Andreessen Horowitz Pivots to Global Strategy to Export US Tech Model

Andreessen Horowitz推进全球战略，输出美国科技模式 Andreessen Horowitz指出，构建运转良好的科技生态系统需要的不仅是资本，更需要技术人才、亲商政策以及对创业者的文化推崇。 在这场小组讨论中，a16z管理层阐述了将其硅谷模式推向全球的战略考量，把科技定位为国际实力与地缘政治威慑的核心舞台。发言者强调，美国的AI模型与软件正以前所未有的速度传播，向日本、沙特阿拉伯和墨西哥等盟国输出内嵌的西方价值观与民主原则。他们探讨了前沿AI在网络安全中的两用属性，指出自动化代码扫描在显著降低防御成本的同时，也增加了未打补丁的遗留代码所面临的风险。最后，Ben Horowitz详细说明了构建运转良好的科技生态所不可或缺的三大支柱：技术人才、亲商法规以及推崇创业雄心的文化氛围。 a16z 正在推进全球化战略布局，依托全球事务（Global Affairs）、全球投资人与 LP 关系、以及国际市场拓展（International Go-to-Market）三大支柱，将战略盟友直接接入美国创新生态。 (1:01) 现代软件与 AI 工具的全球传播速度远快于企业自建团队的节奏，a16z 通过直接对接海外关键政企买家，帮助初创公司在无需投入巨额本地合规建队成本的前提下锁定数百万美元大单。 (9:36) 前沿 AI 模型在网络安全中带来了双刃剑悖论：尽管代码审计与漏洞修补的成本大幅下降，但也使得攻击者能以极低成本在全球范围内对过去二十年积累的遗留代码进行自动化持续试探。 (29:05) 本·霍洛维茨强调再造硅谷无法单靠远程互联，必须集齐三大基石：输送实干型人才的顶尖高校、友好的营商政策环境（如避免对未实现资本利得征税），以及在文化上给予高风险创业充分社会地位与认同的舆论氛围。 (37:00)

Andreessen Horowitz partners argue that software innovation, defense technology, and artificial intelligence have become the primary arena of geopolitical power, shifting deterrence from traditional military scale to software iteration speed. To maintain Western technological dominance, the firm is pursuing an international strategy that connects venture-backed startups with allied governments and top-tier foreign enterprises. Technology as the Arena of Geopolitical Competition (5:06) Anne Neuberger and Ben Horowitz argue that modern deterrence is no longer dictated primarily by conventional military size, but by the speed at which software can counter low-cost, commercially available threats in strategic corridors like the Red Sea. In addition, critical infrastructure—including municipal water systems, electrical grids, and semiconductor supply chains—has become a high-stakes lever of nation-state coercion. Because frontier artificial intelligence models embed cultural ethics, rule of law, and historical interpretations, the speakers contend that ensuring allied nations adopt Western rather than authoritarian models prevents foreign adversaries from censoring sensitive historical events, such as Tiananmen Square. Scaling Software in Top-Heavy Global Markets (9:36) Raghu Raghuram and Jen Kha explain that while AI software products achieve instant global distribution via APIs, commercial scaling still depends on trusted regional networks. Setting up local legal entities and hiring overseas can cost early-stage startups $5 million to $10 million before securing a deal, leading many founders to treat foreign markets as an afterthought. To resolve this, a16z focuses on connecting startups directly to the five to ten enterprise and government entities that dominate purchasing in top-heavy economies, such as North Asia, where Japan, South Korea, and Taiwan represent 15% to 20% of the Global 2000. Horowitz cites Mexican broadcaster TelevisaUnivision, which used portfolio company ElevenLabs to dub its catalog into regional accents and foreign languages with AI voice cloning, enabling expansion into France and Poland and culminating in a distribution agreement with Netflix. Strategic Defense Alignment and Nearshore Partnerships (18:38) The panel details targeted regional investments across Japan, South Korea, the Middle East, and Latin America, viewing these relationships through an economic and national security lens. Neuberger notes that with China's shipbuilding capacity significantly outpacing American output, regional allies like Japan are overhauling defense policies toward large fleets of cheaper, autonomous platforms co-developed with US startups. Nearshore allies like Mexico require autonomous border and coastal surveillance to combat fentanyl trafficking and safeguard Western Hemisphere supply chains. In the Middle East, Horowitz defends backing Saudi Arabia's modernization, arguing its transition into an allied technology hub is vital to global stability. The firm has assisted startups like Adam Neumann's Flow by connecting them with local ministers, executives, and sovereign capital. Dual-Use AI and the Economics of Cybersecurity (29:05) Neuberger and Horowitz contextualize AI cybersecurity within the historical evolution of digital threats, which shifted from PC vandalism to internet-era theft (accelerated by deploying early Windows software with ActiveX online) to modern state-level warfare. Frontier AI models allow attackers to automate vulnerability scanning across global systems at unprecedented scale. However, Neuberger argues the economic balance ultimately favors defenders. While defenders historically suffered from an asymmetry where they had to patch every flaw, AI models now dramatically lower the cost of finding and repairing flaws across twenty years of unmaintained legacy code. Horowitz notes an unresolved dual-use paradox: training models to audit and remediate code inherently endows them with the capability to uncover novel offensive exploits, while portable code assets render traditional export controls largely ineffective. Three Pillars Required to Build Tech Hubs Dismissing claims that the internet has distributed Silicon Valley everywhere, Horowitz asserts that geography and governance still decide where innovation thrives. He identifies three essential ingredients for any durable startup hub: technical universities that train builders, pro-business policies, and a culture that confers high social status on entrepreneurial risk. On policy, Horowitz points to Norway as a cautionary tale, asserting that its tax on unrealized capital gains severely damaged the domestic startup ecosystem by penalizing illiquid equity. Culturally, he argues that Silicon Valley's defining advantage is its celebration of ambition, contrasting it with societies where outsized wealth accumulation is culturally disparaged as theft. While the speakers advocate for projecting Western influence through commercial and defense tech partnerships, the discussion leaves open how governments can effectively manage dual-use frontier models when digital assets evade physical export controls, and whether American policymakers will preserve the regulatory and tax environment that enabled Silicon Valley's original growth.

Ben Horowitz 与 a16z 合伙人指出，软件与人工智能已取代传统硬件和军事规模，成为国家权力和地缘影响力的核心竞技场；为此该机构正从硅谷本土迈向全球布局，通过绑定全球事务、主权资本与商业落地网络，推动国际关键产业体系锚定于西方技术架构之上。 科技主导权成为地缘安全与威慑的核心舞台 (5:06) Anne Neuberger 强调，技术已从辅助性的外交工具演变为地缘政治影响力和国家安全的核心场域。从半导体制造、关键药品原料等关键供应链，到电网与市政水务系统中的软件漏洞，均已被主权国家作为施加地缘控制的战略杠杆。现代军事威慑逻辑也发生了根本转变：在霍尔木兹海峡或红海等热点冲突区，威慑力不再单纯取决于传统军事规模，而是取决于应对低成本、商用化敌方软件与自主系统的迭代速度。面对中国造船产能大幅超越美国的现状，日本等盟国正在重塑二战后的防务政策，寻求与美国初创企业联合开发自主舰艇与 AI 战场系统。 Neuberger 与 Horowitz 进一步主张，前沿 AI 模型必然内嵌特定的伦理、法治与历史价值观。西方专有模型与非西方（如受国家控制的开源或封闭模型）的竞争不仅在于性能，更关乎核心价值观的全球投射。外国企业与机构若运行在受信任的西方底层模型之上，能够确保数据与交互逻辑不会遭受威权指令式的隐性篡改或审查（例如对敏感历史事件的抹除）。 头部市场拓展与政商资源跳板模式 (9:36) Raghu Raghuram 与 Jen Kha 指出，现代软件与 AI 产品的全球扩散速度已远超初创企业设立海外本地团队的速度。以往初创公司往往在年营收达数亿美元后才布局国际化，但借助 API 与即时数字分发，当下产品在发布初期即被海外需求拉动。然而，海外商业扩张本质上高度依赖信任与本土网络，许多国际经济体呈现显著的“头部集中”（top-heavy）结构——例如日本、韩国和中国台湾地区聚集了全球2000强企业中约15%至20%的份额，通常由5至10家核心财阀或龙头企业与中央政府掌握主要的采购预算。 鉴于初创公司在单个海外市场建立实体、招聘合规往往需要耗费500万至1000万美元成本，a16z 将自身定位为直通关键采购方的跳板，帮助被投企业在重资产落地前斩获大单。在拉美市场，TelevisaUnivision 面对跨国流媒体巨头竞争，借助 a16z 投资的 ElevenLabs 语音克隆技术完成全语种动态译配，在保留演员原声情感的同时打破方言与语言壁垒，迅速开拓法国和波兰市场并与 Netflix 达成全球分发合作；在沙特阿拉伯等快速推进经济多元化的中东市场，a16z 则协助 Adam Neumann 的不动产企业 Flow 直连政府高层与核心主权投资方。 双重用途悖论与网络安全攻防的经济学逆转 (29:05) 讨论强调，前沿人工智能在网络安全领域具备不可分割的“双重用途”（dual-use）属性。Horowitz 指出，无法在训练模型自动化排查与修复代码漏洞的同时，不赋予其识别攻击利用途径的能力。回顾网络安全演进历程，威胁形态从早期针对个人电脑的偶发“破坏”（vandalism），升级为互联网商业化时代针对数据的系统性“盗窃”（theft，如早年 Windows 集成 ActiveX 带来的安全漏洞敞口），最终发展为当下的国家级乃至 AI 驱动的“战争”（warfare）。 以 Anthropic 的 Mythos 等模型为代表的 frontier AI 工具正在逆转网络攻防的经济学公式。以往黑客需要人工探测局部系统的安全弱点，而 AI 工具使其能够以极低成本在全球范围内全天候自动化扫描攻击面。Neuberger 认为，这种能力最终有利于防守方：尽管过去二十年遗留的大量旧代码从未针对 AI 规模的刺探进行过安全加固，放任漏洞的风险成倍增加，但企业利用 AI 自动化扫描、审计开源组件与第三方法律依赖并进行修复的成本也大幅降低。鉴于数字代码极易通过移动介质转移而使出口管制难以完全奏效，确保全球数字基础设施基于内嵌安全护栏的透明模型运行成为关键防务策略。 硅谷难以复制的三大支柱与政策文化警示 面对众多海外政要关于如何本土复制硅谷的提问，Horowitz 驳斥了“互联网能使硅谷完全分布式化”的乐观看法，强调法治、政策与社会土壤依然决定创新的成败。他提出支撑可持续创业生态的三大必要支柱：第一是持续输出实战工程研发人才的顶尖理工院校；第二是亲商业的制度与法律环境，包括极简的设立程序、灵活的用工与解聘机制、适度的税负，以及对非流动性资产的有效保护；第三是崇尚进取与风险、赋予创业者崇高社会声誉的文化框架。 Horowitz 特别以挪威为例进行预警：挪威引入未实现资本利得税后，因重创创业者承受非流动性股权风险的能力，严重损害了本土科技生态。与此同时，他批评部分社会文化中将高额财富视同道德原罪的仇富心理，指出硅谷的核心竞争力在于对大胆冒险与商业成功的制度性激励与文化推崇，这种脆弱的生态环境一旦遭遇政策或监管破坏，将极难被复原。 本期讨论系统阐释了 a16z 将风险投资升级为地缘科技战略支点的运作逻辑，主张通过输出西方技术与搭建盟友高层网络来维系美国的地缘科技主导权；但该策略在推进过程中，仍面临如何长期协调主权盟友的本土工业利益诉求、如何在去中心化代码扩散中有效规制 AI 双重用途风险，以及如何抵御美国本土监管政策变动对创新生态潜在冲击等未决挑战。

## Background 该视频由知名风险投资机构 a16z 于 2026 年 8 月 3 日发布在 YouTube 平台。节目中，a16z 联合创始人 Ben Horowitz 携手前白宫高官 Anne Neuberger、前 VMware 首席执行官 Raghu Raghuram 以及合伙人 Jen Kha，围绕 a16z 拓展国际化战略的考量展开探讨，深入分析了前沿技术在推动经济增长、国家安全和全球战略伙伴关系中扮演的关键角色。 ## TL;DR 对话系统拆解了科技如何从传统外交辅助工具蜕变为大国综合国力竞争的核心竞技场。嘉宾们指出，大模型与底层技术天然承载着文化与制度价值观，面对 AI 时代初创企业跨国扩张极速前置的现状，机构通过政策对接、资本纽带与落地赋能直接帮助企业破局海外“关系型经济体”。讨论深入剖析了印太防务现代化、AI 重塑网络攻防经济学的双刃剑效应，并总结了孕育创新生态不可或缺的三大基石：顶尖人才、包容政策以及宽容失败并推崇创业成功的社会文化。 ## Atmosphere & style 全场对话兼具国家安全层面的地缘宏大叙事与一线投资创业的冷峻务实，政界智囊、资深操盘手与顶级投资人的碰撞张弛有度。在探讨防务漏洞、网络战争等严肃话题的同时，嘉宾们不时穿插拉美肥皂剧 AI 译制等生动案例与机智幽默的调侃，让复杂的科技政治议题变得鲜活透彻。节目尾声从严谨的制度分析升华为充满感染力的开拓者信念，洋溢着坚信“我能做到”的纯正硅谷乐观主义底色。 ## Bulletin - [01:01](https://www.youtube.com/watch?v=XwfUzW32cIA&t=61s) a16z 宣布全面迈向国际化，建立由前白宫高官 Anne Neuberger（全球事务）、Raghu Raghuram（国际市场拓展）与 Jen Kha（全球资本）组成的“三位一体”跨国战略架构。 - [05:06](https://www.youtube.com/watch?v=XwfUzW32cIA&t=306s) 现代地缘竞争的核心已从军队规模转向应对低成本、软件化创新的敏捷反应速度，技术研发主导权已从政府转向私营部门。 - [14:17](https://www.youtube.com/watch?v=XwfUzW32cIA&t=857s) 初创企业传统跨国扩张前期沉没成本高达 500 万至 1000 万美元，a16z 尝试通过直接对接海外政商关键决策层来打破出海死局。 - [18:40](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1120s) 印太盟友防务采购正由极少数造价昂贵、交付周期漫长的单一重型平台，转向更大规模、低成本且消耗得起的大量智能化无人系统。 ## Key takeaways - [07:24](https://www.youtube.com/watch?v=XwfUzW32cIA&t=444s) 大模型并非客观中立的产物，底层训练数据天然内嵌了文明立场与价值观导向，主导全球模型底座关乎言论自由与法治理念的投射。 - [05:06](https://www.youtube.com/watch?v=XwfUzW32cIA&t=306s) 半导体供应链与关键基础设施软件漏洞已由普通商业资产演变为国家权力的核心竞技场与战略威慑筹码。 - [09:36](https://www.youtube.com/watch?v=XwfUzW32cIA&t=576s) API 经济与大模型赋能打破了传统软件出海的时钟节奏，初创企业组织扩张速度远落后于产品全球扩散速度，倒逼全球化全面前置。 - [10:30](https://www.youtube.com/watch?v=XwfUzW32cIA&t=630s) 海外大多数经济体呈现高度头重脚轻的“关系型经济”特征，初创企业直接攻克前五大财阀集团和政府买家远比自建庞大海外实体更具回报比。 - [24:41](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1481s) a16z 拒绝传统 VC 依靠本地团队挂牌代持与分成（carry）的模式，主张将全美资本与客户生态作为真实产品赋能具备出海野心的本土成长型企业。 - [27:19](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1639s) 全球多数初创生态具备充沛的早期种子轮资金与技术人才，但在成长期普遍面临成长资本（growth capital）与全球落地通道的双重断层。 - [29:31](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1771s) AI 大模型使得代码漏洞的挖掘门槛与防御成本同时降低，但自动化巡检与修补使长期处于倾斜劣势的网络防御方首次迎来逆转优势的窗口。 - [33:01](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1981s) 在 AI 驱动下，黑客正以近乎零边际成本无休止地“摇晃每一个数字门把手”，网络对抗已从最初的极客恶作剧、经济盗窃升级为大国维度的网络战争。 - [37:00](https://www.youtube.com/watch?v=XwfUzW32cIA&t=2220s) 复刻硅谷必须同时集齐三大支柱：能够源源不断输送动手者的顶尖理工人才、具备正向税制与包容度的创业友好法律政策，以及免遭酸葡萄心理污名化的崇尚商业冒险与成功的社会文化。 - [41:03](https://www.youtube.com/watch?v=XwfUzW32cIA&t=2463s) 硅谷最核心也最脆弱的精神资产在于对待创业成功的包容与崇尚，结合源自《小火车做到了》中“我能做到”的骨髓级乐观主义，构成了颠覆性创新的内在驱动力。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=XwfUzW32cIA&t=0s) 引言与播客背景：技术作为国家权力的新竞技场 节目开篇通过几个核心议题的精彩预告展开讨论。当 AI 成为连接所有事物的统一入口，任何工具与技术都将建立在这些大模型之上；然而大模型绝非客观中立的产物，它们拥有鲜明的立场与观点，在历史、文化和价值观等方面均承载着特定的取向。同时，现代地缘政治的威慑力量已不再仅取决于军队的庞大规模，而是取决于应对廉价、基于软件的技术创新的反应速度。这类技术正由对手——不论是国家政体、网络犯罪分子还是恐怖组织——广泛使用，且它们的研发源头已不再是传统政府部门，而是私营部门。此外，产品和技术的传播速度往往远快于企业组织自身扩张的速度，使得初创企业不得不以前所未有的速度迈向全球化。 ```mermaid flowchart TD accTitle: AI时代科技与国家权力的演进 accDescr: 展示从AI基础设施到全球地缘政治与初创企业出海的相互影响链条 subgraph TechShift["技术范式变革"] A["AI 成为万物通用入口"] --> B["模型内嵌价值观与历史观"] C["私营部门主导关键技术研发"] --> D["廉价软件化技术成为非对称武器"] end subgraph Impact["现实影响力投射"] B --> E["技术出海与数字主权博弈"] D --> F["国家威慑力取决于创新迭代速度"] A --> G["初创企业被迫极速国际化"] end ``` ### [01:01](https://www.youtube.com/watch?v=XwfUzW32cIA&t=61s) a16z 的全球战略构想与“三位一体”框架 a16z 历史上曾是一家高度聚焦于门洛帕克（Menlo Park）本地的投资机构，早期甚至连在旧金山设立办公室都极为谨慎；如今全面迈向国际化是一个标志性的重大转变。Ben Horowitz 阐明了这一雄心背后的根基——对美国独特创新生态的坚信：如果把一个毫无资源、人脉和资金的人置于世界上任何一个角落，只要其拥有让世界变得更好的构想，那么他在美国成功的几率依然是最高的。因此，a16z 的核心使命之一，便是捍卫这一特殊的创新土壤。要维持这一地位，美国必须在经济、文化和军事上保持领先，而这一切的前提是必须确立技术领导地位。 随着这一使命的推进，a16z 意识到必须将全球伙伴和盟友纳入其中。最初，国际投资者渴望参与美国的前沿科技红利，促使 a16z 接触了来自各国的出资人（LP）。在与这些国际伙伴深入沟通后，a16z 总结出伙伴国家的两层深切诉求：一是不仅希望获得财务回报，更渴望直接采用并落地先进技术；二是政府决策层亟需了解如何优化本国政策环境，从而培育本国的创新生态。为了系统化应对这些需求，a16z 搭建了由三大支柱组成的国际战略架构： - **全球事务（Global Affairs）**：由前白宫及情报部门高官 Anne Neuberger 掌舵，负责在国家政策层面与各国政府（如沙特阿拉伯、日本、墨西哥等）建立深度互信与外交对接通道。 - **全球资本与合伙人关系（LP Relations & Capital）**：由 Jen Kha 等团队主导，联结主权财富基金、海外机构投资者以及管理公共养老金的投资机构。 - **国际市场拓展（International Go-to-Market）**：由前 VMware 首席执行官 Raghu Raghuram 负责，协助全球政府和企业级客户直接引入并落地 a16z 生态系统内的前沿科技。 ### [05:06](https://www.youtube.com/watch?v=XwfUzW32cIA&t=306s) 芯片、漏洞与无人作战：科技从外交工具跃迁为国家权力竞技场 针对为何选择加入 a16z，Anne Neuberger 结合此前在国家情报委员会和白宫负责网络与新兴技术事务的履历进行了剖析。她指出，过去技术往往只是外交与国家权力的外在辅助工具，如今技术本身已经变成了“国家权力的核心竞技场（the arena of national power）”。最典型的体现是半导体芯片供应链，它早已脱离单纯的商业 IT 范畴，演变为关乎国家经济生存与国家安全的基础支柱；类似的情况也出现在医药核心原料的供应链上，一旦某一国试图利用供应链对别国进行遏制，它就直接转化为了国家权力的杠杆。 与此同时，软件安全漏洞的概念已被彻底改写。当一个关键漏洞潜藏于供水系统或电网等基础设施时，它就是地缘对抗中的关键战略筹码。Anne Neuberger 强调，现代大国威慑不再仅仅依靠正规军事规模，正如在霍尔木兹海峡或红海危机中展现的那样，面对非国家行为体或犯罪集团利用廉价、软件驱动的无人机及自动化武器发起的不对称攻击，防御方的胜负取决于自身技术创新的反应与迭代速度。由于这些关键技术——包括 AI、自主无人系统以及网络防御系统——几乎全部由私营部门主导研发，因此建立政府与私营部门之间的高效协同，并与西方盟友“共同构建、共同作战（build together, fight together）”，成为确保全球体系稳定与价值观传递的核心基石。 ### [07:24](https://www.youtube.com/watch?v=XwfUzW32cIA&t=444s) 拒绝“网络殖民”：当大模型内嵌文明价值观 针对当前核心大模型多集中在旧金山特定街区开发的现状，Ben Horowitz 进一步深入探讨了技术背后的价值观博弈。没有任何国家愿意在数字空间或赛博世界中沦为“被殖民者”。当 AI 演变成接管人类生活方方面面的通用交互界面——无论是汽车控制、家用电器还是下一代教育体系——所有终端都将运行在底层基础模型之上。 然而，大模型从来不是冰冷客观的数据聚合，它们天然内嵌了主观判断。在历史事实的诠释、文化理解、伦理规范和价值导向等方面，不同文明背景下的模型给出的结论可能大相径庭。战争的历史归因、主权法律体系的评判、甚至对特定国家是善意还是恶意行为者的判定，完全取决于模型基于何种数据与意识形态进行训练。Ben Horowitz 明确指出：“尽管我们自身存在缺陷，但我们的基础价值观与理念——从经济体系的运作方式到法治（rule of law）——为世界上的人类创造了最广阔的机会。”如果承认这一点，就必须致力于让代表自由竞争与法治的西方技术体系成为全球底座，避免全球用户不得不被动接受那些压制个人发展机会的其他价值观。 ### [09:36](https://www.youtube.com/watch?v=XwfUzW32cIA&t=576s) 软件以光速出海：初创公司全球化进程的全面前置 Raghu Raghuram 详细分析了 AI 时代企业出海逻辑的结构性突变：“产品与技术的传播速度永远快于企业组织自身扩张的速度（Products and technology travel faster than companies can）。”在传统的企业软件时代，一家初创企业通常在年收入达到数亿美元、组织规模相当成熟时，才会正式考虑全球化布局与海外落地。但在当前的 API 经济与 AI 时代，一款优秀的产品刚刚发布便能瞬间跨越地理界限，全球开发者和企业可以立即在线调用其服务并在其上构建应用。 这一现象导致初创企业在生命周期的极早期阶段就被推向国际市场，不得不应对多国家、多渠道和多语种的复杂环境。幸运的是，AI 技术的自我演进也加速了这一应对过程：大模型不仅让底层产品被更迅速地发现，其自身强大的语言与理解能力也成为产品本地化和国际化（Localization & Internationalization）的高效引擎，使得规模极小的初创团队也能以极低的成本将其服务快速适配到全球各个细分市场中。 ### [10:30](https://www.youtube.com/watch?v=XwfUzW32cIA&t=630s) 关系型经济体与 ElevenLabs 助力 TelevisaUnivision 出海 Raghu Raghuram 强调，虽然现代技术扩散的速度极快，但要在全球范围内真正实现规模化扩张，商业底层逻辑并没有改变。企业不能脱离本土市场结构，必须在当地建立深厚的关系与合作伙伴网络。许多国家的经济本质上是“关系型经济”（relationship-based economies），而非纯粹的“交易型经济”（transaction-based economies）。因此，当前出海面临一种奇特的张力：技术迭代日新月异，但做生意和建立信任的方式却依旧依赖传统的深厚关系。 Ben Horowitz 紧接着分享了一个生动的案例：a16z 团队近期到访墨西哥，与 TelevisaUnivision 首席执行官 Alfonso de Angoitia 会面。作为 a16z 的全球战略合作伙伴，Alfonso 坦言自己面临来自 Netflix、Disney 等全球流媒体巨头的严峻挑战，这些流媒体拥有强大的美国技术壁垒。TelevisaUnivision 坐拥极其丰富的西班牙语影视内容，但受限于墨西哥西班牙语口音（不同于阿根廷或哥伦比亚口音）以及多语言鸿沟，难以在全球广泛变现。 ```mermaid flowchart LR accTitle: a16z 赋能本土巨头与被投企业走向全球 accDescr: 展示 TelevisaUnivision 与 ElevenLabs 合作后成功打入全球流媒体市场的链路。 subgraph Partner["本地文化巨头"] A["TelevisaUnivision<br/>海量西班牙语内容"] end subgraph Matchmaker["a16z 全球生态"] B["促成技术对接<br/>a16z 被投企业"] end subgraph AI_Tech["前沿 AI 能力"] C["ElevenLabs<br/>多语言/原生音色笑声配音"] end subgraph Global_Market["全球化商业成果"] D["覆盖法国、波兰等全球市场"] E["斩获 Netflix 重大战略合作协议"] end A --> B B --> C C --> D D --> E ``` a16z 随后引荐了其被投 AI 音频公司 ElevenLabs。短短两个月内，TelevisaUnivision 的影视剧便成功输送至法国、波兰乃至全球多国。ElevenLabs 的技术让演员的声音完美保留了原本的情感、语调与标志性笑声，只是换成了各种当地语言。凭借这一出海突破，TelevisaUnivision 随后更与 Netflix 签下了重大合作大单。Ben 指出，这正是 a16z 国际战略的核心：将本土文化深厚但面临天花板的合作伙伴，与前沿盟友技术相结合，实现双赢而非对抗。 ### [12:56](https://www.youtube.com/watch?v=XwfUzW32cIA&t=776s) 模型中的价值观：内容真实性与反审查 Anne Neuberger 顺着 TelevisaUnivision 的案例，切入到许多人经常提出的尖锐问题——如何看待美国基础模型与中国开源模型的差异？她指出，TelevisaUnivision 在对影视剧进行多语言本地化译制时，完全不必担心模型会在翻译中横加政治审查。如果剧本台词涉及敏感话题（例如天安门事件或中国政府相关议题），使用西方或美国模型能够确保内容忠实还原原始表达，不会在模型推理中途被强行修改、过滤或注入政治意志。 这生动体现了“美国与西方价值观”内嵌于核心基础设施中的实际意义：保障表达自由、不搞政治性内容审查。Anne 补充表示，a16z 依然极度重视开源生态（Open Source），因为不同场景和价格敏感度需要多样化的模型选择，但在全球关键基础设施的构建上，底层模型承载的价值观防线至关重要。Jen Kha 随即笑着调侃道：“虽然肥皂剧（Telenovelas）里的戏剧性冲突已经够狗血、够辛辣了，我不知道故事情节里是不是真的需要冷不丁冒出天安门广场这类话题，但剧情总会涉及边境偷渡之类的敏感题材，原汁原味确实很关键。” - **AI 本地化中立性**：避免生成式 AI 在跨国语种翻译和文化输出中，因模型政治对齐策略而篡改剧本原意。 - **开源生态考量**：承认开源模型在丰富场景与低成本部署上的重要性，但强调不可忽视底层技术背后的价值观偏向。 ### [14:17](https://www.youtube.com/watch?v=XwfUzW32cIA&t=857s) 成为权力撮合者：破解初创企业的海外破局成本 Jen Kha 提到，投资作家 Packy McCormick 年初在撰写 a16z 特写文章时，曾将该机构描绘为“权力掮客”（Power Broker）。a16z 正是将这种深厚的政商权力网络转化为赋能创业者的实际产品。通常情况下，像墨西哥这样的市场绝不会排在初创企业国际化路线图的前三名甚至前十名。但如果 a16z 能够直接帮创始团队敲定头部核心客户，并在企业生命周期极早的阶段铺平道路，创始人就会重新评估优先级。 Raghu Raghuram 剖析了海外市场的商业结构：大多数国家的商业生态呈现极其显著的“头重脚轻”（top-heavy）特征。哪怕是在英国这样成熟的发达经济体，相当大比重的 GDP 和软件采购预算依然集中在最顶尖的少数巨头手中。而在发展中国家或拉美地区，真正具有决定性影响力的核心企业通常也就五到十家，再加上政府这个关键买家。对于软件公司而言，大费周折地在墨西哥设立庞大的全功能实体可能并不划算；但如果能通过顶级高层关系直接签下前五大集团中的一两家，其营收价值将远超在其他国家漫无目的地自建团队试错。 Ben Horowitz 进一步算了一笔严酷的账：初创企业传统意义上的跨国扩张极其昂贵。成立海外法人实体、组建本地团队、办理合规与本地化改造，往往在拿到第一份订单前就得烧掉 500 万到 1000 万美元。昂贵的前期沉没成本迫使创业公司推迟国际化进程。但如果 a16z 能提前铺路，让创始人甚至在没有设立实体之前，就直接面对一个 500 万或 1000 万美元规模的实际签约机会，扩张的投资回报率就会发生质的飞跃。 - **墨西哥的战略地缘价值**：Ben 认为硅谷公司把墨西哥排在海外拓展第十名极为遗憾。从美国国家安全与供应链韧性来看，构建紧密的西半球制造与供应链生态至关重要。 - **沙特阿拉伯与 Flow 案例**：通常也没有创始人会第一步奔赴沙特，但沙特目前正处于现代化转型爆发期。沙特摆脱传统部族形态、转型为融入现代世界的开放经济体并成为美国坚定盟友，对全球安全极为关键。a16z 带领 Adam Neumann 创立的居住社区品牌 Flow 进入沙特，直接牵线政府高层、商业巨贾和顶级资本，精准满足了利雅得大量外派专业人才的高品质居住需求。 ### [18:40](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1120s) 重点区域布局与印太防务技术变革 在谈到 a16z 目前重点倾斜的国际战略区域（包括日本、韩国、中东、墨西哥和加拿大）时，Ben 指出，筛选的核心标准在于三要素的深度交织：该国是美国的强力盟友，且 a16z 已经在当地与“政府决策者、商业领袖、顶级投资机构”三大群体建立了牢固且相互联动的网络关系。只有当政府、产业与资本三位一体协同运作时，才能最大限度地缩短技术落地周期。 Anne Neuberger 从地缘政治与国防现代化的视角补充了紧迫性。当前全球格局动荡，供应链武器化加剧，美国在印太地区的盟友正面临前所未有的外部地缘压力。以中国为例，其军事力量迅猛增长，不仅在核力量建设上显著加速，造船产能更是比美国高出数个数量级。 这一失衡倒逼印太盟友彻底重构战略底线： - **日本防卫政策历史性破局**：日本彻底打破了二战后奉为圭臬的和平防卫限制，下定决心打造一支具备实战对抗能力的现代军队。 - **应对“当下的战争”而非“过去的战争”**：日本迫切希望建立能够应对现代冲突的防务体系，但自身缺乏现成的防务工业基础（Defense-Industrial Base）。 - **重构国防采购范式**：过去二十年美国国防改革反思中的核心结论在此适用——传统采购过度依赖极少数极其昂贵、精密且交付周期漫长的单一重型平台（exquisite platforms）；面对现代战争，各国必须转向部署规模更大、成本更为低廉且消耗得起的大量智能化平台，仅在极少数关键节点保留精密重器。这一国防工业思维的转向，为现代技术型初创企业的快速嵌入打开了前所未有的窗口。 ### [21:00](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1260s) 地缘政治与国家安全视角下的战略区域选择 Anne Neuberger 从国家安全和地缘威慑的角度，深入拆解了美国与盟友在技术领域的合作考量。当盟友着眼于建立强有力的军事能力以威慑区域侵略时，他们正积极寻求与新兴科技公司合作，研发自主舰艇（autonomous ships）以及面向一体化战场空间（integrated battlespace）的人工智能技术。关键在于，这些国家不再满足于单纯采购成品装备，而是强烈渴望“与我们共同建造”（build with us）——他们希望将核心能力沉淀在本土，并保护自身的供应链。 这种防务与技术合作在全球关键节点上有着清晰的地理分布： - **印太与澳洲**：构建一体化联防体系，涵盖从前沿传感器、自主海域巡航到分布式作战网络的全套能力。 - **中东地区**：长期处于美国的核心利益区。以色列拥有极具竞争力的军工与情报体系；阿联酋积极拥抱前沿科技；沙特阿拉伯则在全面推进现代化，聚焦 AI 与基础设施投资，全力打造本土的创新与创业基座以参与全球竞争。 - **拉美与墨西哥**：芬太尼（fentanyl）等毒品流入美国边境是严峻的公共安全威胁。通过在墨西哥漫长的海岸线部署自主巡逻舰艇，不仅能强化墨西哥国家安全、维护拉美稳定，更直接关乎美国本土的安全利益。 这种将“美国活力”（American Dynamism）与 AI 基础设施相结合的合作模式，使投资行为超越了纯粹的资本回报，上升为支撑国家安全的战略伙伴关系。 --- ### [22:16](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1336s) 为何优先选择这些市场：政策前瞻、集中型经济与技术“跨越式”发展 Raghu Raghuram 指出，选择落地的国家必须符合多维度的综合评估标准。首先，这必须在战略层面高度符合美国的国家利益。其次，这些国家的政府往往对 AI 持极其前瞻和激进的拥抱态度，积极推行横跨国防与整体经济现代化的顶层政策。第三，这些区域的经济结构普遍呈现“高度头部化”（top-heavy economies）：只要打入少数几家顶尖财阀或核心企业，就能撬动庞大的全球影响力。例如日韩台三地加起来，就贡献了全球 2000 强（Global 2000）企业中 15% 至 20% 的席位。 此外，这些非英语国家的商业运作逻辑与英美等传统市场存在巨大差异。硅谷创业公司在没有本地背书的情况下极难破局，这反而让 a16z 能够提供极为显著的高附加值服务。 ```mermaid flowchart TD accTitle: a16z 国际化落地筛选评估维度 accDescr: 展示 a16z 选择拓展特定海外国家时考量的四大核心要素 subgraph S["战略与政策协同"] A["契合美国核心国家利益"] B["政府高度 AI 前瞻且推行现代化政策"] end subgraph E["市场与生态价值"] C["头部集中型经济<br/>(日韩台占 Global 2000 约 15-20%)"] D["本土商业网络壁垒高<br/>(VC 赋能空间远超英美市场)"] end S --> T["优质国际化落地目标国"] E --> T ``` Jen Kha 补充道，许多新兴市场在技术应用上甚至展现出了“跨越式发展”（leapfrogging）的特征。由于不受历史遗留 IT 系统的拖累，他们直接采纳最新方案。她举例称，在沙特入境时，办理数字签证通关仅需 30 秒，而回到美国却常常面临漫长的排队与低效的流程。技术在不同维度的落地并不均匀，如何利用这一落差加速前沿技术在全球不同场景的应用，蕴藏着巨大的时代机遇。 --- ### [24:41](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1481s) 拒绝“贴牌分舵”：a16z 全球化战略与传统 VC 的本质区别 Jen Kha 指出，传统风投在出海时大多是“投资驱动型”，只在乎把钱投出去，鲜少真正从被投企业的国际化拓展诉求出发。Ben Horowitz 随即回顾了 a16z 创立初期的克制态度——在机构成立的前十年里，团队极少踏出美国本土，甚至很少离开硅谷。 Ben Horowitz 解释说，a16z 衡量投资的唯一尺度是：“我们究竟能为企业带来什么附加值？”（what do we bring to the table?）。在 a16z 的理念中，机构的核心价值是为创业者对接顶尖人才、获取关键企业客户并提供专业指导。在机构规模尚小、无法跨地域稳定输出这套服务时，宁愿聚焦在 Menlo Park 和旧金山把“产品”打磨到极致。 与此相反，传统 VC 的底层逻辑是将“资金本身视作产品”。因此他们的全球化打法往往是：在英国等海外地区物色一支本土投资团队，为其品牌背书帮其募资，随后坐享收益分成（carry）。Ben 直言这种打法完全不符合 a16z 的价值观： > “我们的全部核心都在于：如何帮助你打造一家更伟大的公司？如何加速你的机会与增长？……我们不会去为某些不属于我们公司、不认同我们文化和使命的人募资，让他们打着我们的旗号去执行另一套打法。那根本不是我们。” 随着 a16z 自身的资金规模与全球声望达到特定体量，开启真正的全球化才水到渠成。a16z 进军瑞典、日本、沙特或阿联酋，目的绝不是为了“抢夺当地的投资机会”或挤压本土的天使轮及 A 轮投资人，而是扮演成长伙伴：当本土孕育出的优秀企业展现出走向全球的野心，且面临成长资本与出海网络匮乏时，直接利用 a16z 的全美及全球生态网络为其赋能。 --- ### [27:19](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1639s) 填补成长资本缺口与全球化生态落地 Jen Kha 与 Raghu Raghuram 探讨了全球创业生态中普遍存在的断层。许多地区本土人才储备充裕、早期种子轮孵化活跃，但普遍匮乏“成长资本”（growth capital）以及帮助企业跨入全球舞台的系统化管道。这导致大量潜力初创公司往往过早沦为“本土局限型冠军”，无法蜕变为真正的全球级巨头。 Raghu 结合其在 VMware 担任 CEO 期间的经历分析道，过去只有当企业发展到极其成熟的晚期阶段，管理层才会去考虑全球化扩张；而在当下，一款无论面向 B 端还是 C 端的优秀产品，其产品价值几乎天然具备普适性（universal appeal），这促使企业必须以比以往快得多的节奏走向全球市场。 然而，企业出海面临的最大瓶颈是缺乏支持性的生态体系。企业进入一个全新市场，无法凭空建立深度的本地信任与人脉，而这绝非朝夕之功。a16z 在本地设立办事处、持续扎根经营，正是为了帮被投企业在进入新地域时实现“快速起跑”。与此同时，深度融入本地生态也能让团队近距离发掘具有全球雄心的本地顶尖创业者，以直接投资与全球资源网络助其扬帆出海。在这些国家开展业务，本就是融入一个集体生态（collective ecosystem），而非单打独斗的霸权竞争。 --- ### [29:31](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1771s) 双刃剑困境：AI 对网络安全攻防格局的重塑 对话随后切入到 AI 前沿技术的双重用途（dual-use dilemma）及开源与闭源的博弈。面对这一争议话题，Anne Neuberger 结合其在白宫担任网络安全与新兴技术副国家安全顾问的亲身经历，以网络攻防为例进行了深入拆解。 无论是攻击者还是防御者，核心工作流的第一步都是扫描并发现系统代码中的安全漏洞。如果是防御方，接下来就是打补丁（patching）；如果是攻击方，接下来就是利用漏洞进行入侵利用（exploit）。AI 基础模型在寻找漏洞方面的效率达到了质的飞跃，这让后续的防御和利用门槛同时大幅降低。 Anne 指出，早在本轮前沿 AI 大模型爆发之前，网络防御的经济学模型就已经彻底失衡破裂。在白宫任职期间，令她夜不能寐的梦魇之一便是针对美国偏远地区乡村医院的网络勒索攻击（ransomware attacks）： - 这些乡村医院方圆几十里往往没有替代医疗机构； - 背后受到某些国家暗中包庇的跨国犯罪团伙，仅凭一串勒索代码就能直接瘫痪整座医院，剥夺普通民众获取基础医疗救助的生命通道； - 关键民生基础设施呈现出极高的系统脆弱性。 虽然 AI 工具为攻守双方同时赋能，但 Anne 个人坚定地认为，AI 对防御方的实质助益远大于攻击方。在传统对抗中，防御方必须兼顾辽阔无际的防御阵线，而攻击方只需找到一个薄弱切口即可得手；AI 自动化代码分析和威胁拦截工具的大规模应用，正在帮助防御方以极低的边际成本对系统漏洞进行全面巡检与自动修复，从而有望逆转长期倾斜的攻防不对称劣势。 ### [30:48](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1848s) 网络安全的双刃剑困境与可信模型 当前的数字安全环境正处在一个极其脆弱的十字路口：攻击面极其庞大，存在漏洞的系统不计其数。无论面对全球各国的政府还是大型企业，所有人都在被同一个核心问题所困扰。从主权国家的角度来看，政府通常肩负着两大基本任务：保卫本土安全以及维护海外利益。在这一前提下，最大的挑战莫过于：如何确保这些威力极其强大的技术能力不落入对手或犯罪分子之手？ 在传统的网络攻防工具时代，人们就已经在为此类问题头疼。当一种网络武器或攻击能力可以装进一个小小的 U 盘时，"出口管制（Export Controls）简直就像个笑话"，根本无法实现物理上的封堵。而人工智能模型的全球化扩散，使得基于“可信模型（Trusted Models）”构建基础设施变得至关重要。当全球用户运行的是来自美国的模型时，他们无需担心背后存在蓄意留存的后门，也不必担心聊天机器人的输出会为了迎合威权政府的政治审查叙事而遭到篡改。 因此，当务之急是如何将安全机制内生于大模型之中，以便它们能够在高敏感环境中平稳落地。由于目前最前沿的主流大模型大多出自美国，美国在定义这一技术范式上具备独特的先发优势。然而，这项任务既不可能单靠政府自上而下地完成，也无法完全放手给私营部门自由发展；它迫切需要最懂底层技术的科技界专家与政府政策制定者坐在一张桌子前，共同在“鼓励创新”与“风险控制”之间达成精妙平衡。 ```mermaid flowchart TD accTitle: AI双刃剑模型与网络安全攻防困境 accDescr: 展示AI大模型在漏洞修补与潜在网络攻击之间的两难机制及私营部门与政府的协作逻辑。 subgraph DualUse ["AI双刃剑机制"] A["前沿AI模型"] --> B["自动化扫描代码漏洞"] B --> C["防御侧：先于黑客打补丁修复"] B --> D["攻击侧：攻击者自动化摇动数字门把手"] end subgraph Solution ["应对治理路径"] E["私营部门顶尖专家"] & F["政府政策制定者"] --> G["协同建立内生安全与治理框架"] G --> H["推动创新并提供可信底座"] end DualUse -.-> Solution ``` ### [33:01](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1981s) 攻防演进：从恶作剧、盗窃到战争 正如 a16z 合伙人马克·安德森（Marc Andreessen）常用的比喻那样，科技行业就像是“追上并咬住了大巴的那条狗（the dog that caught the bus）”。世界各国前所未有地迫切索求最新版本的模型，因为各国深刻意识到技术领先直接关乎国家经济、国民福祉与战略利益。这彻底重构了传统的产品出海逻辑。然而，技术本身的双重属性（Dual-use）让治理陷入了经典的悖论： 政府本能地希望限制大模型具备网络黑客攻击的能力，但现实技术逻辑却极为残酷——如果你想利用模型找出自己代码中的漏洞并进行修补，你就必须赋予它发现漏洞的能力；而一旦模型能看穿漏洞，它就自然具备了发起攻击的潜在能力。倘若彻底封禁这种挖掘漏洞的能力，那么最终掌握这些杀伤性能力的，反而会是那些专门越狱（Jailbreak）系统的恶意攻击者（Bad Actors）。 正如业内以 Anthropic 等前沿模型所引发的探讨所示，AI 带来的最大变量是“规模化持续摇晃每一个数字门把手（jiggling every doorknob continuously and at scale）”。网络防御的唯一胜算，就是利用模型“先于攻击者攻击自己的系统”。随着 AI 的介入，修复不安全代码的成本大幅下降，而不去修复的代价却呈指数级上升。网络安全经历了三个历史阶段：最初是极客恶作剧（Vandalism），互联网普及后演变为经济盗窃（Theft），如今在 AI 时代则已彻底升级为国力较量的网络战争（War）。正如微软当年将尚未准备充分的 Windows 暴露在早期互联网上（伴随 ActiveX 等漏洞）引发长达二十年的系统脆弱性，如今全球正拖着过去二十年沉淀下来的海量遗留代码（Legacy Code），被迫迈向 AI 原生的安全新纪元。 - **数字门把手隐喻**：AI 允许黑客以零边际成本在大规模网络范围内无休止地试探系统每一个微小的防护漏洞。 - **攻击范式的三级跳**：计算机破坏/恶作剧（Vandalism） $\rightarrow$ 互联网时代的经济盗窃（Theft） $\rightarrow$ 大国博弈下的网络战争（War）。 - **遗留代码债务**：前 AI 时代积攒了近 20 年的开源依赖项与无维护代码，正迎来一次性被 AI 彻底重构或打补丁的历史机遇。 ### [37:00](https://www.youtube.com/watch?v=XwfUzW32cIA&t=2220s) 复制硅谷的三大基石：人才、制度与文化 当走向国际舞台时，各个国家和地区向 a16z 提出的头号问题无一例外是：“在本地如何复刻一个硅谷？”有些人天真地认为，互联网的全球普及意味着硅谷已经被彻底分布式化、全球化了，本·霍洛维茨直言这纯粹是“自欺欺人的漂亮话（happy talk）”。每个国家的法律、税制、治理模式天差地别，这正是美国至今依然独特的原因。霍洛维茨甚至打趣道，真希望把美国本土那些天天试图拆解、打击科技行业的政客带到这些国家走一走，让他们亲耳听听全球其他国家对拥有一个硅谷是多么梦寐以求。 要想真正培育出科技创新生态，必须同时集齐三项不可或缺的核心要素： 1. **顶尖的技术人才（Talent）**：其显性标志是是否拥有能够源源不断培养出“真正懂得动手创造产品的人”的顶尖理工类大学。世界上许多国家都拥有优秀大学，但它们仍然没能孕育出硅谷，原因在于缺乏后两项支撑。 2. **包容创业的法律与政策环境（Laws and Policies）**：制度必须对商业友好。这包含几个极其务实的衡量标准：注册一家公司是否简单？雇佣和解雇员工是否顺畅？税收结构是否具有正向激励而非惩罚性？此外，近期极其敏感的关键在于：法律是否允许个人持有具备潜在价值但缺乏流动性的资产，还是会粗暴地将其课税或没收。例如，挪威此前推出的未实现资本利得税（Unrealized Capital Gains Tax），几乎实质性地扼杀了当地的科技创业生态；如果缺乏良性的政策保护，绝不会有人愿意承担极端的个人与财务风险去创办企业。 3. **推崇创业与成就的社会文化（Culture）**：这或许是最微妙也最关键的合成因素——整个社会是否给予那些最优秀、愿意付出巨大努力做出贡献的年轻人以崇高的社会地位和声誉回报？去创立一家新公司或者加入一家初创团队，在当地文化里到底“酷不酷”？在大部分国家，商业成功往往遭受冷眼甚至污名化，政客们常煽动平民主义情绪，抛出“赚到十亿美元的人一定全靠偷窃”这类酸葡萄心理（sour grapes）的荒谬言论。唯有在美国（以及很大程度上的中国），追求颠覆性商业成功的雄心才能够获得真正的文化认可与社会地位激励。 ### [41:03](https://www.youtube.com/watch?v=XwfUzW32cIA&t=2463s) 创新文化的脆弱性与美国的“我能做到”信念 谈到硅谷创业生态中最核心且隐秘的精神特质时，本·霍洛维茨（Ben Horowitz）指出，许多地方之所以无法诞生真正的创新，根源在于对成功的敌意与嫉妒心理——人们往往会自我安慰：“那个人怎么赚了那么多钱？哦，他肯定是个小偷；而我不是小偷，所以我现在感觉好多了。”相比之下，硅谷对待创业成功的包容与崇尚是极其罕见的。这种文化“极难被复制，极难被构建，却又极易被摧毁”。霍洛维茨强调，大家必须清醒地意识到这种文化土壤在全世界是多么特殊与珍贵；在尽力帮助其他国家和地区建立类似创新生态的同时，美国自身也必须付出同等努力，千万不要亲手毁掉这种精神基石。 随后讨论借由当下的世界杯话题切入到美国文化特有的乐观主义与雄心壮志。有人提到世界杯期间的一条推文引发了广泛共鸣：“美国队居然真的觉得自己有可能赢得世界杯冠军。”尽管从客观实力来看或许能赢也或许赢不了，但“美国人打心底觉得自己能做到”这一事实本身就是一项无可替代的“产品特性，而非缺陷（a feature, not a bug）”。这种强大的宏大抱负与自信，正是推动科技与创业不断挑战不可能的内在力量。 嘉宾们随即产生热烈共鸣并补充道，正如几乎所有美国家庭都会给孩子阅读的经典童话《小火车做到了》（*The Little Engine That Could*）一样，书中最经典的一句话就是小火车不断默念的：“我想我能做到，我想我能做到，我想我能做到（I think I can. I think I can. I think I can.）。”这种深植于文化骨髓中的信念——坚信自己有能力改变现状、实现飞跃——构成了美国技术与商业创新的精神底色。在对这种文化价值的深刻共识与互致感谢中，全场讨论圆满收尾。 - 创新文化的心态差异：将富裕与成功污名化为不道德（如“他是小偷”的酸葡萄心理）会彻底破坏创业生态，而硅谷对冒险与成功的正向鼓励是极其罕见且脆弱的文化资产。 - 抱负的价值：无论是面对世界杯还是前沿技术突破，无论外界看来多么遥不可及，自始至终深信自己有能力获胜是一种宝贵的驱动力，是美国创新生态的核心特质而非认知偏差。 - 文化传承的载体：经典童书《小火车做到了》中的名句“I think I can”生动概括了美国鼓励个人奋斗与相信可能性的精神底蕴。 ## Notable quotes > [05:06](https://www.youtube.com/watch?v=XwfUzW32cIA&t=306s) 技术本身已经变成了国家权力的核心竞技场（the arena of national power）。 — Anne Neuberger > [07:24](https://www.youtube.com/watch?v=XwfUzW32cIA&t=444s) 尽管我们自身存在缺陷，但我们的基础价值观与理念——从经济体系的运作方式到法治（rule of law）——为世界上的人类创造了最广阔的机会。 — Ben Horowitz > [09:36](https://www.youtube.com/watch?v=XwfUzW32cIA&t=576s) 产品与技术的传播速度永远快于企业组织自身扩张的速度（Products and technology travel faster than companies can）。 — Raghu Raghuram > [24:41](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1481s) 我们究竟能为企业带来什么附加值？（what do we bring to the table?） — Ben Horowitz > [24:41](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1481s) 我们的全部核心都在于：如何帮助你打造一家更伟大的公司？如何加速你的机会与增长？……我们不会去为某些不属于我们公司、不认同我们文化和使命的人募资，让他们打着我们的旗号去执行另一套打法。那根本不是我们。 — Ben Horowitz > [30:48](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1848s) 出口管制（Export Controls）简直就像个笑话。 — Anne Neuberger > [33:01](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1981s) 科技行业就像是‘追上并咬住了大巴的那条狗’（the dog that caught the bus）。 — Marc Andreessen > [33:01](https://www.youtube.com/watch?v=XwfUzW32cIA&t=1981s) 规模化持续摇晃每一个数字门把手（jiggling every doorknob continuously and at scale）。 > [37:00](https://www.youtube.com/watch?v=XwfUzW32cIA&t=2220s) 自欺欺人的漂亮话（happy talk）。 — Ben Horowitz > [41:03](https://www.youtube.com/watch?v=XwfUzW32cIA&t=2463s) 极难被复制，极难被构建，却又极易被摧毁。 — Ben Horowitz > [41:03](https://www.youtube.com/watch?v=XwfUzW32cIA&t=2463s) 产品特性，而非缺陷（a feature, not a bug）。 > [41:03](https://www.youtube.com/watch?v=XwfUzW32cIA&t=2463s) 我想我能做到，我想我能做到，我想我能做到（I think I can. I think I can. I think I can.）。 ## Mentioned Ben Horowitz, Anne Neuberger, Raghu Raghuram, Jen Kha, Marc Andreessen, Packy McCormick, Alfonso de Angoitia, Adam Neumann, a16z, VMware, TelevisaUnivision, Netflix, Disney, ElevenLabs, Flow, Anthropic, Microsoft, Windows, ActiveX, 《小火车做到了》（The Little Engine That Could） ## Video description Ben Horowitz is joined by Anne Neuberger, Raghu Raghuram, and Jen Kha to discuss a16z's expanding international strategy and the growing role technology plays in economic growth, national security, and global partnerships. The conversation explores why America's technology leadership matters beyond Silicon Valley, how AI is reshaping relationships between governments and the private sector, and why countries around the world are looking to adopt frontier technologies while building stronger innovation ecosystems of their own. They discuss AI infrastructure, cybersecurity, defense technology, startup expansion, and what it takes to build enduring technology ecosystems. Along the way, they examine the role of trusted partnerships, the importance of Western technology, and why helping founders expand globally has become an increasingly important part of company building. Timestamps: 00:00 - Intro 01:01 - Why a16z Is Going Global 05:06 - Tech as the Arena of National Power 09:36 - Products Travel Faster Than Companies Can 14:01 - American Values in the Models the World Runs On 18:52 - Prioritizing Geos: Japan, Saudi, Mexico & the Indo-Pacific 24:41 - How a16z's International Strategy Differs from Other VCs 29:31 - AI, Cybersecurity & the Dual-Use Dilemma 37:00 - What It Actually Takes to Build a Silicon Valley Resources: Read more about a16z’s global mission: https://a16z.com/a16zs-global-mission/ Follow Ben on X: https://x.com/bhorowitz Follow Anne on X: https://x.com/AnneNeuberger Follow Raghu on X: https://x.com/RaghuRaghuram Follow Jen Kha on X: https://x.com/jkhamehl Stay Updated: If you enjoyed this episode, be sure to like, subscribe, and share with your friends! Find a16z on X: https://twitter.com/a16z Find a16z on LinkedIn: https://www.linkedin.com/company/a16z Listen to the a16z Show on Spotify: https://open.spotify.com/show/5bC65RDvs3oxnLyqqvkUYX Listen to the a16z Show on Apple Podcasts: https://podcasts.apple.com/us/podcast/a16z-podcast/id842818711 Follow our host: https://x.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see http://a16z.com/disclosures. [Source](https://www.youtube.com/watch?v=XwfUzW32cIA)

[5:06] Anne Neuberger: 技术本身已经变成了国家权力的核心竞技场（the arena of national power）。 [7:24] Ben Horowitz: 尽管我们自身存在缺陷，但我们的基础价值观与理念——从经济体系的运作方式到法治（rule of law）——为世界上的人类创造了最广阔的机会。 [9:36] Raghu Raghuram: 产品与技术的传播速度永远快于企业组织自身扩张的速度（Products and technology travel faster than companies can）。 [24:41] Ben Horowitz: 我们究竟能为企业带来什么附加值？（what do we bring to the table?） [24:41] Ben Horowitz: 我们的全部核心都在于：如何帮助你打造一家更伟大的公司？如何加速你的机会与增长？……我们不会去为某些不属于我们公司、不认同我们文化和使命的人募资，让他们打着我们的旗号去执行另一套打法。那根本不是我们。 [30:48] Anne Neuberger: 出口管制（Export Controls）简直就像个笑话。 [33:01] Marc Andreessen: 科技行业就像是‘追上并咬住了大巴的那条狗’（the dog that caught the bus）。 [33:01] 规模化持续摇晃每一个数字门把手（jiggling every doorknob continuously and at scale）。 [37:00] Ben Horowitz: 自欺欺人的漂亮话（happy talk）。 [41:03] Ben Horowitz: 极难被复制，极难被构建，却又极易被摧毁。 [41:03] 产品特性，而非缺陷（a feature, not a bug）。 [41:03] 我想我能做到，我想我能做到，我想我能做到（I think I can. I think I can. I think I can.）。

[0:00] Ben Horowitz: the belief that an ambitious person from anywhere in the world, equipped only with an idea, has the highest chance of success in the United States. [9:36] Raghu Raghuram: Products Travel Faster Than Companies Can [37:00] the internet is Silicon Valley distributed [37:00] the only way you could make a billion dollars is if you're a thief. [37:00] I think I can, I think I can

[0:00] When AI becomes the interface for [0:01] everything, every tool we have, every [0:03] technology we use is going to be built [0:05] on these models. And the models are not [0:08] objective. They have opinions. They have [0:10] strong opinions on things like history, [0:12] on culture, on values. And those are [0:15] very different depending where you are. [0:18] >> Deterrence is no longer just the size of [0:19] a military. It's the pace of innovation [0:22] in reacting to cheap softwarebuilt tech [0:25] that adversaries may be using whether [0:26] they're a country, criminal, or a [0:28] terrorist group. And those technologies [0:30] today are not being built by [0:32] governments. They're being built by the [0:33] private sector. [0:34] >> Products and technology travel faster [0:37] than companies can. So what happens is [0:39] for startup companies, usually olden [0:42] days, you didn't care about going [0:43] international until you had a few [0:44] hundred million in revenue. Now these [0:46] companies have to go sooner and sooner [0:49] internationally. [0:49] >> The number one question we often get [0:51] when we're visiting different countries [0:52] is how do you recreate Silicon Valley [0:54] locally? [0:55] >> I don't think that's true, by the way. I [0:56] think that's a happy talk. [1:01] >> We are here to talk international today. [1:03] Um, lots of things going on at the firm [1:05] uh for a firm that used to historically [1:08] uh be all centered in Menllo Park and [1:10] and we didn't even want to open up a San [1:11] Francisco office. This is a big change [1:13] for us. So Ben, why don't you take us [1:15] through uh this announcement around our [1:17] global ambition and how all the pieces [1:18] between global affairs, international go [1:20] to market, LP relations on the global [1:23] partnership level kind of fit in [1:24] together. [1:25] >> Yeah. Yeah. Well, it really kind of [1:26] starts with um America, interestingly. [1:30] Uh and you know, it's kind of been our [1:33] belief, our our strong belief that you [1:36] know, if you're drop a person into any [1:39] country in the world and they don't have [1:41] the culture and they don't have uh money [1:45] and capital resources yet, and they just [1:47] have an idea of how to make the world a [1:50] better place, um their best shot is in [1:53] America. And um that's just a very very [1:57] special thing. And so a big mission of [2:01] our firm is kind of trying to preserve [2:05] that that special thing that is America. [2:08] And that means America has to be a [2:10] leader. It has to be a leader uh [2:12] economically. It has to be a leader [2:13] culturally. It has to be a leader [2:15] militarily. Um and to do that we have to [2:18] be a leader technologically. We have to [2:20] be a tech leader. And that's kind of [2:22] really our role in the ecosystem and in [2:25] the world is to help America be a [2:27] leader. And then uh what we realized as [2:31] kind of we've gotten further along with [2:33] this mission is that we need to kind of [2:35] bring in our partners, our allies into [2:38] that mission. Uh and it kind of started [2:42] uh with uh the work that you're doing [2:44] Jen with uh just people from around the [2:47] world wanting to kind of participate in [2:50] the uh kind of US technological [2:52] opportunity and that kind of meant you [2:55] know we started getting investors from [2:57] around the world. Uh and as we went out [3:00] to see those investors and we said you [3:02] know look how can we be a better partner [3:05] um we found some interesting things. So [3:08] one is every country wants to not only [3:12] kind of economically benefit from the [3:15] technology but they'd like to actually [3:16] directly benefit from the technology. [3:18] Can you bring it to us? Can we get a [3:21] chance to use it? Um and then at a kind [3:24] of higher level uh you know at at the [3:27] kind of government policy level people [3:29] wanted to know well how can we be a [3:30] great country for adopting technology [3:33] for innovating like uh you guys are and [3:36] for kind of helping push our country [3:39] into the future. And you know as we uh [3:42] looked at okay how do we kind of step up [3:45] and uh step into that opportunity and be [3:48] a good partner we realized um with our [3:51] country partners we needed relationships [3:53] at the kind of government level at the [3:56] investor level at the adoption level and [3:58] that's what what these kind of three [4:00] components become. So we've got um [4:04] global affairs which uh builds our [4:08] relationships led by an which uh helps [4:11] us have relationships in the right way [4:14] with uh governments from you know [4:16] ranging from Saudi Arabia to Japan to [4:20] Mexico. Um and we've had uh you know [4:23] just great conversations with all of [4:25] them. Then we have people who uh invest [4:29] you know sometimes those are the [4:32] government sometimes they're uh various [4:34] companies in the government sometimes [4:36] they're companies in who actually um [4:38] handle government money in the form of [4:40] pensions uh and then finally um you know [4:44] both the government and the businesses [4:46] want to adopt the technology and that's [4:48] where kind of ragu's effort comes in [4:50] where we have the uh the goto market [4:52] component [4:53] >> um and I'm going to get you involved [4:54] here how much of your decision to join [4:56] A6 andZ was motivated by this mission [4:59] and also the sense of duty and helping [5:01] other countries think about the adoption [5:02] to AI and not wanting to get left behind [5:04] in AI like Ben was alluding to very much [5:07] so. You know what I saw happen in the [5:09] years in the intelligence committee and [5:11] then the White House is technology move [5:13] from being a tool of diplomacy of [5:16] national power to the arena of it where [5:20] the supply chain for chips was not a [5:22] technology issue as much as a [5:24] foundational economic and national [5:26] security issue. The supply chain for the [5:28] components that go into drugs. When used [5:31] by one country to constrain or have [5:35] impact on another country, it becomes a [5:37] tool of national power. [5:39] >> A software vulnerability is now not an [5:40] IT issue. [5:41] >> Yep. [5:42] >> It can be a source of leverage, [5:44] particularly if that vulnerability is in [5:46] key parts of your power systems or your [5:48] water systems. [5:49] >> Yeah. So I saw that shift happen from [5:51] tech being either an IT issue or just a [5:54] tool to the central arena of both [5:57] economic growth and national power. And [6:00] quite frankly the from a deterrence [6:03] perspective which is the language we [6:05] talk about between countries in terms of [6:07] deterring an invasion of a country [6:08] maintaining an international system of [6:10] stability. You know deterrence is no [6:13] longer just the size of a military. I [6:15] think as we see in the straight of [6:16] Hermuz or we saw in the Red Sea, it's [6:18] the pace of innovation in reacting to [6:21] cheap softwarebuilt tech that [6:23] adversaries may be using whether they're [6:25] a country, criminal or a terrorist [6:27] group. [6:27] >> Yep. [6:28] >> And those technologies today are not [6:31] being built by governments. They're [6:32] being built by the private sector. [6:34] Whether that's AI systems, whether [6:35] that's autonomous systems, whether [6:38] that's cyber defense systems, they are [6:40] being built on in the private sector. So [6:42] certainly from a US perspective ensuring [6:45] that we have the relationships between [6:46] government and private sector to [6:47] leverage that for our own economic [6:49] growth and as a force for good on the [6:51] world stage but also that our allies as [6:53] well we build together we fight together [6:56] and to ensure that together we can both [6:59] have the advantages of that and to the [7:01] extent that our tech also represents our [7:04] values [7:05] >> it's important that the world run on [7:07] America and on western tech. Yeah, [7:09] actually that's a great point on the [7:10] values point because Ben you've said [7:12] before no country also wants to get [7:14] colonized [7:15] >> in cyerspace right particularly because [7:18] AI is now just getting built in the six [7:20] block radius [7:21] >> in San Francisco um say say a little bit [7:23] more about that [7:24] >> well you know when AI becomes the [7:27] interface for everything um so every [7:30] every tool we have every technology we [7:32] use every kind of system we have from [7:34] your car to your refrigerator to your [7:37] education system is going to be built on [7:40] these models and the models um are not [7:45] objective. They have opinions um and [7:47] they they have strong opinions on things [7:49] like history, [7:50] >> on culture, on um you know, ethics, on [7:55] values. And by the way, those are very [7:58] different depending where you are. you [8:01] know, the uh you know, the history of [8:03] wars and the you know, history of of of [8:07] a country and laws and whether that [8:09] country was a good actor or a bad actor [8:12] uh are all very different depending on [8:16] um where you're getting the information [8:18] from. And so if you believe that you you [8:22] know as as I do that you know for all [8:24] our flaws [8:26] our fundamental values and ideas about [8:31] everything from the way our economic [8:33] system works to the rule of law uh have [8:37] resulted in [8:39] basically the best opportunities uh for [8:44] humans in the world. then you want those [8:48] values to be projected onto the world [8:50] more so than to be the recipients of [8:53] some other values. Um, which you know [8:57] maybe were good at a time or maybe um, [9:01] you know, have some merits but overall [9:04] don't give people a real opportunity. [9:07] >> Yeah. Giving giving people a chance. [9:08] That's the that was the premise of the [9:10] original announcement we did for our [9:12] fund announcement earlier this year and [9:13] the importance of that and then [9:14] particularly extending that chance to [9:16] the allies of America as well. Um [9:18] regular I want to get you involved on [9:20] this as well because portfolio companies [9:22] are increasingly going international way [9:24] sooner. They're multi-country [9:25] multi-channel multi-product way sooner [9:27] in the life cycle. How much of this is [9:29] also accelerated by what's happening in [9:30] AI and just the ability to push out [9:33] software updates right just immediately. [9:36] >> Yeah. I mean the reality is that [9:38] products and technology travel faster [9:41] than companies can [9:43] >> right because if you got a great product [9:45] everybody learns about it very quickly [9:47] instantaneously you can access it [9:49] through APIs wherever in the world you [9:51] are you can build on top of those things [9:54] so what happens is for startup companies [9:56] usually in the olden days you didn't [9:58] care about going international until you [10:00] had a few hundred million dollars in [10:01] revenue etc now these companies have to [10:05] go sooner and sooner sooner [10:06] internationally. [10:08] So while [10:10] um the products are being discovered [10:12] internationally sooner, also to Ben's [10:15] earlier point, it's easier to adapt [10:17] these products for local markets much [10:21] more easily thanks to AI or as wise [10:24] becomes the interface to localize uh and [10:28] internationalize these products, it's so [10:30] much more easier. [10:32] But in order to capitalize on the [10:34] opportunity at scale, things don't [10:36] change, right? I mean, you still have to [10:38] be in region. You still have to [10:41] understand the market structure of the [10:43] region that you want to operate in. Um, [10:46] you have to understand partnerships, you [10:49] have to build relationships. Lots of [10:51] other countries have very relationship- [10:52] based economies, not transaction-based [10:55] economies. So it's it's this interesting [10:57] blend where you've got to the technology [11:00] moves really fast but the way you do [11:02] business hasn't changed that fast. Yeah. [11:05] Actually so a really great example of uh [11:07] what Ragu was talking about. So uh when [11:10] we were in uh Mexico recently um we were [11:14] meeting with uh the CEO of Tel Ava [11:17] Univision which uh Alonso Angua who is a [11:22] great friend of the firms who one of our [11:24] global partners and we were asking about [11:26] gee you know what are your struggles and [11:28] one of his challenges is you know he's [11:30] really faces uh you know heavy [11:33] competition from some of the kind of [11:35] global streamers like you know a Netflix [11:37] or Disney or so forth. who have this US [11:41] technology that's very powerful. Um, but [11:44] he has, you know, this unbelievable [11:46] Spanish-sp speakaking content. Uh, but [11:48] it was hard for him to get to play on a [11:50] global basis because it's in Spanish. [11:52] It's a Mexican Spanish accent, which is [11:54] different than a an Argentinian Spanish [11:58] or Colombian Spanish accent. Uh, and you [12:01] know, he's like, you know, how should we [12:03] solve this? And we're like, well, we've [12:05] got this great company, 11 Labs. They [12:07] can help you with it. We put them [12:08] together in a couple of months. He was [12:11] in, you know, France and Poland and all [12:13] over the world with uh this incredible [12:16] technology that had his actors basically [12:19] acting exactly as they normal do with [12:22] their right kind of intonations, their [12:24] right laugh, everything but in different [12:27] languages. And he recently um based on [12:30] that success got a big deal with [12:32] Netflix. Yeah. Uh and so here we have um [12:37] something where we can bring our [12:38] partners who you know have a great [12:43] product uh that's culturally local but [12:46] really really interesting and make it a [12:48] global product by you know not competing [12:51] uh with their allies but partnering with [12:53] their allies and and I think that's a [12:55] lot of what the program is about [12:56] >> and I want to tie what Benja said back [12:58] to the values point because people often [13:01] ask us you know Chinese open source [13:03] models, how do you compare and contrast? [13:06] So in that case, they didn't have to [13:09] worry that when the content was dubbed [13:12] into different languages that if the [13:14] speaker said something sensit sensitive [13:16] about tanaman square or sensitive [13:19] regarding the Chinese government that [13:21] using an American model that content [13:24] would be translated and changed along [13:27] the way, right? They knew it would be [13:28] true to what was said. And that's what [13:30] we mean when we talk about American [13:32] values of not censoring, not injecting [13:37] political will into how these models [13:40] operate in the way that one would [13:42] contrast. Now, open source is something [13:44] we we deeply value because at the end of [13:47] the day, there's all different kinds of [13:48] models and all different kinds of price [13:50] points clearly. Yeah. Um but that's [13:52] really a good example of what we mean [13:53] when we say American and western values [13:57] reflected in the models that we deploy [13:59] around our infrastructure. [14:01] >> For sure the te nollas are spicy enough. [14:03] I don't know how much the introduction [14:05] of tenement square needs to be [14:07] introduced necessarily in the story line [14:08] but you never know the border crossings [14:11] etc. [14:11] >> Good point [14:13] they keep their audience well enough. [14:15] >> Exactly. Exactly. Um but but this [14:17] actually is a core part you know um when [14:20] Py McCormack wrote the article about us [14:21] earlier this year he he called us a [14:23] power broker and power being part of an [14:26] offering that we enable our founders to [14:29] tap into and for a lot of companies [14:31] >> you know starting a Mexico office for [14:33] example is probably not one two three [14:36] maybe even top 10 on their list but if [14:39] they have a deal that we're able to walk [14:41] in [14:41] >> and and effectively enable them a shot [14:44] at winning way earlier in the life cycle [14:46] cycle, they might prioritize that. So, [14:47] so maybe if you want to riff on that, [14:49] Raun, Ben, on on particularly how that's [14:52] changed for founders, they think about [14:53] international when they have the benefit [14:55] of being able to go right to the buyers [14:57] in a much more adept way than kind of [14:59] figuring out a market from scratch. [15:02] >> Yeah. I mean, a lot of markets are very [15:04] >> topheavy, [15:05] >> right? So you take any mature business [15:08] and you say how much of your revenue [15:10] comes from even companies like I mean [15:12] countries like UK a significant chunk is [15:15] going to come from the very very top [15:16] companies in UK right so the reality is [15:21] whichever country you look at it's five [15:23] or 10 companies that matter and of [15:25] course the government is a big [15:26] influencer and buyer so the example that [15:29] you gave in Mexico is great right I mean [15:31] there are a few global companies in [15:33] Mexico like the one that Ben talked [15:35] about it and so the only way you could [15:37] it is not worthwhile setting up an old [15:40] office and serving all of the countries, [15:42] right? But if you could get into the top [15:44] five countries in sorry companies in [15:46] Mexico that probably is worth a lot more [15:49] than setting up a full operation in some [15:51] other uh country. So it is the topheavy [15:54] nature and the relationship heavy nature [15:56] that helps in this uh uh helps our [15:59] portfolio of companies, [16:01] >> right? And and the thing [16:04] there's a couple of other aspects that [16:05] are important. One is like you're right [16:07] that Mexico probably isn't a top 10 [16:10] thing that most companies go into first. [16:12] And and by the way, that's a terrible [16:14] shame because [16:16] Mexico is our neighbor. uh and from a [16:20] you know when we think about it from a [16:21] national security standpoint how [16:24] powerful would it be to have a western [16:26] hemisphere supply chain um kind of [16:29] ecosystem that was really really well [16:32] connected so I think it's kind of bad [16:34] for the country that it is number 10 um [16:37] secondly [16:38] you know to open a country uh as a [16:41] company is a very large expense you have [16:45] to set up an entity you have to hire a [16:47] team, you have to like localize the [16:49] product. And you know, that's going to [16:52] be, you know, depending on the country, [16:55] depending on how committed you are to [16:56] it, it could be like easily five or 10 [16:59] million just to get to your first deal. [17:02] Um, and so that is going to delay in [17:05] your life cycle when you can go [17:07] international. But what if, you know, [17:10] what if you before you ever uh put up an [17:14] entity or ever kind of got started in [17:16] that country, you had an opportunity, a [17:19] big opportunity, a $5 million [17:21] opportunity, a $10 million opportunity [17:23] that makes that investment much easier [17:24] to do earlier. And so we can really [17:27] enable both our companies and our allies [17:31] uh by kind of paving the way for that [17:33] kind of thing. A great example is um [17:37] like nobody's going to Saudi Arabia [17:39] first. Uh however like Saudi Arabia is [17:43] booming um and there are great [17:45] opportunities there and they really need [17:47] our help uh in developing the country. [17:50] They're modernizing which is incredibly [17:52] important to the security of the world [17:55] that you know Saudi Arabia is going from [17:57] this uh kind of tribal [18:01] different kind of society I would just [18:02] say to kind of a modern you know like [18:04] highly civilized you know player in the [18:07] modern world and full ally of the United [18:09] States like that transition is really [18:11] important that it succeeds. So we're [18:13] able to take a company like Flo, Adam [18:16] Newman's company, bring them to Saudi [18:18] Arabia, introduce them to the right [18:20] people in the government, introduce them [18:22] to the right people in business, [18:23] introduce them to the right investors. [18:25] You know, it's the optimal place to live [18:28] if you're an expat, you know, kind of [18:30] moving to Riad, help developing develop [18:32] the economy. Um, and it's something that [18:34] just it's not really possible without an [18:37] effort like this. [18:38] >> For sure. For sure. Maybe let's talk [18:40] about geos we're prioritizing because [18:42] there's a handful. Um, Japan, Korea, the [18:45] Middle East, Mexico, Canada. Are there [18:48] themes as we think about different [18:50] countries that make more sense versus [18:52] others? [18:53] >> Yeah, this is this is timely Ragu and I [18:56] have been debating this on. [18:59] So, so I would just say like I think you [19:01] know from our standpoint [19:04] it's very powerful when we when they're [19:07] a strong ally of the US. We've got a [19:10] relationship with the uh people in [19:14] government. We've got a relationship [19:15] with the people in business. We've got a [19:17] relationship with uh the investors. You [19:20] know, because those those things are not [19:22] separate. They they all interact with [19:24] each other. And if we, you know, have [19:27] that, then we can uh really really [19:31] accelerate the opportunities um for them [19:33] and for us. I think what I would just [19:36] say to that point building on Ben's [19:38] point and then I'll turn it over to Ragu [19:40] is what's changed so much as we talked [19:43] about a moment ago is that on the on the [19:46] global stage whether it's in the context [19:49] of or whether it's in the context of [19:52] countries pressuring other countries via [19:54] supply chains there are parts of the [19:56] world which we have close allies who are [20:00] now in increasingly tenuous positions [20:02] >> so if we look at the Indopacific region. [20:05] China has vastly increased its military, [20:07] vastly increased the number of nuclear [20:09] capabilities that it has uses various um [20:13] their ship building capability are [20:15] orders of magnitude more than the United [20:17] States and our allies in the region have [20:19] started responding. Japan has changed [20:21] its really sacrosan policy from in the [20:25] postw World War II period to say it must [20:28] have a capable military. [20:30] >> Yeah. It wants to build the military to [20:32] fight today's wars, not yesterday's [20:34] wars. It doesn't necessarily have the [20:37] military-industrial base. Now, that cuts [20:39] both ways. As we know from the own kind [20:42] of the halting steps the United States [20:44] military has been taken, the number of [20:45] defense reform policies that have been [20:48] written in the last 20 years, they all [20:49] say the same thing, right? Fewer small [20:51] numbers of exquisite platforms, more [20:54] cheaper, larger numbers of platforms in [20:56] addition to very exquisite platforms for [20:58] certain specific use cases. So as Japan [21:00] thinks about building a capable military [21:02] in order to deter aggression in the [21:05] region, it is looking to partner with [21:08] new firms building autonomous ships, [21:11] building AI for an integrated battle [21:13] space and they want to build with us. [21:15] They don't want to just buy because they [21:16] have their own capabilities they want in [21:18] their own country, their own supply [21:20] chains to protect. So I think a part of [21:22] that is thinking about those regions of [21:23] the world. I talked about one in the [21:25] Indopac region that covers Australia as [21:27] well. Obviously, certainly the Middle [21:29] East has been an area of American [21:31] interest for a long time. Israel is an [21:33] incredible capable ally with capable [21:34] military and intelligence community. The [21:37] UAE is as well. Saudi Arabia is [21:39] modernizing and thinking about AIN infra [21:43] how it builds an innovation, [21:44] entrepreneurial innovation base [21:46] >> in order to compete globally. And then [21:48] certainly when we think about Latin [21:50] America, when we think about fentinel [21:52] coming into the United States, it comes [21:54] in via Mexico. So public safety, [21:57] national security, how autonomous ships [21:59] can help a vast Mexican coastline [22:02] >> are not only important to our ally and [22:04] to the stability of Latin America, it's [22:06] important to us as well. So I think [22:08] that's at least the national security [22:09] lens we bring to some of our [22:11] partnerships in American dynamism or in [22:14] the AI infraspace. [22:16] >> Yeah. I mean it's the reason Ben and I [22:19] have been debating is there are so many [22:20] sides to this that you got to bring it [22:22] all together. Mhm. [22:23] >> Firstly, [22:25] um to to Ann's point, you got to be a [22:28] lot it has to make a lot of sense from [22:31] America's interest point of view, right? [22:33] And that is abundantly existing in all [22:35] these countries. These governments are [22:38] very AI forward and they're putting in [22:41] policies that are very [22:42] modernizationcentric whether it's in [22:44] defense or across the economy. Thirdly, [22:48] they are all topheavy economies. I mean [22:50] you go hit the the top companies you are [22:53] actually having global influence. For [22:55] example all those North Asian countries [22:56] that you talked about I think 15 20% of [22:59] the global 2000 are across Japan, Korea, [23:03] Taiwan. [23:03] >> Y [23:04] >> right. Um and then lastly for our [23:07] startups it's this the way business is [23:09] done in these countries is very [23:11] different than let's say UK or some [23:15] English-sp speakaking country right. So [23:18] the value ad that we can provide to [23:20] these to our startups and the value ad [23:23] we can provide to these countries is so [23:25] remarkably higher than in other places. [23:28] >> Yeah. [23:28] >> So when you put it all up and score it [23:30] out that's why we go into these [23:32] countries. [23:32] >> Yeah. And some of these countries are [23:34] also leaprogging too where they didn't [23:35] have to deal with the prior [23:36] >> Exactly. Like technology like Saudi you [23:39] go through the uh visa entry and it [23:40] takes 30 seconds like way that snap [23:43] wasn't good. [23:44] 30 seconds to go through the the visa [23:46] entry which is like unbelievable and you [23:49] know here you get stuck behind uh [23:52] unfortunately a very long line [23:53] increasingly uh for for a number of [23:55] different other reasons but you know [23:57] it's a it's a different kind of uh [23:59] approach to how you think about [24:01] incorporating technology in everyday [24:02] life but it's not evenly distributed. So [24:04] what are ways we can accelerate that in [24:05] different avenues is is also the [24:07] opportunity that a lot of these [24:08] countries are seeing as as a huge log [24:10] up. Um it's very different though when [24:12] we think about international [24:13] priorization compared to maybe other VC [24:15] firms because the typical VC firm was [24:17] that you know then you've talked about [24:18] this before like the typical way that [24:21] other VC firms have done international [24:23] have always been investment driven [24:25] versus actually thinking about the [24:27] portfolio companies and what they would [24:29] like to do internationally. So maybe [24:30] speak a little bit more about that [24:32] contrast and and quite frankly why for [24:34] the first 10 years of the firm it was [24:35] very rare we even did an international [24:37] deal outside of of not just the US but [24:40] even outside of the Silicon Valley. [24:41] >> Well I think we've always looked at [24:43] investing through the lens of what do we [24:45] bring to the table. Um so [24:50] as as a result like if you think about [24:52] what we try to do for a company um you [24:55] know we really try to help them network [24:57] to the right talent to the right [25:00] customers um to you know the right [25:04] expertise and that is very hard to do in [25:09] kind of multi- [25:12] geographies even like San Francisco [25:14] Menllo Park in the early days we were [25:15] like like let's just have a really good [25:18] product for this [25:19] >> and then we'll expand from there. [25:21] Whereas I think [25:22] >> other VCs um think of the product much [25:25] more as the money. [25:26] >> Um [25:27] >> and so if the product is the money then [25:30] the international strategy is okay we'll [25:33] find a team of investors in say the UK [25:37] and we'll help them raise money and then [25:40] we'll take a percentage of their [25:42] winnings for helping get them their [25:44] product. uh and you know and that makes [25:47] sense in that model. I think it doesn't [25:50] make sense for us because [25:53] you know our our whole thing is how do [25:55] we help you build a better company? How [25:57] do we you know how do we accelerate your [26:00] opportunity or growth and so forth. So [26:02] we needed to get to kind of a certain [26:05] level of size and stature for that to [26:08] make sense in more regions. Um but now [26:11] that [26:12] you know now that we are um it starts to [26:15] make a lot of sense and I think you know [26:17] in our role in investing we don't look [26:19] at it as how do we snatch up the [26:22] opportunities in Sweden or Japan or or [26:26] or or Saudi or UAE or that kind of [26:29] thing. We look at it as what is your [26:32] ecosystem missing [26:34] >> uh in terms of kind of capital and [26:37] expertise to go global and is there an [26:40] opportunity for us to plug in. So we [26:43] don't want to displace their a- round [26:45] and seed investors necessarily. We want [26:47] to kind of say okay if you've got [26:50] companies that are emerging as real [26:52] global uh companies then that's where we [26:55] can really step in and help. And if [26:57] we're already in the uh country and [27:00] we're already doing a lot of business [27:02] and we're already kind of understanding [27:05] um them and their entrepreneurs and so [27:06] forth, then we're in great position to [27:08] do that. But we would do that directly. [27:10] We're not raising money for somebody [27:12] who's not part of our firm, our culture, [27:14] our mission and having them go run a [27:17] play. That that that's just not us. And [27:19] also so much of the time when you think [27:21] about these um entrepreneurial [27:23] ecosystems, what they're oftentimes [27:25] missing is growth capital. They actually [27:27] have a great talent base on the early [27:29] stage side, but they're kind of missing [27:31] that component to actually go global and [27:33] they become a localized success versus [27:36] you actually thinking about global [27:37] ambitions. Maybe that actually drawing [27:38] from your experience at at VMware [27:40] Reguing about the evolution of company [27:43] building. How's that kind of [27:45] particularly changed in the the ambition [27:46] for local companies to become global [27:48] over time? [27:49] >> That's dramatically increased, right? [27:51] >> Yeah. [27:52] >> Um, previously, like I said earlier, you [27:55] would not even dream of going global [27:57] until you're like [27:58] >> really well advanced and now because [28:01] your product can be there is such a [28:04] universal appeal for you. If you've got [28:05] a great product, it's such a universal [28:07] appeal whether it's B2B or a B TOC [28:09] product, it behooves you to go earlier. [28:13] Mhm. [28:13] >> What is missing is the supporting [28:15] ecosystem, right? [28:17] >> And being a finding a way to plug into a [28:20] new country, right? And that's where our [28:23] new offering uh really focuses on, [28:26] right? So, it's really help them get the [28:28] fast start in these regions. Y [28:30] >> and for that you need to have you need [28:32] to be in the region to have the [28:34] relationships to build the trust to [28:36] build the credibility. All of those [28:38] things don't happen overnight and that's [28:40] where we can provide that. [28:42] >> Yeah. Yep. And then by spending time [28:43] locally as well, we get to know the [28:44] local ecosystem and then see the great [28:47] entrepreneurs from those regions and [28:48] then to the extent to which it makes [28:50] sense actually fund those and then help [28:52] them go international leveraging our [28:53] network as well. [28:54] >> So businesses in these countries are so [28:56] much a collective ecosystem [28:58] >> rather than an individual company going [29:00] in and dominating something there. [29:03] >> For sure. For sure. um maybe speak a [29:05] little bit more about the prior tech [29:07] cycles of of um adoption and and [29:09] particularly with AI. Ann, this is a [29:11] great question for you. There's so many [29:12] nuances around security and and even the [29:15] debate between open versus closed that [29:17] there are considerations of whether [29:18] that's even viable to build without [29:20] upsetting the Apple card on what is [29:22] allowed with open source given the [29:24] security nuances. Maybe speak a little [29:26] bit more about how you've been thinking [29:27] about that and advising companies on [29:29] that that delicate balance. [29:31] >> So, it's a hot topic right now. um as [29:33] you know and it's a hot topic around the [29:35] world because fundamentally as many [29:37] technologies are AI is dual use right so [29:42] the first step let's use cyber as an [29:44] example right the first step whether you [29:47] are an attacker or a defender is finding [29:50] vulnerabilities in systems and then if [29:51] you're a defender you patch them if [29:53] you're an attacker you exploit them [29:54] >> y [29:55] >> so models make the step of finding [29:58] vulnerabilities in code far easier Mhm. [30:02] >> Um, and quite frankly, it makes the next [30:05] steps, exploit or defend, easier as [30:07] well. The economics of cyber defense [30:10] were broken well before, you know, the [30:12] latest AI models came out. I saw that in [30:15] the White House where my nightmare was [30:16] the ransomware attacks against, you [30:18] know, rural hospitals in America. There [30:20] wasn't another hospital two blocks down. [30:22] So actual Americans opportunity to [30:24] access health care was disrupted by in [30:27] some cases states state harbored [30:29] criminal groups halfway around the [30:30] world. Right? So infrastructure is very [30:33] fragile and the reality is that the [30:36] models help both. I personally believe [30:39] I'm focused on cyber security for a [30:40] moment that they help far more in [30:42] defense because defense is far harder. [30:44] You have to be you know defending a [30:45] broad expanse and attacker has to find [30:48] one way in. Um so as a result I actually [30:51] think we have we are in a difficult [30:53] space at this moment because there's [30:55] such a broad attack surface there's so [30:58] many systems that are vulnerable but I [31:01] think we will see the opportunity for [31:03] much more secure space to your point you [31:06] know when I talk with governments around [31:08] the world or companies around the world [31:10] all are grappling with that question of [31:13] I'll talk about the government side [31:15] given we know it can both help Most [31:19] countries have two missions, right? [31:20] Defending the country as well as [31:22] defending it, you know, as well as [31:24] maintaining its interest overseas. Given [31:27] both, how do we ensure that these [31:28] powerful capabilities don't fall into [31:30] our adversaries hands, don't fall into [31:32] criminals hands to be used for that [31:34] purpose? And I think, you know, in the [31:38] traditional cyber tools arena, we always [31:40] grappled with this, right? When a [31:42] capability is on a thumb drive, [31:44] >> export controls are a joke. You can't [31:46] actually control it in that way. [31:49] And there is real power, as we talked [31:51] about earlier, in the models that the [31:54] world runs on being run on trusted [31:56] models. When people are using an [31:58] American model, they don't have to worry [32:00] there's an intentional backdoor. As we [32:02] talked about earlier, they don't have to [32:03] worry that a chatbot's answers will be [32:05] framed by, you know, censoring to fit a [32:09] particular political story of like an [32:11] authoritarian government. So the [32:14] question for now is to say how do we [32:16] build in the degree of safety in model [32:22] so that they can be used in environments [32:24] that need that safety and [32:27] that is somewhere where because the main [32:30] models are American we're in a [32:32] particular position to think through and [32:34] build that out and I think this is not [32:36] done by government it's not done by the [32:38] private sector what it really needs is [32:40] people who really understand the models [32:42] best from the private sector at the [32:44] table with folks on the government side [32:46] who are thinking about how do we both [32:48] promote innovation and control for these [32:49] risks and building out the right set of [32:52] government policies to enable both the [32:55] use as well as the understanding of how [32:57] do you protect against the risks [32:59] >> that come with it with the models as [33:01] well. [33:01] >> Yeah. Yeah. As our partner Mark often [33:03] says, we technology is the dog that [33:05] caught the bus. And we've never seen [33:07] before that countries are de, you know, [33:09] kind of feverishly demanding the latest [33:11] versions of the model because they [33:12] recognize how competitive it is to their [33:14] economy, to their people, to their [33:16] interests. And this now is unlocking all [33:18] sorts of new paradigms as we think about [33:20] globalization in a way that that feels [33:22] categorically different than just [33:23] selling product internationally, right? [33:25] And and this is hopefully then an [33:27] opportunity for many of our founders as [33:29] well. [33:29] >> No doubt. Although it is complicated. Um [33:33] like so for example you'd like to like [33:36] from a government standpoint you'd like [33:37] to say okay don't release your model [33:40] with cyber hacking capabilities. [33:43] But [33:44] if you do, as an say and you want to [33:48] find the vulnerabilities in your own [33:50] code and patch them, [33:53] you can't do that without also enabling [33:56] cyber attacks because as soon as it can [33:58] find the vulnerabilities, guess what? [34:01] Uh, and so you're kind of in this [34:03] situation where it's like, well, if you [34:08] kind of lock that up, lock that [34:11] capability to understand [34:12] vulnerabilities, then the only people [34:15] who will be able to understand v [34:17] vulnerabilities are the bad actors who [34:18] jailbreak the system. [34:20] >> And so you you we're in a very [34:23] >> I would say interesting interesting [34:25] world on that. [34:26] >> Yeah. And and I think to that point, [34:28] we'll focus on the cyber aspect because [34:30] that's gotten so much attention with [34:33] anthropics mythos model, etc. I think [34:36] what AI models essentially do to Ben's [34:38] point is you want to be the first to [34:40] hack your systems [34:42] >> because, you know, you we often use the [34:44] expression of, you know, digital doorork [34:46] knobs. It now allows attackers to jiggle [34:48] every doorork knob continuously and at [34:49] scale. So your hope for your network is [34:52] using those models first to find where [34:54] you're vulnerable and frankly then be [34:56] able to fix it and that wasn't a [34:58] tractable problem before y [35:00] >> and I think now the models help you [35:03] particularly for code bases that were [35:04] public particularly for if you're [35:07] relying on third party code or open [35:08] source code and that's some of the [35:10] biggest areas of risk or where there is [35:12] dependencies on code that nobody was [35:14] really maintaining. So I think we're now [35:16] seeing you know companies coming [35:18] together taking responsibility for that [35:20] because now the I would say the cost of [35:25] not addressing insecure code has gone up [35:28] and the cost of addressing it has gone [35:30] down. [35:30] >> Yeah. [35:31] >> Which is a good shift from where we were [35:33] before. [35:34] >> We we we may finally build secure [35:36] systems. [35:37] >> Exactly. [35:38] >> Please God. And more importantly, the [35:41] the process for building new code can [35:44] incorporate these as well. [35:46] >> And it's kind of very tricky from an [35:47] industry standpoint because like [35:49] originally [35:51] cyber hacking was vandalism, right? [35:54] Like, okay, I'm going to mess up your [35:55] PC. Then, you know, with the internet, [35:59] it became theft and now it's war. Uh, [36:02] and [36:03] >> that evolution kind of, I think, snuck [36:05] up on people. And so, and we've seen, [36:08] yeah, I mean, the great thing when it [36:10] went from vandalism to theft, the the [36:12] vandalism company, Microsoft, is one who [36:15] created all the theft opportunities by [36:17] putting Windows on the internet when it [36:20] wasn't ready to be on the internet um [36:22] because it wasn't built for that kind of [36:24] idea. And uh you know, with ActiveX and [36:27] all these kinds of technologies that let [36:29] people come into the internet and get [36:30] right to your machine. Uh, and I think [36:33] that we're kind of in a period like that [36:35] where we've got all this technology that [36:37] was built in the pre-AII era that's not [36:40] nearly secure enough for the AI world. [36:42] And um, and we're going to have to make [36:44] that transition. [36:46] >> Yeah. There's a 20 year legacy of code [36:48] that has to be tackled. [36:49] >> Yeah. [36:50] >> Nope. [36:52] >> That AI is increasingly [36:54] >> many opportunities. Yes. Many [36:55] opportunities. Yeah. [36:56] >> Many many. [36:58] >> Um, awesome. Well, maybe in closing, so [37:01] Ben, I I know the number one question we [37:02] often get when we're visiting different [37:04] countries is how do you recreate Silicon [37:06] Valley locally, right? That's let's like [37:08] they want the ingredients to it. And [37:10] there there's a few components to [37:12] success that have enabled Silicon [37:14] Valley, but there's also the the [37:15] proposition, what if the internet is [37:16] actually Silicon Valley or this the the [37:19] actual Silicon Valley just distributed. [37:21] So [37:22] >> yeah. Yeah. [37:22] >> I don't think that's true, by the way. [37:25] And like they think that's a happy talk. [37:28] If you you know like one thing you learn [37:31] when you go to different countries is [37:32] they're very different. Like the laws [37:34] are different, the policies are [37:35] different, the governance are different. [37:36] And that's why you know America has [37:38] continued to be special. We've had the [37:40] internet now for quite a long time and [37:41] like that's absolutely not the case. [37:44] >> Uh so like so what is going on? And and [37:47] by the way, I wish we could bring uh [37:49] some of our own uh policy makers on [37:52] those trips so they could hear people [37:53] say we want Silicon Valley before they [37:55] keep trying to wreck us. [37:58] Not all of many of our policy makers are [38:00] outstanding, but some really don't like [38:02] the fact that we have a technology [38:04] industry. [38:05] Um, so I think that like as we've [38:08] traveled around there's a kind of a a [38:11] combination of ingredients and and they [38:13] are a little bit the combination that [38:15] makes Silicon Valley special and then [38:17] you know America itself special. So the [38:20] first is um you do need the talent and [38:24] uh [38:26] talent is kind of easily kind of most [38:30] easily seen in like is there a great [38:32] technical university um that graduates [38:35] people who know how to build things. Uh [38:38] that that that's kind of like an [38:40] essential thing. And like by the way [38:41] many countries have that that don't have [38:43] a Silicon Valley. Um and what are they [38:45] missing? They're missing the other two [38:47] components. The second is do you have a [38:51] set of laws and policies [38:54] that uh facilitate entrepreneurship are [38:57] essentially good for business? Like can [39:01] you create a company easily? Can you [39:04] hire people easily? Um can you fire [39:07] people easily? Is there a tax incentive [39:10] structure that's not ownorous? Can you [39:14] um own and this is a big one lately [39:18] which uh you know California is kind of [39:21] moving against but can you own an [39:24] illquid asset that has value um [39:28] >> or will it get seized uh and so I think [39:32] Norway uh introduced a um unrealized [39:35] capital gains tax and that that kind of [39:37] effectively ended the tech ecosystem [39:39] there. Uh, and I think that that, you [39:42] know, so those those kinds of policies [39:43] are super important. If they're not [39:45] right, nobody is going to take the [39:47] extreme risk um kind of both personal [39:50] and financial starting a company if it's [39:53] not an environment that's conducive to [39:55] that. And then the third one which is I [39:57] think kind of a combination of [39:59] everything in a way is is the culture [40:02] such that [40:05] young people the most capable young [40:08] people willing to make the biggest [40:10] contribution [40:12] is do they get status social status [40:16] reward is it cool to either start or [40:18] join a new company um and almost no very [40:22] few countries have that uh I would say [40:24] that it's really really good in the US. [40:26] It's um you know it's clearly good in [40:29] China by the way. Uh although they've [40:31] done a few things to mess it up, but in [40:34] most countries you know like in a lot of [40:36] ways like success is frowned upon. [40:38] >> It's like oh you know and in a lot of [40:40] you hear a lot of this in the political [40:41] rhetoric. Oh the only way you could make [40:43] a billion dollars is if you're a thief [40:45] >> which is completely absurd. Um [40:48] >> you know you just look at the people [40:49] almost none of them are thieves. like [40:51] they you can't steal a billion dollars [40:53] is a real truth. But like that's just [40:55] like it's a rhetoric that works because [40:57] if you're you know if you're like how [40:59] >> it's a great sour grapes idea, right? [41:01] Like it's like how how did he get all [41:03] that money? Oh, he's a thief. Okay, I go [41:05] I'm not a thief. So like I feel much [41:07] better about myself now. But but that's [41:09] the you that's a lot of the thing and [41:12] that's the thing that's so hard to [41:13] replicate and it's so hard to build and [41:16] it's so easy to destroy. Uh so it's you [41:19] know something that I think we have to [41:21] just be aware of how special that is and [41:24] unique in the world and then we can try [41:26] and help other people do it but we also [41:28] have to try equally hard not to wreck it [41:30] ourselves. [41:31] >> Yeah, for sure. It's uh it's World Cup [41:33] season and there was a great tweet where [41:35] someone was like, "I think America [41:36] thinks they could potentially win the [41:37] World Cup." And like it's like maybe [41:39] they can, maybe they can't, but like the [41:41] fact that America thinks they can is a [41:44] feature, not a bug. [41:47] Yes, [41:47] >> there's a reason [41:48] >> amition is fantastic. [41:49] >> There's a reason we all read our kids. [41:51] You know that what is it? The little red [41:52] jing. I think I can. I think I can. I [41:55] think I can. That's in America. You can [41:57] believe you can. [41:58] >> That's right. That's right. Amazing. [41:59] That's a good note to end on. Thank you [42:01] all.