Public transcript

Palmer Luckey on Building for the Future: Oculus, Anduril, and The Future of American Power

By: a16z crypto
Length: 1 h 1 min
Language: English
Text from: the video's captions
Transcribed: 13 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

Palmer Luckey joins a16z for a wide-ranging conversation covering his Oculus days, the Facebook acquisition, why he believes VR is succeeding despite press narratives, and the founding and growth of Anduril. He discusses US-China rivalry, defense manufacturing constraints, his new narrow bank supporting stablecoins, and his optimistic thesis that automation will make goods as cheap as food and clothing are today. He argues much current progress is psychological and regulatory catching up to existing science rather than new breakthroughs.

Key points

Luckey says Facebook's commitment of a billion dollars a year in R&D for 10 years was the decisive factor in selling Oculus, not the price.
He argues VR is not failing: Quest 2 sold 20 million units, outselling the Nintendo 64's first three years, but is unfairly compared to smartphones.
His 'free isn't cheap enough' essay argues VR's obstacle is quality, weight, and content, not price — even free headsets would lose most users within a month.
Anduril's 'don't work at Anduril' recruiting campaign tripled incoming recruitment by attracting mission-driven people and repelling the wrong fits.
Anduril runs about 25 products with lean, engineer-led teams, roughly 85% engineering, and every product older than 5 years has scaled.
He says the US can't match Chinese manufacturing costs, and US rules banning Chinese components plus anti-tamper requirements make his drones costlier than Ukraine's Chinese-parts alternatives.
His new bank is a narrow bank — one-to-one reserves, stablecoin support, and a 50% loan-to-deposit ratio — designed to survive scenarios the government can't or won't bail out.
His optimism thesis: automation that made food and clothing nearly free will spread to cars, housing, and everything else, with regulation as the main bottleneck.

Questions it answers

00:00Oculus origins and VR hardware

How did Oculus get started and what made VR hardware so hard?

Luckey started building headsets at 15; DK1 sold 55,000 units with only rotational tracking, and DK2 added six degrees of freedom. Real rooms — glass, reflections, moving curtains, ceiling fans — made inside-out tracking a Herculean challenge, and Samsung was the only viable display supplier, forcing the Gear VR partnership.

Cyber sickness comes from inner-ear/eye mismatch, which the body interprets as poisoning.
Dead pixels were so costly that Oculus tested every display; one asymmetric pixel stands out like a cross-eye puzzle.
Gear VR sold over 10 million units but was 3DOF, which Luckey calls 'not the real thing.'

08:04Bitcoin and the Facebook sale

Why did Oculus sell to Facebook, and what role did Bitcoin play?

Luckey was an early Bitcoin user, nearly Coinbase's first customer, and wanted the Oculus store to accept Bitcoin before lawyers killed it. The team turned down a ~$1 billion offer partly because Luckey didn't care about money, and accepted when Facebook committed a billion dollars a year in R&D for 10 years.

His Forbes 30 under 30 advice in 2014: 'Buy more Bitcoin.'
Raising billions in VC would have cost him control; the Facebook deal guaranteed funding.
He frames it as Oculus taking over Facebook, which has since spent a public $60 billion.

13:57VR success and 'free isn't cheap enough'

Is VR actually failing as the press claims?

Luckey argues no: Quest 2 sold 20 million units, outselling the Nintendo 64's first three years, and total VR users are in the low tens of millions — comparable to early iPhone adoption. His essay argues price isn't the blocker; quality, weight, and content are, so even free headsets wouldn't retain users.

Quest at $399 hugely outsold the $99 Oculus Go.
He's praying Apple and Samsung's high-end pushes work or VR faces another 'winter.'
The anti-tech press compares VR to smartphones, the biggest computing shift in history.

18:58Anduril culture and product strategy

How does Anduril recruit, organize, and ship so many products?

The 'don't work at Anduril' ad tripled recruitment by attracting mission-driven people and repelling the wrong fits. Luckey runs ~25 lean, engineer-led product teams (~85% engineering, little middle management), which he says is easier than one giant product serving everyone, and rapid growth creates leadership roles for proven engineers.

Every Anduril product older than 5 years has scaled.
He keeps up with technology by reading academic journals as an outsourced research team.
Anduril has been doing AI for ~9 years — the company's acronym is literally AI.

32:58US-China rivalry and defense

How does Luckey see the US-China technology race?

Anduril's 2017 deck predicted Russia invading Ukraine and warfare shifting to cheap, proliferated systems. He says China has won cheap manufacturing, so the US must lean into advantages China lacks: capital markets, culture, and allies worldwide. He praises Putin's 2007 AI quote as Bond-villain-level foresight from a competent adversary.

You can't start building bombs after the war starts and expect deterrence.
The cynics who said war never ended were right.
The US should re-industrialize but can't win on that alone.

39:46Narrow bank and stablecoins

What is Luckey's new bank and why did he build it?

It's a narrow bank — one-to-one reserves, not fractional — supporting stablecoin deposits for 24/7 settlement, with a 50% loan-to-deposit ratio as its riskiest product. He built it after SVB's collapse showed hidden systemic risk, aiming to be a survivor in a crisis the government can't financially or politically bail out.

A previous narrow-banking attempt was blocked by the government.
The stablecoin bill debate includes attempts to ban paying interest/rewards to customers.
The banking system only works if everyone 'stays in the matrix.'

46:11Just do things and optimism

Why is Luckey optimistic about the next 20-30 years?

He argues much current progress is psychological and regulatory catching up to decades-old science — nuclear energy, oxytocin applications, bioscience. His thesis: automation that made food and clothing nearly free will extend to cars (~$1,000 F-150 equivalent) and housing, with regulation, not technology, as the bottleneck.

One hour of federal minimum wage buys a week of calories in oats.
Anti-manufactured-housing laws are 'one of the most anti-American sets of laws.'
Oculus almost bought 51% of Nvidia when it was worth ~$7 billion.

53:53Ukraine Q&A: cost and constraints

Why is Anduril's gear expensive for Ukraine, and can it scale?

Luckey says China has won cheap manufacturing and US rules prohibit Chinese components, approved-only navigation, and anti-tamper requirements, making his $2,000 drone uncompetitive with Ukraine's $900 Chinese-parts drones. Government-owned artillery plants block private competition, though he's pushing administrations to commit to scale contracts awarded on price and volume.

Zelensky wanted Anduril border towers before the war; the State Department blocked it.
Anduril towers cover ~30% of the US southern border and 35 bases worldwide.
China sanctioning Anduril's supply chain 'cripples my ability to compete' — a smart Chinese move.

Notes

Oculus and the Facebook deal

Luckey recounts building VR headsets as a 15-year-old hobbyist; DK1 sold about 55,000 units, deliberately beating the previous bestseller (Philips Scuba / Takara Dino Visor, 50,000 combined).
DK2 added six degrees of freedom tracking; DK1's 3DOF made users sick because inner-ear/eye mismatch makes the body think it's poisoned.
Samsung was the only viable display supplier; Oculus built the Gear VR with them partly to secure screens, selling over 10 million units.
Zuck offered roughly a billion dollars on the spot; the team turned it down. What convinced them was Facebook's commitment of a billion dollars a year in R&D for 10 years — Luckey's email: "If Zuck is playing us, he's Van Halen."
He frames the outcome as Oculus taking over Facebook: it became the dominant part of their R&D spend, with a public figure of $60 billion spent.

Is VR failing?

Quest 2 sold 20 million units, better than Nintendo 64's first three years and outselling Xbox; PlayStation VR sold about 6 million, HTC Vive 3-4 million.
His essay "free isn't cheap enough" (on palmerluckey.com) argues price isn't the blocker — weight, sweat, quality, and content are. Even free headsets would lose ~90% of users within a month.
Quest at $399 hugely outsold the $99 Oculus Go, suggesting higher-end products can win.

Anduril

Once called "most controversial company in tech" at 25 people; now ~7,000 including contractors, ~85% engineering.
The "don't work at Anduril" campaign tripled recruitment and filters for true believers over "lily pad jumpers."
~25 products with lean teams; every product older than 5 years has scaled (towers cover ~30% of the southern border; counter-drone is program of record for SOCOM and the Marine Corps).
He keeps current by reading academic literature as an "outsourced research team," distrusting press releases.

Geopolitics and defense

Anduril's 2017 pitch deck predicted Russia invading Ukraine and warfare shifting to proliferated, cheap, mass-manufactured systems.
He quotes Putin's 2007 line that the country that wins in AI will rule the world.
He can't compete with Chinese-component drones or Eastern European artillery costs; US rules (approved navigation, anti-tamper, no Chinese parts) tie his hands, and government-owned shell plants block competition — "a lot of secret communism" in the US.

Bank and optimism

His bank is a narrow bank: one-to-one reserves, stablecoin support, 50% loan-to-deposit ratio — designed to survive a bailout the government can't politically deliver.
Optimism thesis: food and clothing are already nearly free via automation; cars (~$1,000 F-150 equivalent) and housing will follow, with regulation the main bottleneck.

Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.