrial conglomerate that's selling the same thing today that it was selling decades ago, or a mining company that's got 150 years of heritage. That's a hard thing to replicate outside a startup, and it's hard to maintain within a startup, but I think it's really important to getting things done.

Yeah, totally. Talking about getting things done, building that team, and being able to hire, one thing that jumps out is that both of your companies are building real facilities that will create real jobs. Turner, your initial lithium and copper projects should add over 500 construction jobs and additional full-time jobs in the next 18 months, with many more as you scale operations. Drew, Heron is getting ready to build out its first large factory, which should also be something around 500 jobs, and that's just the first factory of many.

What have you both learned about building an industrial workforce in the US in 2026?

I think you have to be creative here in the US. We are reindustrializing, and I can't just go to a talent pool of power electronics manufacturing engineers or production associates. In my background, I was responsible for building, along with my team, the 4680 program manufacturing facility—a 50-gigawatt-hour battery facility in Texas. At that point, there were really not a lot of battery operations in the United States.

So instead, you have to look for analogs. I was hiring people out of high-speed bottling plants and syringe manufacturing facilities where they're making billions of syringes. If you can get that creative hat going, you find that there's immense depth of talent in the US. People are excited to work in new industries, and you build that shared vision of the future. I'm very positive about what you can get accomplished here.

I would say that looking at analog industries is a great point. For the mining industry, we're in a similar position where we've had 35 years of meaningful attrition in the labor pool. But the oil and gas sector has a bunch of extremely good talent. In the software space, a lot of the underlying optimization algorithms that we're writing for our plants look very similar to the optimization algorithms in dog-walking apps, Uber ride optimization, underwriting loans, and ad optimization.

There is transferability in the broader US talent pool. What's important is building that talent magnet. It's an interesting one for us because the mining industry, like the villains in every movie, is the resource-extraction sector. We have to combat that and say—

Make mining sexy again.

That's right. That's right.

Yeah.

Finally, I know we're out of time, but if you had one specific and actionable ask for the people in this room that would materially speed up production, onshore manufacturing, and create jobs in the next 12 to 24 months, what would you say? You have the floor.

I'll go first. I think if we have a minerals mandate, what we should do is look at everything that was done in the last 50 years for oil and gas, when we had an energy mandate—and we still have an energy mandate. That's a lot of asks boiled down into one, but there are a lot of tools in the toolkit.

I think the most important thing is providing the right incentive structure that mobilizes the private capital markets behind these projects, so that they're confident there is a market in the long term and that the rug isn't going to get pulled out from under them in an industry that, for the last 30 years, really hasn't been built out in the US.

Yeah, I think durable industrial policy that you can plan around. I'm very pro-manufacturing in the United States, building these technologies in the United States, but my suppliers, maybe my financiers, are not as certain. So, durable industrial policy driving in this direction.

I think a concerted effort between the federal government and the states to identify areas of energy and manufacturing buildout, so you can get those colocated supply chains that I mentioned before, would be major—where the local jurisdictions are getting to yes with you rather than trying to find ways to say no all the way along the project.

The last thing is, I'm a fan of the electricity sector. I think it's enabling so much growth. I like the idea of a federal highway trust fund for the grid. It never has existed. That's sort of why we have this patchwork. How do we find a master plan of buildout of linear infrastructure that maybe connects those manufacturing and energy buildout zones to improve resilience, reduce costs, and really move us forward as a nation?

Awesome. Thanks so much. You heard it here. Great thing.

Thank you.