Anduril & Palantir: How Silicon Valley Is Rebuilding America's Military
In a Nutshell
Palantir and Anduril leaders detail Silicon Valley's revival of U.S. defense tech, emphasizing deterrence through rapid innovation, massive manufacturing (e.g., Anduril's 5M sq ft Ohio Arsenal One campus), and software-hardware integration to counter China's 10,000:1 drone gap and 223x shipbuilding edge by 2027 Taiwan risks. They critique post-Cold War industrial atrophy (86% defense spending to specialists vs. 6% in 1989), advocate private R&D competition over cost-plus primes, and stress rebuilding stockpiles (only 8 days munitions vs. 800 needed) with attritable autonomous systems for quick, decisive wins. Patriotism demands rejecting anti-defense taboos, trusting democratic oversight on ethics/privacy, and re-industrializing for economic primacy to avoid a Chinese vassal world by 2040.
These notes were generated by AI and may contain inaccuracies.
Trey Stevens and Sean Sankar join the All-In podcast at the Hill and Valley Forum. They have known each other for a long time. Sean gave demos in the early days of Palantir. Trey, working in an intel agency, saw a demo, got frustrated with bureaucracy, and decided to leave his basement job to join Palantir. Trey applied and came to Palo Alto wearing a full suit, tie, cufflinks (CIA cufflinks). Palantir had about 20 people wearing t-shirts and second-hand clothes. The receptionist told him to ditch the tie. They loved him immediately, and he helped build the government business.
Trey joined Palantir as employee 13 in early 2008, when there were 25-30 people. Peter Thiel incubated it from the beginning.
Palantir had a slow start. The five co-founders, including Peter Thiel, believed post-9/11 it was insane to argue privacy vs. security—both are important. Push the efficient frontier: for any security, more privacy; for any privacy, more security. This changed the dialectic. Technical approach and privacy/civil liberties followed. Started myopically focused on solving counterterrorism problems for a handful of institutions.
Is war good? War is awful, categorically bad, but not always avoidable. People use might to make right, so protect your people and interests. It's about deterrence: don't want war, but be prepared to win decisively and quickly. No general wants to call parents about children dying in combat. Goal at Palantir and Anduril: make it unthinkable for adversaries to challenge you.
Beautiful consequence of the peaceful post-Cold War world. Silicon Valley origin: Lockheed largest employer in 1950s; Corona spy satellites built there. Soviet threat changed posture. Post-Cold War "end of history," globalism made threats seem unreal. Silicon Valley woke up with Russian tanks in Ukraine. Many protesters had Ukraine flags in bios—policy mismatch. Large tech companies saw themselves as global, not American; protests often from non-US citizens.
Awareness of geopolitical complexity has increased; less controversial now than 2017-2018. 20-year overnight success: Anduril raising at $60 billion valuation, $20 billion Army contract; Palantir worth $400 billion.
Industrial base that won WWII and early Cold War was American industrial base, not pure defense. Chrysler made Minuteman ICBM (also minivans). General Mills built torpedoes and inertial guidance using grain R&D. Ford built satellites until 1990. Entire economy invested in prosperity and underwriting freedom.
Post-Cold War consequence: In 1989 (Berlin Wall), only 6% major weapon systems spending to pure defense specialists; 94% to dual-purpose companies. Today, 86% to defense specialists. Perverse narratives on mobilization. History: Took 18 months to ramp up in WWII, starting for allies earlier. Ukraine used 10 years production in 10 weeks—five-alarm fire. Deterrence calculus wrong: factory, not stockpile.
10,000:1 drone production gap vs. China; 223x shipbuilding disadvantage; 2027 Taiwan window. US spends more than any nation. Joint force best in world, but adversaries moving fast. Loss of deterrence: Crimea annexation 2014, Spratly Islands 2015, Iran breakout 2017, pogrom in Israel, Houthis in Red Sea, present conflict. High-end capabilities amazing, but can't shoot $2M interceptors at $20K drones.
Lost industrial base means lost volume, R&D stimulus (e.g., bathtub methodology for low-cost cruise missile). Stuck with expensive platforms.
Globalization sent capabilities away, gutted US communities. Trey's family: GM, Ford, Frigidaire, National Cash Register, Armco Steel—all Ohio factories closed. Anduril building 5 million sq ft factory campus in Columbus, Ohio, tapping undermployed knowledge base.
At-scale manufacturing for new 2000s company: only Tesla. Atrophied ability to spin up quickly. Must invest ahead of conflict.
Arsenal One: Factory campus in Columbus. Arsenal platform OS reduces automation overhead. Working with Palantir Foundry. Modular design like contract manufacturers (VR headsets for Facebook, Samsung, Apple; optics systems). Network scale for lower costs, supply chain. Pivot to produce Furies, Roadrunners, Barracudas as needed.
Counterexample: Ukraine needed Stingers, Javelins; post-inventory, no assembly lines—retirees recalled to teach.
Anduril does private R&D investment, sells product output—different from primes responding to specs. Requires tremendous capital as new entrant. Hopeful for competition, not just Anduril. Not like normal tech company.
Defense as monopsony: single buyer with power, often wrong. Innovation history: heretics like Churchill building tanks. Soviet threat forced 8 competing ICBM programs—peacetime seen as wasteful. Need competition of ideas, execution variance, free market. Competition inside government too.
Originally 51 major defense contractors, now 5-6 primes subcontracting.
Anduril builds products cheaper, better, faster, competes on merits vs. primes. Capital concentrating: Anduril, SpaceX, OpenAI, few others. Power law like venture: concentrate on winners, not peanut-butter spreading (innovation theater). Space tech: SpaceX; crypto: Coinbase; social: Facebook. Capital allocators forget to concentrate.
1993 post-Cold War "peace dividend": DoD told 15 of 51 primes to consolidate (budget slashed). Led to 5 primes; nearly 4 in 1999, but Justice Dept. blocked Lockheed-Northrop.
Window for new companies: Not obvious early. Palantir not welcomed, sued customer to compete. Validated from field backwards—deployments where free market existed, bending rules to use software. Created facts on the ground. Long road.
Anduril benefited from Palantir's lessons, avoiding growing pains. Palantir took about 5 years to reach $10 million annually in revenue; Anduril did it in 22 months. This was due to founders living through Palantir's 5 years of challenges. There's enough experience now to work within the government system rather than fighting from outside.
Early pitch to Palantir before Anduril started: software is hard, hardware would be less hard. Government values hardware via bill of materials, margins easier to justify. Government historically made its own hardware, like Cold War ammunition factories. Software harder to value, often seen as lines of code or irrelevant metrics. Still hard; government wants cost-plus for software, ignoring R&D spread across commercial and government.
At Founders Fund, reviewed companies bidding on federal contracts via public lists and Crunchbase. Met hundreds; only one investment: KDM (renamed Expanse, acquired by Palo Alto Networks). Nothing else worth investing. Identified gap for next-gen prime: software-defined, hardware-enabled (not hardware-defined, software-enabled). Team encouraged starting Anduril.
Hardware needs higher capital, longer R&D cycles. Question on venture capital funding next-gen hardware defense.
Fairchild (1968): 96% integrated circuits sold to Apollo program (monopsony). Bob Noyce (Fairchild, transistor co-inventor) limited government R&D funding to 4%, chased Moore's Law independently. Enabled 1980s precision guided munitions for Assault Breaker vs. Soviets. Nvidia analogy: Jensen Huang sold 93% graphics chips as OEM until dominance. Space Shuttle: $50,000/kg to orbit; Starship: under $20/kg via heavy reuse. Cost-plus deletes profit when costs decrease.
SpaceX sued Air Force using same federal procurement law as early Palantir.
Shift: Palantir offer letter examples at $15B/$20B valuations seemed ridiculous; now seed rounds priced at $20B. Good: proves path to success (Palantir, SpaceX, Anduril). Bad: high capital/valuations add risk. Advice to startups: raise less at lower price to avoid playing chicken with unrealistic numbers (Silicon Valley HBO scene reference). Anduril climbs down multiples tree each round; Series H down significantly from Series G for discipline toward IPO.
$200B capital deployment (Emil Michael). Helps like 1950s-60s Air Force bootstrapping titanium supply chain. Address bottlenecks (e.g., brushless motors for drones). First customer (government) less economical than end customer.
Drone history: US invented (Abe Karem, Predator, General Atomics), but ITAR and FAA beyond-line-of-sight rules killed domestic market. Counterfactual: General Atomics consumer subsidiary owns market like DJI, driving down costs for national security.
Pharmaceuticals: 80% APIs for generic drugs from China. Conflict choice: defend world/sovereignty vs. 5-year-old dying from curable ear infection. Globalist error: innovate here, produce there. Innovation follows productivity. Google 2017 "Attention is All You Need" paper from 3% Translate improvement. 50% new clinical trials in China.
Semiconductors: US led (Fairchild), lost to TSMC. No domestic competitor investment. Hard to fix on Taiwan 2027 timeline.
Offsets: 1st nuclear, 2nd precision/stealth, 3rd decision advantage (outthink/out-execute). Tight thesis: interoperable, earned in exercises/tests/combat. Innovation messy; frameworks like MOSA destroy it.
Reimagine munitions/drones as consumables: order with expend/test/replenish cycle for next-gen demand. More buyer autonomy/flexibility vs. rigid commitments.
Office of Strategic Capital: critical minerals, refining, offtake agreements (vs. Obama-era Solyndra failure). Procurement revisions (Ash Carter's Third Offset). Private sector input on changes.
Not political, but people/leadership. Ash Carter (Democrat) and Trump admin focused fixes. Bureaucracy entropy toward sclerosis (David Packard's 7-page 5000 series acquisition now 2000 pages; JADS blown up). Kelly Johnson (Skunk Works): built 41 airframes (SR-71, U-2); rule to keep bureaucrats out. Big corps same issue.
Heretics/heroes as founders: U-2 via intel community (military didn't want); Benny Schrieber (ICBMs), Rickover (nuclear navy). Now systems by committee (F-35: components in 400 congressional districts; SLS in all 50 states vs. Starship). Calvinist humility hides founder figures (e.g., Gene Kranz/Apollo, John Boyd/F-16, Rickover's Navy).
Spear analogy: tip sharp, shaft needs work. DoD supply-demand disconnect: demand from combatant commands, supply from services/industrial base. Man, train, and equip
Producing material requires driving consensus on supply and demand, which is managed with great difficulty in the Pentagon. Agility to respond to scenarios is weak. Private sector companies have more precision: changing from X barracudas to Y barracudas and Z furies percolates quickly through the supply base. In government, decisions are locked in and cannot be revisited. Over time, entropy leads to unexpected bills, taking money from programs, resulting in minimum rate production to keep lines open, which is not deterrence but how we frog-boiled our way here.
Tie production to consumption: everything built, especially munitions, needs to be consumable to replenish, giving primes and neoprimes demand signals to produce cheaper and make more money. Contrarian idea: military-industrial complex is a crappy business trading at less than two times revenue; it's not profitable enough.
Trey Schlaepfer said we have 8 days of munitions on hand for a major conflict with China versus 800 days needed. How far away is the supply chain and production capacity to meet that?
There's a high-low mix: exquisite multi-million dollar interceptor missiles vs. dumb munitions like bombs. Not the same across the stack; some munitions woefully undersupplied with low readiness, others feel good. Anduril opened Arsenal yesterday to produce Furies, ramping as fast as possible. Over next 18 months, not taking foot off the gas, but hitting run-rate capacity limits as lines stand up.
With unlimited cash starting today, over next 18 months, country could get on track for sustainable industrial base. But we'll trickle it out over time without real political leadership, staying in similar situation unless leadership drives it forward. Take advantage of current urgency understood by administration leadership.
How much of the spear is autonomous systems? Is everything going autonomous, just drones, or on-ground autonomous?
Wars of today fought with weapons of yesterday, which are awesome: B2s, bombs dropped last summer in Iran, Patriot missiles—high-tech, exquisite, performant, hit targets almost every time. Problem: incredibly expensive, cannot resupply, way behind the eight-ball.
For wars of the future, build attritable mass systems today for inventory. Number of Anduril systems at readiness level, deployed and active in combat today, but small percentage of how wars waged. Over next 5-10 years, hi-lo mix must shift massively, evident in Ukraine and Iran. Department making steps for better divide of low-cost attritable vs. exquisite systems, but needs political leadership.
Anthropic refused Claude model for Maven without human oversight; Pentagon labeled them supply chain risk. What role should humans have, who hits the kill switch? Vendor responsibility to Department of Defense, where draw line on ethics?
Responsibility to ethics differs from belief in democracy. People elected representatives to make hard combat decisions. Fully autonomous weapons not new: SeaWhiz on naval vessels shoots down aerial threats fully autonomously—no time for decisions on incoming missiles. Has accountability: person on boat responsible for actions.
Future of autonomous systems: accountability baked in, like soldier with gun or naval captain with SeaWhiz.
Ethically: society progressed from rocks/sticks to knives, guns, bombs; WWII plateaued with nuclear weapons. Decided not to make more powerful nukes forever; combat engagement came back down chain: precision-guided weapons, non-explosive missiles into apartment windows avoiding casualties.
AI as command center for better decisions, precision, discrimination, fewer civilian casualties—ethically improved from dumb bombs on cities. Abstention from building national security tech is not morally neutral; it's a moral decision. Building company is ethically just, trusting democracy to deploy tools for American interests.
Palantir accused of enabling surveillance state; protest at All-In Summit in September where Alex Karp spoke.
Hard to respond: unclear what surveillance means—Terminator fear. Palantir doesn't collect data; brings your own lawfully collected data together for decisions—like Excel. Unabashedly patriotic, serving US military, colors views.
It's Excel with cell-by-cell auditability. Most insane platform for illegal activity—you'll be caught. Enhances privacy/security, civil liberties protections: system enforces laws/rules, not relying on people.
Need epistemic humility: elected officials accountable for policy. Lawful use perspective; salami-slicing policy is tyranny by tech bro—unaccountable few constraining democracy.
Example: Soviets got nuclear bomb via traitors—committed communists like Klaus Fuchs, and Theodore Hall (18-year-old Manhattan Project scientist, brother Edward Hall built Minute Man missile). Theodore thought two countries with bomb = world peace; walked into Soviet trade mission in New York 1944, gave secrets. Instead, every communist death since 1949 on his hands—no accountability.
Criticism of Palantir implies not wanting civil servants best tech to enforce policies—lack of belief in democracy. Examples: traffic cameras (libertarian concerns, but enforce red lights); autonomous speeding tickets in San Francisco (speaker got three last month); watching guy do heroin near elementary school while speeding.
Same for tax law enforcement. Position against tools is morally bankrupt.
Scenario: government corruption with advanced tech. If saw illegal use breaking law, do you report?
100%. Entire mechanism: every agency has Inspector General (IG)—independent, anonymous/non-anonymous complaints, statutory unfettered investigations (DoD, HUD, etc.). Used, e.g., weaponized against Colonel Drew Cukor (Maven inventor)—false claims like hiding illegal immigrants investigated by Naval Criminal Investigative Service.
First schism: Vietnam—lied to about war, fundamental academia-defense distrust unhealed, escalates through society. Second: few prior service/connected see military as humans with informed mental model; fears fill gaps, cartoon version without oversight/doctrine.
Snowden (15 years ago): little discussion of investigations—data collection not abused; <12 documented cases from tech errors, not policy. Policy renewed by multiple administrations/parties over decades.
Distrust of institutions creates stories. At Stanford guest lectures: no undergrads have immediate family in military—veterans rare. Divide from elites; less connection to veteran community vs. Cold War/WWII. Echoes JD Vance criticisms.
External influence driving culture? Social/mainstream media, destabilization attack vectors seen in science/industry—fear articles go viral.
100%. Soviets spent $7 billion (2026 dollars?) funding Vietnam peace/anti-war protests—organic element, but gasoline on fire for division. Today, CCP money to organizations protesting Palantir on domestic issues—successful adversary strategy.
Brilliant strategy from adversaries. US does counterintelligence: modern art movement CIA-funded to undermine Soviet art control, spreading cultural values.
Shot at aligning defense/tech with patriotism? Writing book Home Mobilize to do that. Clear-eyed world view: lost deterrence; WWII may have started already, perceivable in 10 years. Create clarity on stakes for children/future to get participation.
America's story: underdogs—17th largest army pre-WWII; rag-tag farmers vs. British in Revolutionary War. Success breeds nealism/self-loathing. Legitimacy of institutions: deserve best software; no doors falling off planes, provide services without fraud/corruption—breeds nealism otherwise, younger generations want to tear down, making things worse. Fight daily for functioning institutions.
America's story unwritten, new arc now. Rising socialist movement—impact on defense/resiliency next decade as adversaries rise?
Greatest threat: not homicide, suicide—internal discord, division, self-loathing, socialism symptomatic. Root: elite not caring about prosperity; globalization/NAFTA callousness (lose manufacturing job, learn to code). Not unabashed free market or communist—marginal decisions matter, need leadership.
Hopeful: socialism doesn't work. San Francisco: elected DA, recalled; board of education recalled; Daniel Lur as mayor—hit rock bottom, realized it fails. Seattle and Washington State are likely
Seattle and Washington State are likely going to lose a large number of their biggest employers. As that happens, they'll come through on the other side, probably taking 7 to 10 years. It might be painful and take a long time, but they'll eventually come to their senses.
The speakers bickered quite a bit at Palantir. Usually Sean was right about something and the other took a long time to come around. Well-functioning cultures are rooted in debate.
Debate on Communication Strategy
One disagreement was whether to show up to a tech company interview in a suit. A key part of Anduril's story: at Palantir, it was a bad decision to be as quiet as they were. They needed to get out there and tell their story to counter claims like "we don't have any data, we are Excel." Palantir had a quiet, reserved comms strategy. They went back and forth on that. When starting Anduril, they implemented all positive lessons from Palantir about doing business with the government, except they decided to go out and tell their story. That has worked incredibly well. Palantir has come along to that side of the debate.
If we don't do things right, what does 2040 look like economically and in defense? If we do things right?
If We Get It Wrong
National security is a means to economic prosperity for the American people. No country has done more to develop the world: the US sent integrated circuits and microelectronics to Southeast Asia, benefiting from trade. The US spent its own capital to rebuild conquered nations like Japan and Germany, leading to stability. Wrong path: a Chinese century where everyone is a vassal state to China, might makes right. For the CCP, it's not enough for China to prosper—America must fall, an explicit strategy. Example: smuggling agricultural funguses into the US so we can't grow soybeans—a zero-sum frame.
If We Get It Right
Massive re-industrialization of America followed by the West. Thriving middle class, defined qualitatively as one that believes their children's future will be better than their own—a fundamental promise broken over the last 30 years. Belief in institutions once again.
Education and Re-industrialization
Completely agree. Big component around education: figure out how to educate young people to successfully enter a marketplace that needs them and benefits from their services. Haven't nailed that yet. Re-industrialization is central to countering China's decades-long Belt and Road strategy, preventing a position where the US literally can't do anything.
Do we need military and industrial primacy, or can we operate in a multipolar world sharing influence and economic prosperity with China, Russia, or one or two other nation states? The challenge with not being the leader is you don't set the terms of engagement. Since the end of World War II, the US has had the primary seat at the table to set rules for semiconductors, supply chains, protecting trade lanes. Step back, and someone else sets the incentives—you play by their rules, and it doesn't stay multipolar for long.
Thank you to the guests. This has been great. Thank you. I'm going all in.
Keep All-In Podcast in your library
Save the videos and channels worth coming back to, and find them again in one place.
