How to Build the Next SpaceX or Tesla
Summary
- Defense demand can finance an industrial rebuild that serves much larger commercial markets. Erin Price-Wright’s premise is that “you don’t have a defense industrial base without an industrial base.” Early defense customers can absorb factory costs that commercial buyers would reject, helping establish capacity for products ranging from robots to hearing aids.
- Industrial AI gains depend on redesigning entire physical operations. Mariana Minerals owns assets through refining, while Hadrian improves “15 or 20 different things” across factory workflows. Connor Love’s boundary condition matters: “AI cannot turn a wrench yet,” and he says that capability remains far off.
- Flexible manufacturing can compress both product development and production ramp-up. Price-Wright frames the challenge of high-mix, low-volume manufacturing; Love describes technicians using Claude to reconfigure lines in 20 minutes, versus three months and hundreds of thousands of dollars through a systems integrator. The broader principle is “design for manufacturing”: prototypes and eventual production capacity develop together.
- Private capital is allowing defense startups to build before receiving formal requirements and funding. Love contrasts fragmented procurement with Castelion’s roughly three years of privately financed hypersonic-missile development before government contracts and purchases. Price-Wright says its team resisted product expansion: “No, we need to do this one thing really well.”
- Better domestic components could make new robotics markets financially viable. Price-Wright says defense production builds demand for the same motors, actuators and switches that robots need. Her concrete evidence is a portfolio company whose PCB acceptance improved from roughly 50% to 70%, 80% or 90% after switching to emerging suppliers.
- Capital availability is expanding faster than physical production can. Love says defense attracted more risk capital in the first half of 2026 than throughout 2025, which itself drew many times the amount invested in all previous years combined. Banks, private equity and credit firms are joining venture investors; Price-Wright says hundreds of billions, possibly trillions, are being directed toward the category over the next 10-plus years. Factories still take years to build.
- Power expansion ultimately depends on willingness to build and communities sharing the gains. Asked about America’s flat power production versus China’s new energy production, Price-Wright says, “I don’t think that is a technology problem.” Her question is “Do we want to grow?” Love adds that communities need to share in the economic upside of development.
Deep dive
1. Defense demand can finance a much broader industrial base
Price-Wright opens with the episode’s governing claim: “You don’t have a defense industrial base without an industrial base.” Decades of industrial hollowing now have battlefield consequences for American soldiers and allies. Elena Burger frames the AD50 around three connected layers—inputs extracted from the ground, factories making components, and finished systems—rather than treating defense manufacturers as a self-contained sector. Each layer constrains what the next can deliver.
Price-Wright says “dual use” can tempt startups to pursue too many things at once, even though the industrial base being rebuilt can serve many markets. Government buyers may pay costs that commercial buyers would reject, funding machinery and first-of-a-kind capex. Love illustrates this with a new U.S. PCB fab: a defense buyer can help pay for the equipment, while PCBs are also used in robots and hearing aids.
Price-Wright contrasts earlier government-led technology development—in space, the internet and nuclear energy—with private industry and capital now often providing the impetus. Government can still subsidize technical and process learning, while discerning American consumers push capability higher and prices lower than government buyers would. Love uses DJI to illustrate the transfer: after consumer drones, the company’s business expanded into public safety and national defense in the late 2010s. He says the people behind DJI are now the world’s largest robot producers, since drone motors are similar to robot actuators, which he calls the hardest part of building a robot.
2. Asset ownership can unlock full process redesign
Price-Wright points to SpaceX and Tesla as rare modern companies that have built at scale and as examples of vertical integration. She cautions against integrating everything, though: a unique, exquisite system such as a missile may require it. Love adds that SpaceX and Tesla built capabilities because the U.S. industrial base to support them did not exist, while high-quality suppliers should be used where available. Price-Wright notes Tesla had to recreate a battery supply chain; Love adds that its power electronics and drivetrain were built from scratch. Price-Wright says the underlying papers had been collecting dust since the 1970s.
Mariana Minerals illustrates why ownership can matter. Price-Wright says incumbent miners would not adopt technology quickly enough, so Mariana acquires and develops assets, mines and operates them, refines materials, and sells finished commodities across multiple minerals. Love says owning the asset can be necessary to change it. Price-Wright emphasizes the feedback loop: software across planning, operations, mining and refining captures data that feeds back into the process.
Hadrian supplies Love’s factory counterpart: improvement comes from “15 or 20 different things,” rather than one breakthrough. He cites quoting and accepting orders, making parts, inspection and movement between steps. Building factories from scratch—or taking over a mine—lets founders question whether practices inherited from the previous 100 years remain appropriate, rather than optimizing only one isolated task within an unchanged operation.
Love’s limit on the AI thesis is direct: “AI cannot turn a wrench yet.” He expects that eventually, but says “we’re far off.” Applying AI to industry therefore looks different from deploying it in legal work or customer service. Operators need deep knowledge of physical processes and how to manipulate them; software’s contribution sits inside that operational understanding, with robotics and autonomy still developing.
3. Flexible factories connect faster iteration with production scale
Price-Wright introduces the distinction between low-mix, high-volume production and high-mix, low-volume work. Love explains that traditional automation is expensive to reconfigure for varied products. Manufacturing technicians understand the process, and new software lets them reconfigure lines without necessarily having software-engineering training. Love’s example: a manufacturing engineer using Claude can make a change in “20 minutes,” instead of waiting three months and spending hundreds of thousands of dollars with a systems integrator.
Price-Wright says a more adjustable line makes it possible to prototype faster and then reach production rate sooner; otherwise, companies may have to rejigger subsystems after finalizing a production-level system. Love sums up the principle as “designing for manufacturing from the outset.” Price-Wright points to Anduril’s facilities in Ohio and Baltimore and Saronic’s shipyard, saying they are on a near timeline to produce thousands, tens of thousands and, hopefully, hundreds of thousands of systems. Burger notes that this is no longer theoretical: real products are coming off the lines.
Price-Wright treats repeat site visits as a way to observe execution, returning roughly every three months. Her image is an empty warehouse with epoxy floors, a flag, pallets and a founder excited about a rented forklift. Six months later, technicians fill the building and the company needs the city to provide more power because too many machines are plugged in. The transformation makes operating speed visible.
4. Building ahead of procurement rewards focus and operating experience
Drawing on his Army service, Love describes defense procurement’s disconnected decision makers: “the buyer is not the user,” and the person appropriating dollars for the buyer is further removed. Historically, primes waited for the government to specify and fund a system, producing products based on what was needed a decade earlier. Venture financing lets founders interpret military needs and build ahead of a formal request, aligning products with current technology and soldiers’ needs rather than waiting for the procurement sequence.
Love cites Castelion as an example: private venture capital funded its hypersonic-missile development before government contracts and purchases arrived. Price-Wright says the team spent 2.5 to 3 years focused on its first missile, resisting suggestions to build a small drone instead: “No, we need to do this one thing really well.” She connects that focus to SpaceX’s culture of engineers taking personal responsibility for whether their component works.
Price-Wright sees supplier startups emerging from problems founders encountered at Anduril, SpaceX, Tesla and Saronic. Existing relationships provide early customers willing to tolerate failures while solving their own difficult problems—much like YC companies selling to one another. Love adds that delivering the first product builds process knowledge, supplier relationships, design-for-manufacturing capability and customer trust, making subsequent products easier. But “it doesn’t work unless you deliver on that one core thing.”
Love describes Mariana CEO Turner traveling to remote locations without electricity or cell service and spending weeks or months away from family. The people founders engage with may have different timelines and cultures, and can be tough to negotiate with. Love says, “It sounds kind of sexy and romantic, but when you’re actually doing it, it’s rough.” Price-Wright adds that if founders are going to put themselves through that, “it better be damn worth it.”
5. Domestic suppliers can improve components and open new markets
Price-Wright argues that defense production creates demand for actuators, motors, valves and switches that future robots will also need. Improving their quality and cost could make previously uneconomic robotics applications viable. She estimates there is a “99% chance” that most components a robotics company buys today come from China. Love says companies want to bring purchasing onshore for quality, cost and speed of iteration, as well as geopolitical reasons, and notes how much of a supplier ecosystem has developed in just the last three years.
Price-Wright’s evidence is a portfolio company that, three years earlier, ordered twice as many PCBs as needed because it expected to reject 50% for quality problems. Emerging suppliers using software and modern manufacturing changed that: the company now sees 70%, 80% or 90% passing. Her conclusion is that close customer-supplier collaboration and co-development can materially improve quality, beyond simply relocating purchasing.
Love calls the next opportunity the “manufacturing middle.” His Boeing–Honeywell landing-gear example explains the obstacle: suppliers rationally repeat an existing design, potentially 50 years old, while lower-tier manufacturers remain in “receive mode.” That arrangement makes sense if the existing part meets requirements and suppliers are paid the same either way. But new systems and materials create reasons to rethink components. He expects suppliers to co-engineer with end manufacturers, pushing improvements upward through the chain. He calls this creating an “American manufacturing asset class,” distinct from buying factories and improving costs.
Price-Wright sees a parallel in electricity: mechanical infrastructure deployed in the 1970s and 1980s is giving way to power electronics, battery storage and more software-native architectures, opening the supply chain to redesign. Burger raises the possibility that demand may not grow fourfold and argues that even delivering half the anticipated requirement could start a snowball. Love agrees that “you’re not going back”; Price-Wright adds that a new architecture can bring better cost, efficiency, serviceability and ease of manufacturing.
6. Capital is following evidence that physical businesses can become valuable
Love’s distinction is that these companies can now be among the most valuable, as well as the most important. He cites Palantir, Anduril and SpaceX as proof that companies in these categories can become very large, and says most of the world’s ten most valuable companies are hardware companies. Burger quips that some of these overnight successes took 20 years.
Love argues that America’s private capital markets can follow entrepreneurs and demonstrated success flexibly, without a top-down-directed industrial policy. He cites JPMorgan committing trillions toward investing in the physical world and growing participation from private equity and large credit firms. Capital-intensive factories and systems require that broader financing stack. Price-Wright says hundreds of billions, if not trillions, are being directed toward the category for the next 10-plus years.
Love says more defense risk dollars went into the ecosystem in the first half of 2026 than in all of 2025; he says 2025 itself was many times the amount invested in all previous years combined. He also says the financing “valley of death” around contracts and capex is starting to disappear with capital-market support. That change extends beyond defense into energy, robotics and critical minerals.
7. Expansion requires patience and communities participating in the upside
Burger asks what would close the gap between essentially flat U.S. power production and China’s rising new energy production. Price-Wright answers, “I don’t think that is a technology problem.” She frames it as a question of national willingness: do Americans want conforming projects built, and “Do we want to grow?” If so, she believes technology and capital will follow.
Price-Wright says the snowball has started to roll, but asks whether emergent winners will bring more capital into the ecosystem. Physical projects require moving dirt and understanding supply chains, so “these things take years”; she says capital markets need patience. She is optimistic, but warns that government and national choices can obstruct the process: “it’s ours to lose.”
Love says physical projects make economic participation important: unlike software’s more abstract benefits, a mine, data center or power plant changes surrounding communities. Residents have “a say” and “a stake.” He expects companies to figure out how to bring communities along and make participation economically worthwhile, even if the transition is short-term painful.
Price-Wright gives AMA as an example of a company partnering with small machine shops, working with them to build new products rather than cutting management and firing people. She says companies need to work with communities rather than against them. She expects AD50 companies to bring jobs and prosperity to Middle America; Love adds that broader growth could reduce wealth concentration in elite coastal pockets.