"It's not just that you get a factory a building with some operators that they go and recruit. It's you get a full engineering team that co-designs the product for you." - Sam De Miko (Discussing the reality of modern Chinese contract manufacturers) [00:08:22](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h8m22s)
"There is a reinforcement learning with factory feedback step that unless you build stuff at high volume you will never achieve." - Sam De Miko (Explaining why theoretical design is useless without scaled physical production) []()
Disclaimer: Orignal content owned by or sourced from third parties. It does not represent the views of 'Nuggets' platform or it's team. AI is used extensively across this platform including for summaries. Accuracy is not guaranteed, there can be mistakes. Any info or content on this platform is not a financial, legal, or investment advice. Do your own research. Refer for complete disclosures:- Terms of Use · Full Disclaimer
"We have only so much tribal knowledge left... once it's gone it's gone. And preserving tacit knowledge is very hard because it's not like the matrix where you just jump in and download I know kung fu." - Aaron Slodov (On the urgency of the Re-industrialize movement) [00:12:13](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h12m13s)
"We built up scaffolding of environmental laws from the 60s onward that have basically made it illegal to do stuff... California bit by bit has restricted what I would basically call chemistry at like a foundational level where it's like basically illegal to do chemistry with metal." - Sam De Miko (Discussing regulatory roadblocks to the electric stack) [00:21:41](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h21m41s)
"Less than 5% of everything in defense is produced at higher than 10,000 of anything per year. And this is actually going to cause reinforcement learning into stupidity." - Sam De Miko (Critiquing the lack of volume in defense procurement) [00:42:41](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h42m41s)
"Yes while you're exporting files, the factory has been micromanaged by Apple to such a degree that it is effectively an Apple facility." - Sam De Miko (Clarifying the myth that Apple just sends blueprints to China) [01:01:07](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=1h1m7s)
"This is the way that advanced hardware is made today, you can choose to participate in that world or you can go Ted Kaczynski and live in a cabin in the woods." - Sam De Miko (On the necessity of engaging with China for hardware startups) [01:16:11](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=1h16m11s)
2. Executive Summary
The United States has experienced a massive loss of manufacturing capability over the last 40 years, outsourcing the crucial, high-volume repetition required to maintain physical "tacit knowledge." China has utilized this massive volume to evolve from low-end assembly to leading global R&D in advanced electronics, batteries, and automation.
To reverse this trend, the US must overcome systemic bottlenecks—such as over-regulation, NIMBYism, and misaligned capital markets—by capitalizing on high-volume commercial demand (like the AI data center boom) to forcefully onshore the fundamental "electric stack."
Volume dictates capability: You cannot retain manufacturing prowess through low-volume or purely digital design; factories require millions of repetitions (reinforcement learning) to optimize tacit processes.
The Electric Stack is the new baseline: Future hardware (from EVs to robots to stoves) relies on a uniform foundation: batteries, power electronics, embedded compute, and electromagnetics. Mastering these components locally is a national imperative.
Regulation has banned core foundations: Decades of localized environmental rules in places like California have effectively banned "chemistry with metals," making it legally impossible to build vertically integrated electric supply chains in the US's prime engineering hubs.
Venture Capital mismatches hardware realities: Startups go to China because Chinese CMs offer "fractional capacity" and deep engineering teams. US capital markets and banks are no longer structured to fund the massive upfront CapEx required to build isolated factory ecosystems from scratch.
Defense procurement lacks the volume to save US manufacturing: Because the DoD rarely orders components in volumes exceeding 10,000 units, defense contractors act more like consulting firms than mass-manufacturing powerhouses, failing to build the "muscle memory" required for commercial scale.
Data centers are the Trojan Horse for re-industrialization: The current massive demand for AI data centers (which utilize the full electric stack) presents a golden, un-tariffed opportunity to mandate and onshore advanced US manufacturing capacity if capitalized on correctly.
The US's global share of manufactured goods plummeted from 30% in the 1980s to roughly 16% today, with China capturing the difference. Aaron Slodov explains that this isn't just a loss of jobs, but a loss of "tacit knowledge"—the undocumented, tribal understanding of how to physically make things.
Because manufacturing processes are incredibly stateful and impossible to perfectly simulate, they require constant, high-volume repetition to perfect. By offshoring assembly, the US inadvertently offshored its engineering feedback loop.
Sam De Miko outlines his time at Google Glass and Oculus, noting how China rapidly evolved. Initially, the Oculus DK1 was built in a damp "toy factory" that previously made Guitar Hero plastics [00:07:03](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h7m3s).
Within years, Chinese factories transitioned into advanced engineering powerhouses. Today, hiring a Chinese CM is like accessing a hardware LLM: you provide high-level design input, and an embedded team of engineers leverages an immense, localized supply chain to co-design and scale the product for you [00:08:49](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h8m49s).
To rebuild capacity, the US doesn't need to make everything; it needs to master the "Electric Stack" (lithium-ion batteries, power electronics, embedded compute, electromagnetics, and vision hardware). However, placing factories in the US is hampered by deep regulatory issues.
In California—home to the best engineering talent—decades of reactionary environmental laws have effectively outlawed basic "chemistry with metals" (e.g., anodizing, battery formulation) [00:21:41](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h21m41s). This forces companies to either offshore or break their supply chains across state lines, breaking the vital link between the engineering office and the factory floor.
While companies like Anduril are celebrated for boosting US defense tech, the defense sector cannot independently save US manufacturing. The DoD rarely orders anything (besides bullets/shells) in volumes over 10,000 [00:42:41](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h42m41s). Without consumer-level volume, US suppliers become bloated "consulting firms" utilizing obsolete methods because they lack the throughput required to drive process innovation.
Conversely, the massive commercial demand for AI data center racks—which are highly advanced electromechanical systems—represents a massive, untapped opportunity to spin up high-volume US supply chains [00:31:14](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h31m14s).
Founders building hardware (like Impulse stoves) are practically forced to manufacture in Asia due to Capital limits. A startup cannot justify buying a million-dollar injection molding machine to let it sit idle 75% of the time.
China offers "fractional capacity"—the hardware equivalent of AWS—allowing startups to pay only for the machine time they use while leveraging local expertise. Until the US recreates these fractional ecosystems, or changes the way CapEx is funded, reshoring will remain economically unviable for new entrants.
The ultimate test of physical manufacturing supremacy will be humanoid robots. Because humanoids require a completely new, massive supply chain of advanced actuators and batteries, the race to scale them is critical.
The host and guests suggest that whoever scales humanoids first will essentially build a "training ground" for physical AI that self-perpetuates [01:08:50](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=1h8m50s). If the US relies on China to build the physical bodies for its AI brains, it risks handing over the ultimate leverage in the next era of automation.
6. Data & Figures
Data Point
Value
Context
Timestamp
US Manufacturing Share (1980s)
30%
The percentage of the world's manufactured goods produced by the US in the 1980s.
Tesla's 747 Bumper Delivery [00:29:04](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h29m4s): To highlight how critical low-level components are to final assembly, Aaron notes that in its early days, Tesla had its massive bumper injection molds cut in Japan and flew them to California on a 747 just to shave two weeks off production timelines.
The Oculus DK1 Toy Factory [00:07:03](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=0h7m3s): Sam recalls that when Oculus was acquired, its dev kits were being built by the exact same Chinese contract manufacturer that made the plastic guitars for Guitar Hero and Rock Band, illustrating China's rapid leap from plastic toys to spatial computing hardware.
Google Glass's False Start [01:04:15](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=1h4m15s): Google attempted to onshore the manufacturing of Google Glass in a San Jose facility to prove a point to Apple. However, Sam notes that while final assembly happened in California, he still had to constantly fly to China because the core components (camera modules, flex circuits, plastics) had no local supply chain.
Apple Exporting Engineers, Not Files [01:00:06](https://www.youtube.com/watch?v=TRcvUUMaBcA&t=1h0m6s): A leaked image showed Apple was United Airlines' top customer, spending roughly $100 million primarily on flights to Asia. Sam uses this to debunk the myth that Apple just "emails blueprints" to Foxconn; Apple physically exports armies of engineers to micromanage the factory floor.
Concept: Physical manufacturing cannot be fully simulated. Process engineering parameters (like chemical laminations or machine heat settings) act like weights in an AI neural network.
Application: You only discover the correct "weights" through the high-volume throughput of mass production. If you don't scale to millions of units, you never learn how to actually build the product optimally.
Concept: The modern foundational layer of hardware consisting of 4-5 pillars: lithium-ion batteries, power electronics, embedded compute, electromagnetics, and computer vision hardware.
Application: Rather than trying to reshore every random product, the US should focus on subsidizing and building high-volume capacity exclusively for these core pillars, which power everything from drones to EVs to induction stoves.
Concept: Just as software startups rent fractional server space on AWS rather than building server farms, hardware startups need to rent fractional time on expensive manufacturing machines (like injection molders).
Application: China dominates because its CMs offer this fractional capacity alongside embedded engineering, a model the US lacks due to different capital market incentives.
Companies/Entities:BYD (used as the ultimate example of co-locating engineering, factories, and worker housing), Starlink, Atomic Industries, Impulse, Anduril.
10. Speakers & Credentials
Aaron Slodov: Founder building Atomic Industries. Background in physics and engineering. Early engineer at Waymo. Started the "Re-industrialize" movement after facing immense supply chain roadblocks while trying to build consumer hardware stateside.
Sam De Miko: Founder of Impulse (building advanced battery-integrated induction stoves). Background in electrical and software engineering. Former engineer at Google X (Google Glass) and Oculus/Facebook, giving him a decade of firsthand experience scaling consumer electronics in Chinese factories.
Host ("Relentless"): Facilitates the overarching timeline and prompts the guests on macroeconomic and industrial trends.
11. Actionable Next Steps
Embrace the Data Center Opportunity: Policymakers and industrialists should leverage the massive CapEx currently pouring into AI data centers to mandate and onshore the production of high-end power electronics and server racks in the US.
Align Housing with Industrial Zones (Industrial YIMBYism): To make US factories viable, municipalities must streamline permitting for dense, affordable worker housing immediately adjacent to manufacturing sites, mimicking the BYD campus model.
Learn Mandarin & Go to Shenzhen: For founders and engineers looking to build advanced hardware, visiting China to witness the speed, scale, and tacit knowledge of modern CMs is a necessary rite of passage to understand the physical world's baseline.
Reform Local Environmental Scaffolding: State governments (especially California) need to review and modernize 1960s-era environmental laws that inadvertently ban core industrial processes ("chemistry with metals") needed for the green tech transition.
Full Episode: The AI Industrial Revolution | 2 Jun 2026 | Naval and Nivi
Context: Host Naval Ravikant introduces a roundtable discussion on the "AI Industrial Revolution" with three frontier deep tech and software founders who build their own physical factories and tech infrastructure from first principles rath…
Hyperscaler Data Center Demand
10,000s of units
Volume of data center racks ordered by hyperscalers, each costing $3M-$4M.