Trey/Shyam: Thanks for having us. Doing great. Good to be here, too.

Trey/Shyam: A really long time.

Trey Stephens: Well, I'll start. You can fill in all the gaps.

Shyam Sankar: I feel like I know the story you're going to tell and it's going to be very uncomfortable.

Shyam Sankar: So Trey, I think you know in the early days of Palantir, I was roaming around giving demos to anyone who could possibly want to see them. Trey was working in an intel agency and happened to see one of these demos and he should tell you his version of it, his side of it, his frustration with bureaucracy, but I think he realized like, hey, this might be really cool. Maybe I should leave the hell hole I'm in in the basement of this building getting nothing done, talking about sports with other people to go join this crusade. So Trey reached out and applied. He made a big faux pas. He came all the way to Palo Alto and he wore a full-on suit, tie, cufflinks—CIA cufflinks, mind you.

Shyam Sankar: Coming to interview with, I don't know, we were probably 20 people, you know, who wore t-shirts and, you know, second hand-me-down clothes. And he was intercepted in the lobby by a receptionist who really cared about him and told him to ditch the tie and try to dress down and don't screw it up too bad. But we loved him immediately and he helped us build the government business.

Trey Stephens: Yeah. I came in in early 2008, but there were still, you know, 25-30 people at that point.

Shyam Sankar: Yeah, it was one of those things that was kind of a slow start. There was a real idea amongst the five co-founders, including Peter, that it's kind of insane to live in a world post-9/11 where people are arguing about what's more important, privacy or security. Aren't they both really important? And who is actually spending time pushing out the efficient frontier? For any amount of given security, you should have more privacy than you had before, or any amount of given privacy, you should have more security. This sort of changing the dialectic there was really the entire impetus of what we started with. Now there's a technical approach that follows from that, an approach to privacy and civil liberties. But really we started as a business that was pretty myopically focused on solving a handful of problems in counterterrorism for a few institutions in the world.

Shyam Sankar: Well, anyone who's been to war would tell you that war is awful. War is bad. Categorically bad. That doesn't mean it's always avoidable. There are people who will want to use might to make right to define a set of rules. And you have to be in a position to protect your people and your interests accordingly.

Trey Stephens: At the end of the day, it's all about deterrence. You don't want to go to war, but you want to be prepared so that if you do have to go to war, that you will win decisively and quickly. I've never met a general that has said, "You know what I really want to do today? I want to make phone calls letting parents know that their children have died in combat." Nobody wants to do that. And I think that that's really the goal of everything we were working on at Palantir and what we're working on at Anduril, which is make it unthinkable to your adversaries that they should ever challenge you.

Shyam Sankar: I think it's a beautiful consequence—and I mean it unironically—of the peaceful world that we lived in. The origin story of Silicon Valley is actually defense. Lockheed was the largest employer in Silicon Valley in the 1950s. The Corona spy satellites were built there. In a world gripped by the threat of the Soviet Union, Silicon Valley有完全不同的姿态。但在“冷战结束”、“历史终结”后的世界，在全球化万岁的观点下，这些事情被看得更加愤世嫉俗，威胁似乎并不真实。

Shyam Sankar: I think many of the people who were protesting Silicon Valley's involvement in working on national security priorities were the exact same people that had a Ukraine flag in their bio. So there's clearly just a policy mismatch or an understanding mismatch. The other thing I would mention is that these large tech companies, unlike many epics prior, were global technology companies. They didn't necessarily view themselves as American. A lot of the signatures on these protests were not coming from US citizens.

Trey Stephens: I think that there's just an increased awareness of the complexity of the geopolitical situation. It's certainly a lot less controversial to work on these topics than it was in 2017 or 2018.

Shyam Sankar: The industrial base that won World War II and the early Cold War was not a "defense" industrial base. It was an American industrial base. Chrysler made the Minuteman ICBM—they were the prime contractor—同时他们也造面包车。General Mills 这家麦片公司曾有一个机械部门，利用研发谷物处理的技术来制造鱼雷和惯性导航系统。Ford 直到 1990 年还在造卫星。整个经济不仅在投资繁荣，还在为保障繁荣的自由提供担保。

Shyam Sankar: That figure today is 86% goes to defense specialists. We have a very different structure of the US economy as a result. And it leads to very perverse narratives of what it would be like to mobilize if things got really bad. It took 18 months to do that [in WWII]. When Ukraine went through 10 years of production in 10 weeks of fighting, that probably should have been a five-alarm fire that we got the fundamental calculus on deterrence wrong. We thought the stockpile was going to deter our adversaries. It was always the factory.

Shyam Sankar: Well, our joint force is the best in the world. But our adversaries are moving very quickly. Deterrence is eroding. You had Crimea in 2014, the militarization of the Spratly Islands, Iran, the Houthis. We have amazing high-end capabilities, but you can't keep shooting $2 million interceptors at $20,000 drones and have that math work very long. When you lose the general American industrial base, you lose volume and you lose the R&D stimulus to be creative—比如用造浴缸的方法去造廉价巡航导弹。我们现在被困在那些极度先进但单价 200 万美元、无法大规模生产的平台上。

Trey Stephens: We sent most of these capabilities away during the last 30 years of globalization, and that gutted entire communities. In my family, relatives worked at GM, Ford, Frigidaire—every single one of those factories in Ohio closed. What we're doing with Anduril right now is building a 5 million square foot factory campus in Columbus, Ohio. We are tapping into that underemployed knowledge base. Since 2000, only one new company has built major manufacturing muscle: Tesla. That's it. We've atrophied massively.

Trey Stephens: Arsenal-1 是我们在哥伦布建立的工厂园区。其背后的软件平台旨在降低自动化和流程管理的成本。我们正与 Palantir 的 Foundry 平台合作。核心理念是模块化。如果我们只固定生产某种无人机，在冲突中就会受限。我们的思路更像合同制造商（如富士康），我们建立的是一整套构建光学系统、动力系统的能力，并利用规模网络降低零部件成本。这样我们可以根据需求，瞬间从生产 A 产品转向生产 B 产品。相比之下，乌克兰战争初期，我们需要更多“毒刺”和“标枪”导弹，但库存耗尽后，生产线已不存在，甚至要从退休人员中找人回来教大家怎么造。

Trey Stephens: Our business model is very different from the "Primes" (传统五大防务商). They respond to requirements directly from the customer. We are doing all of this as private R&D investment and then selling the output as a product. This is a fundamental change. It requires a tremendous amount of capital. I don't think the market will support 100 new Primes, but I hope there is more competition. This isn't like building a normal tech company.

Shyam Sankar: Defense works as a monopsony where there's a single buyer. If you look at the history of defense innovation, typically the monopsony is wrong. Churchill as head of the Royal Navy built the tank because the British army was not smart enough to realize that having horses was not going to work. Every one of these innovations is an act of heresy. You need this competition.

Shyam Sankar: Just like venture capital, there's going to be a power law here. A grave mistake we've made is "innovation theater"—把资金像涂花生酱一样平摊给每一家公司。这无法达到规模化。你可能投了 10 个赌注，但在某些时刻，你必须把资金集中在那些真正奏效的事物上。

Trey Stephens: If you're a space tech investor and you didn't invest in SpaceX, you probably lost money. Capital allocators have a very short memory.

Trey Stephens: I think there's a real tension around the amount of capital raised and the valuations. My advice is: You can raise less at a lower price. This avoids "playing chicken" with numbers that don't make sense. In Anduril, we've focused on "climbing down the multiples tree"—我们不希望下一轮的营收倍数高于前一轮。这种纪律非常重要，尤其是考虑到中期 IPO 的目标。

Shyam Sankar: 太多了。除了战争武器，我更担心那些影响我们战斗意志的事物。制药就是其中之一。我父亲是药剂师。80% 的通用药物活性药物成分（API）由中国生产。 如果美国人在保卫自由世界与自家 5 岁孩子死于耳炎（本可轻易治愈）之间做选择，你觉得他们会选什么？我们沉溺于全球化愿景：我们负责创新，他们负责生产。但创新其实是生产力的结果。

Trey Stephens: Semiconductors is another big one. We basically just allowed it to happen somewhere else and now almost no amount of money is going to fix this problem on a timeline of relevance to the risk Taiwan has in 2027.

Trey Stephens: My belief in democracy is central here. The people of America elect representatives to make hard decisions. Fully autonomous weapons are not new. 我们有部署在舰艇上的 Phalanx CIWS（密集阵近程防御系统），它能完全自主击落空中威胁，因为人类根本没时间反应。但系统中存在问责制。船上有一个人对该武器系统的行为负责。未来的 AI 自动系统也将如此。

Trey Stephens: We are shooting non-explosive missiles into windows of apartment buildings, avoiding unintended casualties. AI leads to better precision and less civilian casualties. Abstention from participating in building technology for national security is not a morally neutral decision.

Shyam Sankar: 这种指责很荒谬。我们不收集数据，也没有数据。指责我们就像指责 Excel 是监控工具一样。Palantir 只是让客户把他们拥有合法授权收集的数据整合起来做决策。实际上，就像 Alex Karp 所说，由于我们的系统有极其严格的审计追踪，这是世界上最难进行非法操作的平台。

Shyam Sankar: It's like accusing Excel of being a surveillance tool. If you are "salami slicing" the policy as a tech person, that's actually tyranny by tech-bro. A small number of people are constraining the maneuver space of a democracy with no accountability.

Trey/Shyam: 100%. 每一个机构都有监察长（Inspector General）办公室。这套机制确实在发挥作用，有时甚至会被用来对付像 Maven 项目创始人 Drew Cukor 上校这样的英雄。

Shyam Sankar: 如果搞砸了，我们将迎来一个无法恢复的“中国世纪”，全世界都将成为中国的附庸国。CCP 的策略很明确：不仅要中国繁荣，美国必须倒下。如果做对了，我们将看到美国的大规模重新工业化，以及随之而来的西方复兴。我们将拥有一个繁荣的中产阶级——其定义是相信子女的未来会比自己更好。

Trey Stephens: The re-industrialization point is going to be the central point of making sure that China's belt and road strategy is not going to put us in a position where we literally just can't do anything.

Trey Stephens: 挑战在于，如果你不是领导者，你就无法设定参与规则。二战结束以来，我们一直坐在桌子的主位上，决定了半导体、供应链和贸易路线的规则。一旦你退后，别人就会制定游戏规则，而多极状态通常维持不了太久。

Trey/Shyam: Thank you.