Dell and Kalanick Shift Focus to Physical AI and Wealth Transfer at SXSW

戴尔与卡兰尼克在SXSW聚焦物理AI与财富转移 Michael Dell与Travis Kalanick认为，未来十年的经济主导权取决于物理世界基础设施的自动化，以及通过让每个美国儿童在出生时获得股权来捍卫资本主义。 本期节目在德克萨斯州奥斯汀现场录制，深入对话了Uber联合创始人Travis Kalanick和Dell Technologies创始人Michael Dell。Kalanick结束隐身模式，披露了他的新创企业Atoms，详细阐述了他在餐饮、采矿和机器人领域的物理AI与现实世界自动化愿景，并反思了他从洛杉矶迁至德克萨斯的经历。Michael Dell讨论了推动Dell营收扩大至预计1400亿美元的规划，概述了公司在AI服务器基础设施领域高达500亿美元的巨额押注以及本地化边缘AI的兴起。节目最后，Dell与Brad Gerstner解释了“Invest America”项目，这项规模达62.5亿美元的创纪录慈善行动旨在为2500万儿童设立初始投资账户，以捍卫自由市场资本主义。 Travis Kalanick 正式结束长达七年的隐身模式，宣布将母公司 City Storage Systems 正式更名为 Atoms，致力于通过物理自动化变革现实世界的物流与重工业。 (0:00) Kalanick 提出“基于原子的计算机”框架，并透露 Atoms 即将完成对自动驾驶工业设备初创公司 Pronto 的收购，以将其物理 AI 技术栈拓展至采矿等重工业领域。 (7:20) Michael Dell 回顾了 42 年前以 1,000 美元在宿舍创立戴尔的经历，预计公司年营收将达到约 1,400 亿美元，并深入探讨了得克萨斯州吸引企业与创始人的商业环境优势。 (36:03) 戴尔详述了其 500 亿美元规模的 AI 基础设施战略，其推出的“Dell AI Factory”企业级架构已被逾 4,000 家客户采用，以支持端侧与私有化模型的运行。 (40:30) Michael Dell 与妻子 Susan 宣布向“Invest America”计划捐资 62.5 亿美元，为全美 2,500 万名儿童每人注资 250 美元建立投资账户，推动全民所有制与长期财富积累。 (62:51)

Travis Kalanick and Michael Dell argue that technological progress is shifting from purely digital software toward physical automation, localized enterprise compute, and systemic economic realignment. Speaking in Austin, Texas, both founders emphasized the need to build specialized physical infrastructure, overhaul legacy corporate workflows, and expand direct capital ownership to sustain American innovation. Automating Physical Atoms and Robotics Infrastructure (3:16) Travis Kalanick announced that City Storage Systems has rebranded to Atoms, outlining an operational model that mirrors digital computing in physical space: manufacturing functions as processing, real estate as memory, and logistics as transport networks. Beyond food preparation spaces like CloudKitchens, Atoms is near closing the acquisition of San Francisco autonomous equipment startup Pronto to automate heavy mineral extraction. Kalanick argued that autonomous operations alter the economics of mining by lowering safety hazards and labor footprints in remote deposits. Addressing robotics and self-driving vehicles, Kalanick cautioned against humanoid forms optimized for kinetic demonstrations, advocating instead for specialized, non-humanoid robots designed for concrete tasks like food packaging. Analyzing autonomous vehicle platforms, he noted that systems like Waymo consume roughly 100 times more energy than the 100-watt human brain. To bridge this divide, Kalanick argued developers must adopt Vision-Language-Action (VLA) compression models to filter out background irrelevancies and target critical navigation parameters. Enterprise AI Factories and Edge Decentralization (40:30) Michael Dell described an unprecedented acceleration in enterprise hardware, noting that Dell’s AI infrastructure business grew from $2 billion to $10 billion, $25 billion, and is currently tracking toward $50 billion annually following the launch of Nvidia H100 servers. Through its 'Dell AI Factory' architecture, adopted by over 4,000 enterprise and sovereign customers, Dell enables organizations to run AI models on proprietary internal data. Even so, Dell estimated that only 10% to 15% of large enterprises have truly overhauled their operations around AI, warning that brand equity and existing customer relationships are rapidly depreciating assets for slow-moving incumbents. Dell also forecast a widespread decentralization from public clouds to edge inference, driven by high centralized compute costs and latency. With 10,000 customers embedding Dell systems into medical devices, industrial facilities, and telecom gear, computing is shifting directly to the point of data creation. Dell emphasized the rise of open-source architectures—including models like Gemma and Nemotron alongside local frameworks like Ollama—which allow enterprises and individuals to retain proprietary data security without transmitting information to centralized providers. Regional Governance and the Business Exodus to Texas (16:34) Both founders framed the migration of technology companies from California to Texas as an inevitable response to divergent governance and economic incentives. Kalanick detailed his decision to relocate to Austin, contrasting the region's lower living costs and forward-looking environment with San Francisco's governance, which he criticized for inefficient infrastructure outlays and lagging public safety enforcement. When asked about legislative reform in California, Kalanick favored direct ballot initiatives over lobbying to bypass political stagnation. Dell pointed to Texas’s long-standing regulatory stability, lower taxes, and access to energy and land as key drivers for data center development and industrial manufacturing. Contrasting the rapid ground-up construction of Tesla's Gigafactory in Texas with the constraints of retrofitting older factories in California, Dell noted that Texas now accounts for 10% of U.S. births and hosts more NYSE-listed companies than any other state. Scaling Direct Ownership Through Invest America (62:51) Michael Dell and investor Brad Gerstner outlined the 'Invest America' initiative, intended to build an ownership base for citizens who feel disconnected from national wealth growth. The effort is initiated by a $6.25 billion personal pledge from Michael and Susan Dell, funding $250 investment accounts for 25 million children between ages 2 and 10 in zip codes with median household incomes of $150,000 or below. The program complements a federal legislative provision passed in the April 2025 reconciliation bill that establishes an account framework for every child born on or after January 1, 2027, tied directly to their Social Security number and seeded with $1,000 in federal funds subject to four-year reauthorizations. Gerstner framed the initiative as a nonpartisan effort to safeguard market capitalism, pointing out that fewer than half of Americans under 40 view capitalism favorably. Governed by a Giving Committee co-chaired by Dell and Gerstner, the program aims to channel $5 trillion into family-held assets over 15 years. Accounts are linked to a mobile interface displaying fractional shares of prominent S&P 500 companies, enabling family contributions via QR codes and corporate stock endowments. While Kalanick and Dell present physical automation, local AI infrastructure, and universal investment accounts as viable remedies for economic stagnation, significant challenges remain. These include managing the disproportionate power demands of physical AI compute, overcoming deep organizational inertia among the 85% to 90% of enterprises lagging in AI adoption, and maintaining bipartisan consensus to preserve federal seed funding for universal child accounts.

Travis Kalanick 与 Michael Dell 分别从物理世界自动化与数字化硬件基础设施两个维度，阐述了技术与实体产业的深度融合逻辑；同时结合德克萨斯州对创新生态的吸引力，以及旨在通过建立全民儿童投资账户来重塑大众所有权的“投资美国”倡议，探讨了资本、政策与实体产业转型的交汇。 物理自动化与 Atoms 体系 (3:16) Travis Kalanick 正式宣布将其母公司 City Storage Systems 更名为 Atoms，结束长达七年的潜行运营。Kalanick 提出了“基于原子的计算机”理论框架，将传统以 CPU、内存和网络为主的数字计算，类比为由制造、房地产与物流构成的物理处理体系。他主张，餐饮外卖无法仅依附于产能利用率仅约 20% 的实体餐厅，必须建造类似亚马逊履约中心的高负荷专用基础设施以平抑成本。基于此逻辑，Atoms 正在布局餐饮厨房、矿业开采及非人形特种移动机器人等细分工业载体，并接近完成对旧金山自动工业设备初创公司 Pronto 的收购。 在技术路径上，Kalanick 指出双足人形机器人在动作展示中消耗超过 1200 瓦电力，能效严重偏离实际任务需求，其研发核心集中于执行特定打包分拣任务的非人形轮式硬件。针对自动驾驶领域，他指出人类大脑进行空间导航仅需 100 瓦电力，而 Waymo 等自动驾驶方案能耗高出约百倍。他主张引入基于语言的高效压缩模型（如视觉-语言-动作 VLA 模型）以过滤无效计算，并认为具备全垂直整合能力的 Tesla 已成为该领域的标杆，迫使业内初创公司直面其竞争壁垒。 区域营商环境与全球制造业竞争 (16:34) 围绕科技产业从加州迁往德克萨斯的趋势，Kalanick 与 Michael Dell 强调，奥斯汀具备居住空间翻倍、成本减半以及政策友好的优势。Dell 援引数据指出，全美每十个新生儿中就有一名出生在德州，德州拥有全美最多的纽交所上市公司，并在奥斯汀聚集了顶尖科研与产业转化资源；例如 Tesla 能够迅速建成得州超级工厂，而在加州则受制于旧厂房改造与低效的市政基建决策。Kalanick 批评部分城市治理抑制了社会秩序的免疫系统，导致警务士气受挫与犯罪执法缺失，促使大量科技、文化与餐饮企业向外迁移。 在审视国际制造业格局时，Kalanick 强调中国制造业中心呈现出惊人的迭代速度。他回顾深圳在二十年间从相对陈旧落后的环境超越了现代新加坡水准，以小米为代表的企业展现出极强的进取心与制造效率。Kalanick 呼吁美国不应采取破坏规则的封闭对抗方式，而应专注于提升自身竞争力，并在高技能移民政策上打破阻碍，简化全球顶尖人才进入美国的行政阻力，以维持技术和生产力层面的良性竞争。 企业 AI 工厂与边缘算力回流 (40:30) Michael Dell 梳理了戴尔 AI 服务器业务的激进扩张，该业务年收入从 20 亿美元、100 亿美元快速跃升，预计本财年将触及 500 亿美元，基础设施部门季度同比增速达 73% 并指引下季度增速达约 100%。尽管联邦政府提供针对数据中心资本支出首年 100% 的加速折旧税收激励，但 Dell 观察到目前大型企业中仅有 10% 至 15% 真正实现了底层业务与 AI 的深度融合，多数企业仍停留在表面推进阶段。与此同时，Stripe 数据显示 2025 年创立的初创公司借助 AI 原生工具，扩张速度达到 2018 年同期的四倍。 在架构演进层面，Dell 认为集中式公有云并非终局，因为高额账单正在倒逼企业寻求成本最优解，推动算力大规模向边缘和本地环境回流。目前已有超过 4000 家企业采用“Dell AI Factory”在私有数据上部署模型，还有超过 10000 家客户将戴尔算力嵌入至医疗、工业与安防等专用设备端，实现就地推理。Dell 指出，开源模型（如 Gemma、Nemotron）与 Ollama 等本地化运行工具的普及，使企业得以在私有数据壁垒之上构建差异化竞争优势，同时大幅降低对外部云巨头的依赖。 投资美国倡议与全民所有权构建 (62:51) Brad Gerstner 与 Michael Dell 披露了推动“投资美国”（Invest America）倡议落地的细节。该法案被纳入 2025 年 4 月的和解法案，规定自 2027 年 1 月 1 日起，每位出生的美国婴儿都将自动获得一个与社保号绑定的永久性投资账户，联邦初始注入 1000 美元种子资金（每四年需国会重新授权），预计每年覆盖约 370 万新生儿。此外，Dell 夫妇以个人名义捐资 62.5 亿美元，为生活在中位收入 15 万美元以下邮编区域的 2500 万名 2 至 10 岁儿童各提供 250 美元初始启动金，目前每日注册儿童人数已达约 10 万名。 Gerstner 解释，该机制被视为“社会保障 2.0”，其核心资产为解构后的标准普尔 500 指数，通过移动应用向家庭展示其持有的具体公司零碎股份，以培养长期的财务认知。推动者认为，鉴于 40 岁以下群体对资本主义的好感度已不足半数，让占人口 70% 的低财富阶层直接共享国家财富增值，是维护自由市场制度的防御举措。该计划目标是在未来 15 年动员包括企业 IPO 股票捐赠与富豪定点资助在内的多方资本，向工薪家庭账户累计输送 5 万亿美元，探索一种规避官僚损耗的直接资产积累模式。 节目展现了两位创始人对实体工业自动化与去中心化 AI 算力基建的长期押注，以及用金融所有权化解制度信任危机的尝试；然而，以物理 AI 解决现实工业瓶颈仍面临巨大的能效与硬件工程制约，而全民投资账户能否在两党政治周期与长期资本运作中稳定兑现财富代际流动的承诺，依然有待时间检验。

## Background 本期内容由知名科技投资播客《All-In Podcast》于 2026 年 3 月 17 日在 YouTube 平台发布。节目现场录制于得克萨斯州奥斯汀的 Arena Hall，聚焦特拉维斯·卡兰尼克（Travis Kalanick）与迈克尔·戴尔（Michael Dell）两位传奇创始人的深度对谈，内容涵盖卡兰尼克结束长达七年隐形模式公开新公司 Atoms 与自动化愿景、加州治理危机与得州崛起、戴尔科技年营收 1400 亿美元背后的 500 亿 AI 基础设施布局，以及迈克尔·戴尔联手布拉德·格斯特纳（Brad Gerstner）发起的 62.5 亿美元“Invest America”全美儿童资本账户计划。 ## TL;DR 特拉维斯·卡兰尼克正式打破长达七年的保密状态，系统阐释了从“食物计算机”拓展至矿业自动化的母公司 Atoms 愿景，并犀利剖析了 Waymo 与特斯拉纯视觉方案的自动驾驶竞局，同时直言加州体制与治安恶化促成其出走奥斯汀。迈克尔·戴尔随后登台，回顾戴尔从大学宿舍起步到年营收 1400 亿美元的复利奇迹，详述戴尔在 AI 算力潮中迎来的 500 亿美元业务跃迁与边缘推理演进，强调企业唯有自上而下激进重构才能抵御 AI 原生颠覆者。最后，布拉德·格斯特纳惊喜登台，共同揭晓戴尔夫妇出资 62.5 亿美元的“Invest America”史诗级慈善项目，旨在为全美 2500 万儿童建立终身投资账户，借复利重塑社会契约与资本主义信心。 ## Atmosphere & style 现场气氛极其热烈且充满硅谷老友相聚的松弛与张力，四位主持人在嬉笑调侃、内幕八卦（如私人群聊中的“达斯·弗里德伯格”）与极其严肃硬核的商业研判之间自如切换。从卡兰尼克重返聚光灯下的犀利、从容与野心，到戴尔坐拥千亿帝国却依旧谦逊幽默的得州主场从容，整场对话兼具宏大科技叙事与辛辣体制批判。随着后半段格斯特纳登台揭示白宫椭圆形办公室博弈与世纪捐赠的诞生，全场情绪从对技术变革的躁动升华为对社会阶层流动与国家未来的深沉共鸣，充满极具感染力的理想主义与笃定力量。 ## Bulletin - [05:47](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=347s) 特拉维斯·卡兰尼克（Travis Kalanick）正式结束长达七年的隐形模式，宣布母公司更名为“Atoms”，专注通过物理 AI 推进餐饮与矿山装备自动化，并透露即将完成对自动驾驶矿卡企业 Pronto 的全资收购。 - [41:13](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=2473s) 戴尔科技董事长迈克尔·戴尔（Michael Dell）透露，公司 AI 基础设施业务规模从 20 亿美元飞跃至 500 亿美元，得益于 100% 加速折旧政策红利与企业自上而下的流程重构，基础设施业务当季指引翻倍增长。 - [52:14](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3134s) 戴尔指出 AI 正在从云端向边缘扩散，“最便宜的 Token”必然是在数据源头设备就近生成，企业为避免昂贵云账单正大力采购本地嵌入式与桌面算力部署开源小模型与自主智能体（Agents）。 - [01:04:03](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3843s) 迈克尔·戴尔夫妇宣布为《投资美国法案》（Invest America Act）个人捐赠 62.5 亿美元，为人均符合条件的 2500 万全美儿童各注资 250 美元种子基金，打造出生即享的终身复利投资账户。 ## Key takeaways - [03:00](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=180s) 卡兰尼克提出“原子计算机”哲学，将比特世界的计算、存储、网络对应到现实世界的制造业、仓储地产与交通物流，以重构重工业和餐饮基础设施。 - [07:36](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=456s) 特种工业机器人区别于人形机器人，关键在于适配地形的特化底盘与具体工种落地，实体采矿自动化能打破劳动力短缺与恶劣环境限制，让深层稀有资源开采具备商业可行性。 - [11:00](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=660s) 特斯拉在物理 AI 领域占据类似早期 Google 与微软的绝对统治地位，其纯视觉路线正在挑战高难度的底层第一性原理突破；而 Waymo 虽有先发验证，却受困于整车制造成本与量产规模短板。 - [12:51](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=771s) 语言是极高密度的信息压缩工具，多模态智能体协同可大幅降低物理自动驾驶感知系统的数据冗余与算力能耗。 - [16:18](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=978s) 加州高税收、过度监管、昂贵基建与激进地方检察官对司法的破坏迫使大量科技精英出走，奥斯汀凭借宽广空间、充沛电力、低税收与 UT Austin 人才生态成为新科技制造中心。 - [25:21](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1521s) 具身智能面临严峻的能耗与硬件瓶颈，当前人形机械手噱头大于实用，投资焦点正向基础材料科学、高效执行器以及能直接创造商业收益的专用设备迁移。 - [29:23](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1763s) “资本作为武器”仅在特定高度竞争场景下是生死门槛；受中东地缘冲突与沙特股市大跌 20% 冲击，卡兰尼克旗下中东实体原定上市进程受阻。 - [32:14](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1934s) 中国如深圳等地的先进制造业与供应链展现出惊人的工程落地效率，中美之间应展开良性竞合，同时美国亟须改革合法高技能移民机制以保留全球顶尖创业人才。 - [43:49](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=2629s) AI 带来的范式转变比以往技术浪潮快 5 倍，但企业获取超额 ROI 的阻碍在于组织惯性与文化，唯有自上而下重塑业务流程与梳理私有暗数据才能避免被 AI 原生颠覆者淘汰。 - [01:00:39](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3639s) AI 的核心本质是“人类潜能与能力的放大”，能加速医疗、科学与能源创新，而并非单纯为了降本减员，当前社会公众的抵触主要源于拟人化包装带来的焦虑。 - [01:09:45](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=4185s) Invest America 推动自 2027 年起为全美新生儿自动开立绑定社保号的个人终身资本账户，预计 15 年内绕过官僚中介直接向工薪家庭转移 5 万亿美元资产，重构现有社会保障机制。 - [01:12:54](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=4374s) 通过极简可视化移动端（如类 Robinhood 界面）持有标普 500 指数股权，结合生日礼金、企业配捐与各州领养模式，帮助年轻一代重塑对自由市场与资本主义的长期信念。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=0s) 走出长达七年的“隐形模式”：全员保密的硬核创业 现场开场时，主持人 Jason Calacanis 兴奋地向观众提问自己有史以来最得意的天使投资是哪一家，台下齐声高呼特拉维斯·卡兰尼克（Travis Kalanick）的名字。卡兰尼克在热烈的掌声与欢呼声中登台。Jason 感慨道，卡兰尼克过去整整七年时间几乎完全“泡在实验室里埋头造物”。这期间 Jason 每年都力邀他参加 All-In 峰会，但卡兰尼克每次都干脆回绝，表示自己正在专心打磨产品。他甚至将整个公司置于极致的隐形模式（stealth mode）之中：不仅行踪隐秘，公司数千名员工——哪怕是极其需要向外拓展人脉的招聘人员和销售团队——都绝对禁止在 LinkedIn 个人资料上填写公司真名，全网只能挂上笼统的“Stealth”。现场嘉宾笑称，这些员工的父母一度以为自己的孩子是替中情局（CIA）执行机密任务的间谍。即便是作为早期投资人的 Jason，也被要求签署严格的保密协议（NDA），连入股的事实都不能对外公开。 卡兰尼克笑着回应，公司全员打上“Stealth”标签确实让大家在“地狱困难模式（hard mode）”中生存，甚至导致洛杉矶一度出现大量身份神秘的初创公司员工聚集的奇景。为了在全球保持隐形，他们在进驻的 30 个国家分别使用了极其普通、互不相关的本土化名称：在美国，其厨房业务被称为 CloudKitchens；在韩国则叫作 Kitchen Valley；在中东地区命名为 Namah；在拉美部分国家则取名为 Cocinas Ocultas。这种去中心化且刻意平淡的命名方式，使得外界多年来始终无法拼凑出这家庞然大物的完整全貌，而今天他终于正式打破沉默，从隐形走向公开。 - **隐形规模**：全球横跨 30 个国家开展业务，拥有数千名员工，全员在职场社交平台隐藏雇主信息。 - **各地曾用主体**：美国使用 CloudKitchens，韩国使用 Kitchen Valley，中东使用 Namah，拉美地区使用 Cocinas Ocultas。 - **投资插曲**：Jason Calacanis 虽然位列股东名册（cap table），但签署了严苛的保密协议，多年来被禁止公开自己投资人的身份。 --- ### [03:00](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=180s) 从比特到原子：构建“原子计算机”与 City Storage Systems 的诞生 在被问及从最初的“共享厨房”如何一步步演进时，卡兰尼克系统性地阐述了自己从 Uber 时期延续至今的底层哲学——“物理世界的数字化（digitizing the physical world）”。他回顾计算机科学的核心常识指出，由迈克尔·戴尔（Michael Dell）等先驱奠定的经典“比特世界（bits world）”由三大核心计算资源构成：负责处理与操纵比特的 CPU、负责存储比特的内存/硬盘（Storage），以及负责在节点间搬运比特的网络（Network）。如果将物理世界中的“原子（atoms）”等同于比特，那么人类本质上是在构建一台以原子为基础的庞大计算机： 1. **操纵原子**：对应实体产业中的**制造业（Manufacturing）**； 2. **存储原子**：对应现实世界的**房地产与仓储空间（Real Estate / Storage）**； 3. **搬运原子**：对应**交通运输与物流体系（Transportation / Logistics）**。 正因如此，他当年特意为母公司挑选了一个极其无趣、看似平平无奇的名字——城市仓储系统（City Storage Systems，简称 CSS）。卡兰尼克将他们打造的第一个实体计算集群定义为“食物计算机（Food Computer）”，即面向餐饮领域的制造、地产和物流全链路基础设施。其使命是“为更好的食物提供基础设施”，终极目标是将一份新鲜烹制并配送到家的餐食成本降到接近“去超市买菜做饭”的水平。卡兰尼克断言，一旦达成这个拐点，他们对家庭厨房造成的颠覆，就如同当年 Uber 对私家车造成的颠覆一样彻底。 然而，改造实体厨房的难度远超当年的 Uber。在 Uber 创办之初，现成的道路网络和大量闲置的私家车早已存在，创业团队某种程度上只需“在应用商店上线一个 App”即可启动飞轮；到卡兰尼克离开 Uber 时，旧金山全城整整 13% 的车辆行驶里程都来自 Uber。但在餐饮领域，传统独立餐厅通常只有约 20% 的富余产能，Uber Eats 和 DoorDash 接入后迅速将其填满，却根本无法支撑大规模、高通量、工业化级的出餐与外卖运转。这就像传统电商崛起必须依赖亚马逊自建巨型现代化仓储物流中心一样，餐饮外卖要完成向电商化的高级跃迁，必须重构底层的生产与供应链基础设施。 ```mermaid flowchart LR accTitle: 比特世界与原子世界的系统映射 accDescr: 对比比特计算的三大基础要素与原子世界中对应的制造、地产及物流产业。 subgraph Bits["比特世界（Bits）"] B1["CPU 处理"] B2["存储 Storage"] B3["网络 Network"] end subgraph Atoms["原子世界（Atoms）"] A1["制造业（Manipulate）"] A2["房地产 / 仓储（Store）"] A3["交通 / 物流（Move）"] end B1 --> A1 B2 --> A2 B3 --> A3 ``` --- ### [05:47](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=347s) 全新品牌 Atoms 亮相：进军矿业自动化与 Pronto 收购 随着底层技术与工业能力的延伸，卡兰尼克宣布母公司正式更名为“Atoms”（原子），其全新愿景被定义为：“通过物理世界自动化转型支柱产业，推动世界运转（physical automation to transform industries and move the world）”。业务版图已不再局限于最初的“食物计算机”，而是向更多承载实体经济运转的重工业领域强力扩张，其中最为核心的新赛道就是实体矿业（Mining）。 卡兰尼克强调，这里所说的采矿是实打实地从地底开采矿物和稀缺资源，其部门使命是“打造更高生产力的矿山，为地球工业提供动力”。在硬件路线上，Atoms 专注于针对工业特种场景的机器人，而非泛用的人形机器人。他解释道，特种机器人在物理世界行动的前提是必须配备适配地形的底盘（wheelbase）。虽然当下的公众视线大多聚焦在城市乘用车自动驾驶（如特斯拉 Robotaxi 或正在奥斯汀街头测试的 Waymo），但现实中有着海量需要移动作业的重型设备，采矿机械便是其中体量极其庞大的一类。为了加速在重工业自动化领域的布局，卡兰尼克透露他们正在完成对总部位于旧金山的矿业重卡自动化企业 Pronto 的收购，该交易已处于“距离最终交割仅剩咫尺之遥”的阶段。 - **全新品牌定名**：Atoms，定位为面向全球实体工业场景的自动化与智能化基础设施。 - **两大战略抓手**：继续推进餐饮自动化（Food Computer），全面开拓实体矿山机械自动化（Mining Automation）。 - **关键收并购事件**：即将完成对旧金山自动驾驶与矿山装备自动化初创公司 Pronto 的全资收购。 --- ### [07:36](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=456s) 采矿业的自动化革命：降低地缘与生存成本的终极解法 针对采矿业务的技术可行性，主持嘉宾 David Friedberg 提出了行业痛点：采矿的核心难点在于前期的勘探选址，以及向深处掘进时急剧上升的成本——能量与开采成本通常随钻探深度呈几何级数（距离平方）暴增。Friedberg 询问自动化技术究竟是在前端帮矿企精准寻找富矿，还是通过重塑深地作业的成本结构来释放产能。卡兰尼克认同自动化是一把关键钥匙，甚至半开玩笑地提到马斯克的地下隧道公司 The Boring Company 若能将自动化掘进机制用于采矿，也将大有可为。 随后卡兰尼克破除了大众对“稀土（Rare Earths）”的认知误区。他指出，稀土元素在地球岩层中的储量其实非常丰富、随处可见，它之所以显得“稀有”，真正的瓶颈在于两点：其一是环保开采工艺极其严苛且对地表环境具有高度破坏性；其二是绝大多数已知储量分布在自然环境极其恶劣、法规严苛、人类根本不愿意长期定居作业的偏远地区。自动化重构了整个采矿经济模型：首先，它能使现有存量矿山的运转效率实现倍数级跃升；其次，通过派遣特种无人装备进驻那些高危、酷热、极寒或深藏地底的不毛之地，人类技术人员只需在远端控制室进行实时监控与调度。这不仅彻底消除了人员人身安全隐患和海外用工门槛，更直接打破了过去受制于劳动力和环境条件的地缘开采限制，让深埋在“世界尽头”的资源开采在商业上首度具备了可行性。 ### [10:00](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=600s) 物理 AI 技术栈与“特斯拉即当代的 Google” 谈及物理世界的自动化，构建完整的物理 AI 技术栈（Physical AI Stack）远不止算法计算和模型训练那么简单。除了常规的物理 AI 模型与算力集群，整个链路还必须把土地开发（land development）、基础化学（chemistry）以及硬核制造（manufacturing）等重资产环节统统纳入底层架构。放眼整个产业，能将这一整套复杂链条完全通吃并掌控的企业极为罕见。正因如此，人们审视当下的竞争格局时往往会发出感叹：“特斯拉已经把这套东西吃透了。” 在物理 AI 领域，特斯拉扮演的角色正如 2000 年代的 Google 或 90 年代末的微软——当年任何创业者去见投资人，面临的第一个问题必然是“Google 为什么不会捏死你？”或者“如果 Google 顺手做了你怎么办？”；到了 2010 年代，Uber 在移动出行领域也曾拥有过类似的统治级威慑力，创业项目总会被质疑“万一 Uber 在 App 里加个按钮呢”。而如今在物理 AI 赛道，特斯拉就是悬在所有人头顶的巨无霸。即便面临这样的行业巨头，创业者依然必须把握机会、放手一搏（"shoot your shot"）。 ```mermaid flowchart TD accTitle: 物理 AI 完整技术栈构成 accDescr: 展示从数字模型到物理现实落地的完整物理 AI 技术栈层次 subgraph DigitalLayer["数字与智能层"] A["物理 AI 模型与推理"] B["高密度 AI 算力与基础设施"] end subgraph PhysicalLayer["物理实体制造与能源层"] C["材料科学与化学 (电池/作动器)"] D["先进工程与大规模制造"] E["土地开发与工厂基地建设"] end DigitalLayer --> PhysicalLayer ``` --- ### [11:00](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=660s) 自动驾驶战局：Waymo 的规模短板与特斯拉的极客硬核 作为自动驾驶领域的先驱之一，Travis Kalanick 对当下群雄逐鹿的自动驾驶赛道给出了犀利判断。当前市场充斥着过多的舆论声量，呈现出“雷声大、雨点小”（"more bark than bite"）的虚火状态。真正处于前沿的核心玩家其实依然聚焦在 Waymo 和特斯拉两家身上，其余初创小玩家尚未展现出真正颠覆性的实力。 Waymo 无疑处于领先地位，且已经给出了无人驾驶商业化落地的“存在性证明”（existence proof is there）。然而，Waymo 的致命痛点在于整车制造能力、规模化量产（manufacturing and scale），以及在市场推进上的紧迫感与狼性（"urgency and fierceness"）。相比之下，特斯拉走的是一条截然不同的道路：坚持第一性原理与底层科学，直接挑战难度放大了一百倍的“地狱困难模式”（"hard mode times 100"）。特斯拉采用不依赖激光雷达等额外传感器的纯视觉方案，行业关注的核心焦点在于他们何时能够迎来视觉领域的“ChatGPT时刻”。这场较量的谜底究竟会在明天揭晓，还是需要再等上五年，依然充满变数。 - **Waymo 当前优势与瓶颈**：已实现落地运营与存在性验证，但受困于高昂造车成本、制造规模化能力匮乏以及推进节奏偏保守。 - **特斯拉破局路径**：全栈坚持底层纯视觉与第一性原理，系统通用性强，但关键突破点的发生时间窗口仍高度不确定。 --- ### [12:51](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=771s) 视觉语言动作模型与人类语言的高效压缩 随着移动自主系统的深入演进，行业正热议是否会出现一个类似 Android 的“视觉-语言-动作”（Vision-Language-Action, VLA）通用操作系统。当人们可以用自然语言发出指令，模型就能瞬间理解语意并在物理世界中识别物体、付诸操作。回顾从 GPT-3.5 到 GPT-4 的跃迁，世人见识了认知智能带来的颠覆，而如今物理 AI 领域也正站在类似的历史临界点前。 过去几十年来，深度学习与机器学习始终像一个不可解释的黑盒，输入数据吐出答案，人类无法洞悉其思维脉络。而多模态智能体（Agent）的引入正在改变自动驾驶的协同机制——例如由一个智能体负责操盘驾驶，另一个智能体负责警戒并实时提醒，如同人类驾乘时的协作交谈。主持人与嘉宾在此打趣调侃起日常乘车时的口误与玩笑（“亲爱的，那辆车还在两百米开外呢，别慌”），引得全场大笑。 抛开幽默，语言本身正是人类演化出的极其精妙的信息压缩机制。人类大脑运转仅需消耗约 100 瓦的能量，而现阶段驱动一台 Waymo 自动驾驶系统所耗费的能源，足足是人类司机的 100 倍以上。早期自动驾驶感知系统的误区在于试图穷尽环境中的所有细枝末节——即便头顶云层的变幻对地面行车毫无意义，传感器也会一股脑采集处理。如果能够借助类似语言的高密度压缩机制在驾驶系统与安全防护系统之间进行信息交互，便能以极高的效率剔除无关噪点、精准锁定物理危险。 --- ### [16:18](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=978s) 告别加州的痛苦抉择与奥斯汀新据点 Travis Kalanick 正式确认已于上一年的 12 月彻底搬迁至得克萨斯州奥斯汀，成为当地的法定居民。实际上，他早在 2021 年就在奥斯汀湖（Lake Austin）购置了房产，作为一个狂热的滑水爱好者（avid water skier），过去五年里他每年至少有 15 个周末飞到这里滑水度假，私下里还经常与主持人们切磋西洋双陆棋（backgammon）和纸牌。 对于离开加州，生于洛杉矶、父母皆是洛杉矶本地人的 Travis 流露出复杂而沉重的情感。他开玩笑称自己的家族资历几乎算得上“洛杉矶的联合创始人”，整个家族与成长根脉都深深扎在加州。看着故乡一步步滑向混乱与怪诞，不得不选择离开的心理过程堪称一场“漫长的心理哀悼”（"almost like a mourning process"）。加州恶劣的政策环境率先逼走了马斯克（Elon Musk），随后大批科技从业者、创业团队、乃至喜剧演员、音乐人和餐饮经营者纷纷出走得州。 许多身处加州的成长期科技企业 CEO 普遍面临想搬却搬不走的现实困境——团队、办公室和实验厂房均已在加州重度绑定。对此，Travis 认为当某件事让人感到无比阵痛且非做不可时，实际行动带来的冲击往往并没有想象中那么可怕，领导者必须拿出魄力带头推进。目前他正在奥斯汀紧锣密鼓地搭建新业务团队，并且正在洽谈一处紧邻湖畔的独立办公楼；他甚至打趣道，等这处办公室敲定后，自己未来每天可以直接骑着水上摩托艇（jet ski）去公司上班。 - **Travis Kalanick 的得州渊源**：自 2021 年起在奥斯汀湖安家，长期在当地度假滑水，已于去年 12 月正式将税务和居住身份迁至奥斯汀。 - **加州人才外流链条**：由头部企业家带头，逐步蔓延至中坚科技创业者、实体企业制造端及文化创意社群。 - **奥斯汀新业务规划**：新团队正在当地扩建落地，正在谈判紧邻湖岸的办公场地，推崇高效、贴近自然的全新工作形态。 ### [19:43](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1183s) 告别旧金山奔向奥斯汀：四亿美元一英里的公交道与被扼杀的城市 全场畅聊起迁往得克萨斯州奥斯汀的切身体验。几位主持人感慨，搬到这里的人几乎都有完全相同的反应：“天哪，我住的地方空间大了一倍，花费却只有原来的一半；这里的人很酷，食物棒极了，每个人身上都带着一种我们在共同建设未来的劲头，氛围轻松又积极。”在经历过纽约、洛杉矶和旧金山这三座全美标志性大都市之后，反而是奥斯汀最让人有“家”的归属感。这里的产业与人才背景极其多元，各行各业的人都在踏踏实实开拓事业，让人真切感受到未来的脉搏。 特拉维斯·卡兰尼克（Travis Kalanick）谈及旧金山时坦言，虽然自己亲手在那里创立了 Uber，每次回去心头依然难免泛起一丝怀旧与心动的“蝴蝶煽翅感”，毕竟那座城市曾拥有无与伦比的魔力，但如今的治理却令人痛心疾首。作为一个习惯“从道路交通视角审视世界”的创业者，卡兰尼克直指旧金山随处可见那些耗费天价却空空荡荡的非机动车道和公交专用道：“你看着那些从来没有一辆公交车或一辆自行车经过的专用道，建一英里居然要花 4 亿美元。这简直就像是一种潜意识里想要把整座城市彻底扼死的冲动。”他特别提到旧金山著名的市场街（Market Street），整条主干道被直接刷成红色只准公交车通行，美其名曰环保倡导，实际上车内空无一人、街道死气沉沉，沦为了纯粹的“美德绑架（Virtue Signaling）”，整座城市的运转机制已经彻底瘫痪。 --- ### [21:20](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1280s) 群聊里的“达斯·弗里德伯格”与五亿美元萨克拉门托自救计划 话题随即转向几位主持人在私密群聊里的日常吐槽，现场爆发出一阵欢笑。杰森·卡拉卡尼斯（Jason Calacanis）当众揭底戴维·弗里德伯格（David Friedberg）：平日里弗里德伯格在公众面前温文尔雅，总是一副“我认为我们可以找到更好解决方案”的理智科学家形象；但在内部私人群聊里，几杯红酒下肚后的他就会彻底黑化成“达斯·弗里德伯格（Darth Friedberg）”，在群里怒发冲冠地痛骂“这些白痴简直是蠢货，他们正在彻底毁灭文明”。其他好友纷纷附和，调侃弗里德伯格一喝到第四杯酒，就会在群里甩出悬崖照，甚至冲动地跑到推特（X）上直接开撕国会议员，第二天酒醒了又不得不赶紧删帖。 哄笑过后，弗里德伯格抛出了一个严肃的政治与商业议题：目前加州正有一群科技与商业界人士筹备集资 5 亿美元，组建一个强大的科技商业联盟，正面进军州府萨克拉门托游说施压。长期以来，硅谷科技精英一直极力回避去萨克拉门托与政客纠缠，但眼下加州治理的崩盘已经到了悬崖边缘。弗里德伯格追问卡兰尼克：砸下 5 亿美元去正面扭转州政策究竟还有没有胜算？这个州是否已经无可救药？ --- ### [22:44](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1364s) 真理与正义是社会的免疫系统：失控的检察官与加州治安崩塌 卡兰尼克对此回应称，任何试图挽救局面的努力都值得支持，但他剖析了科技界深层的心态缺陷。过去的科技从业者成长于一种浓厚的自由意志主义（Libertarian）传统中——信奉“离我远点，我只想安静做产品搞建设”，长期对政治避之不及，但如今这种鸵鸟策略早已行不通。他认为加州的选民公投提案机制（Ballot Initiatives）是一项强大的宪政工具，能够绕过僵化的官僚体系直达立法核心，而眼下最需要重锤整治的靶心，正是那些选择纵容犯罪、彻底放弃起诉的地方检察官（DA）。 ```mermaid flowchart LR accTitle: 社会机体健康机制示意图 accDescr: 展示真理与正义如何作为免疫系统保护社会下游秩序 subgraph Defense ["核心免疫机制"] direction TB TJ["真理与正义<br/>Truth & Justice"] LE["严格司法追责<br/>打击犯罪与制度清明"] TJ --> LE end subgraph Downstream ["下游社会生态"] direction TB SEC["公共安全稳定"] ECO["经济繁荣创新"] SEC --> ECO end subgraph Disorder ["免疫力被压制时"] direction TB DA["极左检察官放弃起诉"] CHAOS["犯罪横行 / 毒品失控 / 城市衰落"] DA --> CHAOS end Defense ==> Downstream Defense -.->|机能退化| Disorder ``` 卡兰尼克分享了他审视世界的一条核心法则：“真理与正义是人类社会的免疫系统；当免疫系统受到压制时，所有的社会病灶就会全面爆发。”当他感叹世道“变得诡异”时，本质上指的就是真理与正义的基石遭到了侵蚀。杰森和萨克斯随后提到了旧金山由《All-In》播客团队带头罢免的激进地检官切萨·鲍丁（Chesa Boudin），以及洛杉矶的地检官乔治·加斯康（George Gascón）。他们感叹加州大城市几乎像被《蝙蝠侠：黑暗骑士崛起》里的反派贝恩（Bane）所接管，整座城市竟然在为罪犯服务。卡兰尼克动情地提到，他认识洛杉矶很多原本将保家卫民刻进骨血里的终身老警察，如今纷纷愤然离职，因为在眼前的畸形体制下，警察即使目睹暴行也被禁止出手阻止，执法者得不到任何体制背书，甚至因此患上了严重的心理创伤（PTSD）。 --- ### [25:21](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1521s) 松鼠能效与机器人之手：从人形狂热到“能打工的专用机器人” 话题在严肃的政治探讨后重新转回前沿技术。弗里德伯格提出，在自然界中，一只松鼠在树丛间自如腾跃飞扑，整个生物系统的能耗仅有区区十分之一瓦特；而相比之下，人类用金属构件、电动机和执行器（Actuators）堆砌起来的机器人，仅仅让其走上四英尺距离，系统累计消耗的能量就高达 1200 瓦。生物在物理行动上的能量转化效率极其完美，而现有的物理 AI 硬件却充斥着能效损耗。弗里德伯格询问卡兰尼克，要让具身物理 AI 真正实现工业级扩展，底层的材料科学与执行器该如何演进。 卡兰尼克指出，大众和资本如今大部分注意力都被人形机器人（Humanoids）所吸引，陷入了一种好莱坞电影《终结者 2》式的执念——所有人都在研究怎么做出一只完美复刻人类动作的机械手，或者去攻坚机器人类似人体“阿基里斯腱”的关键运动部件。但他澄清，自己所全力投入的新创业方向并非人形机器人，而是深耕于餐饮和食品处理的专用机械设备（Food Machines）。 - **专用精细操作的挑战**：卡兰尼克团队解决的是如何在不撕裂纸袋的前提下将纸袋撑开、并平稳将一碗热食精准放入包装袋中。这虽然不需要复杂的本体感觉（Proprioception）去穿针引线或打网球，但对实际工程精度的要求极高。 - **噱头 vs 经济价值**：市面上几乎所有双足人形机器人的演示视频，开场永远是让机器人跳舞或打太极、打功夫；而卡兰尼克的路线是走专用机器人赛道，打造能够直接进入商业闭环、真正创造经济收益的“在岗就业型机器人（Gainfully employed robots）”。 --- ### [28:34](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1714s) 硬件追赶软件：材料科学的重度投资与“资本作为武器”的重演 面对物理 AI 的演进瓶颈，卡兰尼克完全赞同弗里德伯格关于“材料科学与执行器硬件蕴藏巨大风投机遇”的判断。如今物理 AI 的上层模型与多模态操作软件正以惊人的速度迭代，但如果底层的机械工程、传感器与材料硬件无法跟上步伐，软件的先进潜力就无法转化为真实的生产力。物理 AI 的整套堆栈庞大无比，硬件层正迎来一场必须补课的大规模投资潮。 正当讨论深入到这一技术拐点时，主持人话锋一转，将硬科技投资与卡兰尼克最著名的商业操盘风格联系在一起：当年卡兰尼克在执掌 Uber 时，率先开创了将“资本作为武器（Capital as a weapon）”的经典战法——通过超大规模的融资直接清空市场上可用的风险资本池，从源头断绝竞品的弹药来源，形成绝对不对称的压制壁垒。如今这一打法已被硅谷主流深度吸纳，并在 AI 和具身智能时代的资本军备竞赛中被推向了新的极致。 ### [29:23](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1763s) 资本作为战略武器与中东主权财富的变局 谈及结束隐身模式并展现更宏大的野心，特拉维斯·卡兰尼克（Travis Kalanick）被问到是否打算再次将“资本作为武器”（Capital as a weapon）。特拉维斯澄清，单纯为了资本而将资本视作战略武器毫无意义；只有在特定竞争态势下，它才会真正演变成生死攸关的战略武器。他回顾了 Uber 创立初期的残酷竞争：“在那个年代，无论你的产品 App 做得多好，只要孙正义（Masa）给你的竞争对手砸进 10 亿美元，你第二天就会丢失 20% 的市场份额。”因此，团队必须将融资打造成世界顶级的核心能力，必须比任何人都擅长募集资本，否则等待你的只有出局。 话题随即转向当前中东局势及伊朗危机对资本流动的冲击。有传言称沙特、阿联酋（迪拜）、卡塔尔等海湾主权基金正在收紧流向美国的资金。特拉维斯以切身经历回应：他的中东业务原本计划于当年 1 月上市，但沙特股市在短短两个月内下挫了约 20%，导致整个上市计划严重受阻。这一方面是因为油价经历了大跌，在沙特，所有人都在期盼油价回暖。面对当前的战局波动，他表示自己并不会在这种时候去到处打电话四处募资。 但特拉维斯依然保持着底层的乐观主义。他将目前的动荡比作此前的关税风波——初看天崩地裂，不久之后便迅速消化平复。他坚信只要度过震荡期，世界便会走向更美好的阶段。尤其是在人工智能（AI）、物理 AI（Physical AI）领域正在发生的革命，将带来巨大的生产力跃升，从而引领社会走向富足与繁荣的“黄金时代”。 - **资本与竞争**：Uber 早期若缺乏极致的融资能力，对手获得巨额注资后即可瞬间蚕食市场份额；顶级融资能力曾是生死门槛。 - **中东上市受挫**：特拉维斯旗下的中东板块原定 1 月公开上市，受沙特股市两个月暴跌 20% 及国际油价重挫影响，进程陷入停滞。 - **长远预期**：地缘震荡与关税类似，短期带来“震惊与敬畏”，随后会回归常态稳态；AI 驱动的物理世界自动化将释放长期生产力红利。 --- ### [32:14](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1934s) 探访中国制造：深圳奇迹、竞争意识与中美竞合 谈及正在推进的中美贸易磋商以及自身的中国之行，特拉维斯坦言，只要你亲身走进中国、考察当地的制造业与城市面貌，就必定会大为震撼。他对比了过去几十年见证的巨大变迁：几十年前的深圳宛如湿热且落后半个世纪的堪萨斯城，而如今的深圳已经超越了新加坡，散发着无处不在的现代化与先进感。在制造前线，无论是小米的超级工厂，还是无数充满干劲、极具进取心的创业者与工程师，其拼搏劲头（hunger）令人侧目。 特拉维斯提到了 2008 年北京奥运会开幕式带给世界的心理冲击：那种排山倒海的气势展示了他们登顶世界的决心与雄心，如今这一图景正在实体产业中照进现实。他对这种强大毫无嫉妒与敌意，反而充满敬意：“最优秀的想法正在赢得胜利，他们在极其激烈地追求真理与进步，并将其转化为现实。”他呼吁美国本土也必须提升自己的竞技水平（step up our game），同时主张中美之间应展开成年人式的、健康的良性竞争，而不是像上世纪 90 年代底特律活塞队那样互相使绊子甚至上演恶意冲撞。 特拉维斯还分享了一个切身体会：他的团队长期在中国打造餐饮后厨设施，有一位合作十余年的核心骨干，妻子是美国公民，两人目前都在中国生活。特拉维斯极为渴望能将这位得力伙伴带到美国共同开创新事业，但如今的签证与移民流程却困难重重。他直言当前的合法移民机制被搞得一团糟，美国本就是靠吸纳全球各地的杰出创新者追求“美国梦”才走向伟大的，应当给予合理、顺畅的跨国人才流动渠道，而不是给跨国夫妻和顶尖技术人才设置重重壁垒。 - **深圳的发展跨度**：从几十年前潮湿简陋的县城景象，蜕变为在基建与城市体验上超越新加坡的超级科技重镇。 - **制造前线的冲击**：以小米为代表的制造生态展现出强大的工程落地速度与野心，展现出类似 2008 年北京奥运开幕式的组织动员力。 - **良性竞合理念**：拒绝恶性零和博弈，主张通过提升自身竞争力进行友好对抗；呼吁改善高技能合法移民机制，避免阻碍全球顶级人才赴美创业。 --- ### [36:03](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=2163s) 迈克尔·戴尔登场：从宿舍 1000 美元到千亿美元商业帝国 随着全场热烈的掌声，特拉维斯结束分享，得克萨斯本土传奇企业家、戴尔科技（Dell Technologies）创始人兼首席执行官迈克尔·戴尔（Michael Dell）登台。出生于休斯敦的戴尔回忆起自己的创业起点：42 年前，就在得州大学奥斯汀分校（UT Austin）的大一宿舍（Dobie Center）里，他怀揣仅有的 1000 美元创办了戴尔电脑，当时距离大一学期结束只剩大约 10 天。 面对这段传奇经历，戴尔幽默地调侃道：“目前看来进展还算不错，中间虽然有些波折，但总的来说过得去——今年我们的年营收大概在 1400 亿美元左右。”主持人打趣复利的力量，戴尔则笑着接梗：“从小处做起，然后持续不断地往上累加，事情就是这么简单。”四十二年的复利增长，将一间学生寝室里的个人组装电脑生意，最终推向了全球企业级 IT 与 AI 基础设施领域的千亿美元巨头。 - **创业起点**：42 年前，迈克尔·戴尔在 UT Austin 的 Dobie 宿舍以 1000 美元启动资金创办公司，时值大一期末前夕。 - **当期体量**：戴尔科技当前年营业收入已达到约 1400 亿美元规模。 --- ### [36:59](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=2219s) 为什么是得州？低税收、大学生态与全美人口经济大迁徙 现场主持人和嘉宾许多已从加州移居奥斯汀，他们向戴尔提问：为什么是得州？为什么奥斯汀能够持续吸引企业家与资本？迈克尔·戴尔分析，得克萨斯州长期保持着低税收、亲增长的政策环境，以及务实进取的商业氛围。从宏观数据来看，剔除得州后的美国其他地区增速相形见绌，得州的经济轨迹简直就像是“全美经济的一个升级增强版”。 戴尔列举了一组令人瞩目的数据：奥斯汀即将迈入全美前十大城市行列，届时全美十大城市中将有四座坐落于得州；全美每出生十个新生儿中，就有一个出生在得州；在得州设立总部的纽约证券交易所（NYSE）上市公司数量，已全面超过纽约州本土及其他任何地区。他特别强调，得州大学奥斯汀分校（UT Austin）是这一繁荣生态的源头活水，源源不断地为戴尔科技以及驻扎于此的无数企业输送顶尖工程与商业人才，并沉淀了深厚的开拓者精神与创业底色。 主持人和现场嘉宾也感慨道，纽约、洛杉矶和旧金山湾区等曾经的伟大都会，如今政策倒退、治理失序，正在变得愈发“背离美国精神”，促使大量中产阶级和精英家庭出走。戴尔坦承，自己在奥斯汀结识了无数新迁来的邻居与朋友，大规模人口迁徙正在重塑版图。对于在此扎根数十年的本地人来说，得州的宜居与活力从来不是什么新鲜事，但在这个大分化的时代，它的独特优势正以前所未有的确定性展现在全美面前。 - **全美十大城市格局**：随着奥斯汀迈向全美前十，全美十座最大城市中将有四座属于得克萨斯州（休斯敦、达拉斯、圣安东尼奥、奥斯汀）。 - **人口活力与资本集聚**：全美约 10% 的新生儿出生于得州；得州拥有的纽交所上市企业数量已超越纽约州。 - **创新源泉与迁移浪潮**：UT Austin 为奥斯汀的高科技走廊提供源源不断的人才储备；高昂税收与城市治理退步促使沿海科技精英向得州发生历史性大迁徙。 ### [39:58](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=2398s) 德州数据中心热潮与 AI 基础设施建设优势 现场谈及埃隆·马斯克在德克萨斯州建造超级工厂（Gigafactory）的经历，指出德州最大的特点就是“真正允许你把事情干成”。马斯克此前曾感叹这是不可思议的体验，因为在加州，重重审批和监管根本不允许建造这样的新工厂，就连特斯拉位于弗里蒙特的工厂也只是改造了一座年久失修的老厂房。这种建设自由如今正在数据中心领域完全重现，德州正迎来一场鲜有人充分意识到的大规模算力基础设施建设狂潮，这与 Dell 当前的核心业务重心紧密契合。 迈克尔·戴尔指出，AI 基础设施的爆炸式扩张需要海量的土地与充沛的电力，而德州相比其他各州具备无可比拟的先天优势。特别是在西德州的一些城镇，地广人稀、居民稀少，当地对在荒无人烟但能源充沛的区域拔地而起的大型数据中心没有任何抵触情绪。在这种极度宽松的建设环境与电力资源支撑下，德州迅速成为了全球算力版图的核心支柱之一。 - **德州建厂优势**：极低的土地阻力、充沛的电网与能源供应、对大型工业与科技基础设施的高度包容。 - **与加州监管反差**：加州受制于环保法规与土地政策，阻碍重工业和超大型数据中心落地，迫使先进制造业与算力中心向德州转移。 ### [41:13](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=2473s) 从计算到思考：Dell 的 500 亿美元 AI 业务跃迁 全球对 Token 的需求正在呈现指数级爆发，Dell 不仅在德州，更在全世界范围内承建各类 AI 数据中心。戴尔回忆了一个极具戏剧性的商业时间点：Dell 推出全球首款搭载英伟达 H100 芯片的服务器时，距离 ChatGPT 正式发布仅仅早了几个星期。随着生成式 AI 的引爆，Dell 这一业务条线的营收增长轨迹惊人，从最初的 20 亿美元迅速攀升至 100 亿美元，随后跨过 250 亿美元，预计当年将直接达到 500 亿美元规模。 戴尔强调，人类计算史在经历了过去 60 年的算力运算与数据计算之后，迎来了一场根本性的“相位转变”（phase change）——机器如今已经能够独立思考并协助人类思考。虽然眼下的基础模型能力已令人叹为观止，但它们实际上是“未来最差的一代”，仍在以极高的速度演进迭代，算力供给远远追不上旺盛的需求。这种爆发不仅局限于超大规模云厂商（Hyperscalers），更全面席卷了超过 4,000 家部署“Dell AI 工厂”（Dell AI Factory）的企业客户，以及 Palantir 等致力于保护私有数据同时调用 AI 的主权 AI（Sovereign AI）场景。早在几年前，一些顶级量化交易机构和财富 100 强企业便开始自行采购服务器构建私有模型，由于架构复杂度极高，Dell 在两年前顺势打包推出了交钥匙式的“Dell AI 工厂”解决方案，如今已在企业级市场全面铺开。 ```mermaid flowchart TD accTitle: Dell AI业务演进与企业重构路径 accDescr: 展示从计算范式跃迁到Dell业务规模增长，再到企业自上而下通过AI工厂进行组织流程重构的全流程 subgraph ComputingShift["计算范式转变与基础设施扩张"] A["60年传统计算时代<br>数据运算与信息处理"] --> B["AI新纪元：机器主动思考<br>Token需求与算力大爆发"] B --> C["Dell AI业务跨越式增长<br>$2B -> $10B -> $25B -> $50B"] end subgraph OrgTransformation["企业落地：Dell AI Factory 实施路径"] C --> D["自上而下审视业务与技术演进"] D --> E["简化与标准化流程<br>沉淀并归集私有数据"] E --> F["部署端到端 AI 工厂<br>彻底重塑组织架构与运转流程"] end ``` ### [43:49](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=2629s) 资本支出折旧、ROI 辩论与企业自上而下的激进重构 面对千亿美元级资本支出的商业回报周期疑问，主持团队探讨了政府出台的加速折旧政策（Bonus Depreciation）。该政策允许企业在购置数据中心和硬件基础设施的当年实现 100% 税前费用扣除，期限长达十年，极大地刺激了企业投资。对于外界质疑 AI 是否只是泡沫、终端企业能否在第一年拿到 20% 的投入资本回报率（ROIC），戴尔直言，在 Dell 自身的运营及客户实践中，已经看到了大量效率与生产力提升突破 20% 甚至更高的切实案例。 然而戴尔特别厘清，获取超额 ROI 绝非“按下一个按钮就能自动发生”。企业现有的流程、工具与技术架构，往往只是当初成立时的历史妥协产物。想要吃到 AI 的红利，必须自上而下彻底重构：跳出原有框架，看清技术迭代的方向与期望的最终结果，先进行业务流程的精简与标准化，把工具和数据梳理整齐，最后再将 AI 深度注入核心业务。任何寄希望于各个部门各自为政、自发改进的想法都是徒劳的，这必须是一场彻头彻尾的“全盘组织重塑”（wholesale re-architecture）。戴尔提到，早在三年前他就向管理团队敲响警钟：五年后行业内一定会诞生一个在各项业务上都比 Dell 更快、更具创新力、成本更低的颠覆者，并将 Dell 逼入绝境；而 Dell 唯一的自救方式，就是“在被击败前，自己抢先变成那样一家公司”。这种激进变革带来了敏捷度的极大跃升，其上一季度基础设施业务同比激增 73%，并在当季给出了翻倍（100%）增长的亮眼业绩指引。 - **税收政策红利**：100% 加速折旧政策让企业在投资数据中心的首年免除巨额税收负担，为期十年的确定性显著提振了重资产下注意愿。 - **重构方法论**：拒绝部门孤岛式微调，推行高管层主导的端到端流程清理与数据归集。 - **业务飞轮验证**：基础设施业务季度营收增长从 73% 进一步指引至 100%，在如此体量的巨头中极为罕见。 ### [47:23](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=2843s) 5倍速范式转移：组织文化壁垒与 AI 原生颠覆者 回顾从个人电脑革命、客户端-服务器架构、互联网再到移动互联网的多次技术浪潮，过去每一次范式转移都用了十到二十年才完成对打字员队伍、收发室等旧式工作模式的系统性替代。而如今这场 AI 浪潮的演进速度被戴尔形容为“一个季度堪比以往的一年”，整体节奏至少快了五倍。然而，目前大约只有 10% 到 15% 的大型企业真正参透了 AI 的落地逻辑，其余绝大多数仍在盲目跟风，仅仅为了向董事会交差而敷衍地开发一些表面化的 AI 项目，如同小孩子拿着半成品向长辈邀功。 戴尔深刻指出，企业采用新技术的最大阻碍从来不是技术本身，而是“文化、领导力与变革的勇气”。在传统企业中，大多数员工和中层的奖金激励机制本质上是“不求有功、但求无过”，这让砍掉低效部门或终结过时流程的深层改革变得极其痛苦。在探讨现有企业能否转型成功时，戴尔引用了电商时代梅西百货（Macy's）与西尔斯（Sears）的对比，以及亚马逊等原生数字巨头的崛起逻辑。他明确断言，未来几乎所有行业都会涌现出彻底颠覆旧秩序的“AI 原生”（AI-native）企业。Stripe 创始人科里森兄弟（Collison brothers）提供的数据直接印证了这一趋势：在 Stripe 平台上，2025 年创立的企业新群组（Cohort）其业务扩张速度，已经是 2018 年新创企业的整整四倍。由于从诞生第一天起就全面基于最顶尖的 AI 工具和全新架构构建，每一个年份诞生的新生代企业都在以远超前辈的加速度全面重塑商业版图。 ### [50:23](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3023s) 传统巨头的生存危机与生产力扩张视角 Michael Dell 指出，对于拥有知名品牌、雄厚资产负债表和深厚客户关系的现有成熟企业而言，这些既有优势本质上都是“价值正在不断贬值的资产”（expiring value assets）。如果企业不能迅速变革并彻底融入这一波浪潮，最终必将被淘汰出局。Dell 提到，早在三年前他就曾向公司内部管理团队发表过这番警示，强调必须采取极其大胆的行动进行彻底革新，不仅是为了渡过难关，更是为了让戴尔（Dell Technologies）在即将到来的 2030 年代持续繁荣。 针对外界普遍担忧“AI 是否只会导致企业用更少的人做同样的事”，Dell 持完全不同的看法。他承认短期内确实会有一部分成本优化和人员替代，但核心逻辑在于生产力的根本跃升。当人类掌握了更强大的工具时，并不会停步不前，而是会去做海量以往无法做到的事情——大幅加速科学探索、攻克之前束手无策的科学难题，并发明全新的事物。他认为把视野局限在单纯的“降本减员”完全低估了这项技术的巨大推动力。 - 既有企业的核心壁垒如果脱离了前沿技术演进，就会迅速沦为过期的折旧资产。 - 戴尔的战略重心始终聚焦于提前布局，确保组织完全适配 2030 年代的技术范式。 - 工具升级的终极影响是大幅拓宽人类的问题解决边界，而非单纯维持现状的同时削减劳动力。 --- ### [52:14](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3134s) AI 基础设施演进：边缘计算与“最便宜的 Token” 当主持人 Jason Calacanis 问及目前市场对 AI 基础设施的认知是否存在误区——即原先以为一切都必须集中在大型云端数据中心与专有闭源模型中，未来是否会全面转向边缘端与开源生态时，Dell 表示答案是“兼而有之”（all of the above）。无论是工业制造、自然资源开采、先进制造、零售还是物流行业，边缘端的推理需求都在爆炸式增长。数据在哪里产生，AI 的算力与推理就必须部署在哪里。 Dell 戏谑地指出：“每个人都爱公有云，直到他们收到账单的那一刻。”原本许多企业以为迁移到云端是为了省钱，结果账单金额却极其惊人。因此，**成本最低的 Token 必然是直接在数据产生的源头设备上生成的 Token**，未来智能手机、PC 以及各类嵌入式设备都将独立产生大量的 Token。戴尔目前拥有超过 10,000 家将戴尔硬件深度嵌入自身产品的企业级客户，涵盖医疗设备、工业中控系统、网络存储及安防设备等，这让戴尔对边缘算力网络的重塑有着清晰而直接的洞察。 ```mermaid flowchart TD accTitle: 数据生成与就地推理架构 accDescr: 展示数据产生后在边缘端设备就近推理，避免高昂云端成本的流程链路 subgraph EdgeLayer["边缘端就地推理（成本最低 Token）"] A["数据产生源头\n(智能手机 / AI PC / 医疗与工业设备)"] --> B["本地嵌入式计算硬件\n(Dell 嵌入式系统 / 本地芯片)"] B --> C["即时本地推理\n(低延迟、无传输带宽开销)"] end subgraph CloudLayer["传统集中式公有云"] D["海量私有数据回传云端"] --> E["高额网络带宽与云账单开销"] E --> F["云端集中式模型推理"] end A -. 避免高昂账单成本.-> D ``` --- ### [54:20](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3260s) 桌面算力复兴、开源生态与自主智能体（Autonomous Agents） Jason 与 Dell 展开了一段轻松的风趣互动。Jason 谈及当前开源生态大火，许多极客与开发者纷纷重拾组装工作站的热情，甚至联想到当年大家打电话找戴尔定制 PC 的场景。Dell 笑着纠正道：“他们现在是在 Dell.com 下单……有一种叫互联网的东西，Jason，它运转得挺好的。”Jason 随后调侃这台下并非广告赞助，但确实看到大众为了保护个人数据、隐私和工作流技能，不愿将敏感数据拱手让给云端服务商，转而愿意花上万美元自建本地高配桌面主机。 Dell 强调，虽然不确定是否所有人都会买一台一万美元的工作站，但这一趋势极其显著。戴尔在 Hugging Face 上建立了官方认证与适配入口，全面测试并验证了几乎全系机型运行各种开源模型的能力。如今无论是 Google 的 Gemma 系列、OpenAI 的开放模型，还是英伟达的 Nemotron 系列，都在小型设备上运行得极其出色。当现场询问有多少观众配置过 OpenClaw（开源智能体工具）时，约有 20% 的台下观众举手。Dell 强调，企业内部自主智能体（Autonomous Agents）的普及已经是不可阻挡的大趋势，随之而来的关键挑战是必须建立严格的安全防护、身份验证以及行为控制机制。 - 戴尔与 Hugging Face 深度整合，对全系计算硬件进行了主流开源小模型的认证适配。 - 本地化运行模型有效满足了企业与个人对数据私密性及知识产权自主可控的核心诉求。 - 自主智能体的大规模落地，必须搭配对智能体身份凭证、权限边界和审计追踪的严密安全架构。 --- ### [57:07](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3427s) 跨越智能阈值、暗数据挖掘与 AI 的公众认知困境 在谈及通用人工智能（AGI）及超级智能（Superintelligence）的时间节点时，Dell 坦承自己无法给出一个确切的日期，但他与团队在后台交流时一致认为，随着近期发布的各个前沿大模型（如 Gemini 3.1、Opus 4.6、OpenAI 5.4），业界似乎已经跨过了一个至关重要的“智能阈值”。以往戴尔内部技术团队需要耗费数月甚至九个月才能完成的研发任务，现在借助最新模型在一到两周甚至一天内就能搞定，创新的迭代速度令人震惊。 Dell 提到，除模型本身的强化学习进展外，未来最大的潜能在于各行各业尚未被大模型触碰过的“私有暗数据”（private dark data）。真正的护城河在于利用企业自有的历史沉淀数据重新微调标准模型，同时严格保障数据的专有性。 当被问及若今天从零开始重新创办一家非硬件领域的“AI 原生”公司该如何组织人力时，Dell 实话实说：“我其实没花太多时间想这个，光是考虑怎么管理好戴尔就已经够难了。”他更关注的是推演未来技术工具的轨迹——思考 2027 年、2028 年、2029 年的工具会进化成什么样，并倒逼企业现在加快向那个阶段演进。 谈到公众舆论时，Jason 提及在一项公众好感度民调中，“AI”一词的受欢迎程度跌入谷底，甚至排在以严苛著称的移民与海关执法局（ICE）之后，与 ISIS 和民主党等标签处在同一争议区间。Dell 幽默地回应称，AI 之所以引发如此大的社会反弹，很大程度上是因为它被包装并表现得太像一个活生生的人类；“如果我们当初别叫它人工智能，而是把它老老实实称作‘线性代数矩阵计算’（linear algebra matrix calculations）或者‘矩阵乘法器’，大家或许就不会这么抵触了。” ### [01:00:39](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3639s) AI 的社会焦虑与人类潜能的放大 现场关于人工智能与公众沟通的讨论继续深入。主持人指出，目前外界对 AI 的定位与沟通存在偏差，未能向普通大众讲清技术带来的普惠价值——例如 AI 能改善医疗保健、延长寿命、提高儿童教育水平，甚至压低住房和食品成本。随后，主持人直切技术浪潮带来的深层隐忧：抛开信息传递不谈，AI 是否会加剧就业岗位的剧烈动荡？是否会造成一部分人收入爆发式增长，而另一部分人深感被时代抛弃？这种撕裂正在助长社会不满，促使政客们跳出来呼吁“叫停数据中心建设”。 Michael Dell 对此展现了科技企业家的长期乐观主义。他指出，所有技术周期中网络效应都是不可避免的，但更重要的是人们将借助新工具做更多事情。无论是教育水平的飞跃、科学发现的井喷，还是医疗健康和新能源技术的发展，各类未解难题的突破都在被大幅加速。Dell 强调，AI 的本质终究是对“人类潜能与能力的放大”（amplification of human potential and capability），是拓宽人类认知与行动的边界。 现场谈及 AI 本质时还爆发了一阵幽默的互动。大家意识到，除去先进半导体与庞大数据中心，AI 归根结底只是运行在电脑上的软件。“有人说‘我们不想要这个’，但你怎么可能全面封杀软件？”主持人调侃起纽约州提出拟立法禁止 AI 模型提供医疗或法律建议的举措，戏称这简直是“反软件法”，讽刺道：“如果这也算非法，那干脆把医书和法律指南也全禁掉好了。” - **AI 解决的核心领域**：加速攻克医疗健康、长寿、教育普及、能源转型及科学发现等历史难题。 - **技术本质与监管困境**：在云端之外，开源模型可以直接在个人终端本地运行，单纯靠行政立法很难真正封禁个人在电脑上使用软件获取信息。 --- ### [01:02:54](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3774s) 调侃伯尼·桑德斯与 Invest America 的横空出世 在讨论科技资本与民粹反弹时，主持人即兴惟妙惟肖地模仿起参议员伯尼·桑德斯（Bernie Sanders）激昂的口吻：“Michael Dell！你搞这些数据中心，就是在为那 1% 中的 1% 极少数人服务！你为何要这样对待我们伟大国家的人民？……与此同时，你却把真金白银塞给全美儿童设立 Invest America 账户！这算什么？富豪不打招呼就给孩子们发钱，孩子们回头买了父母根本没要的东西怎么办？！” 面对这番戏谑，Dell 笑着澄清事实：孩子们要等到年满 18 岁才能动用这笔资金，绝非随意挥霍的零花钱。主持人继续打趣道：“那赋予你什么权力去给全美孩子资助教育？这算哪门子慈善？根本讲不通嘛！”在一片欢笑声中，Dell 敏锐地发现台下观众席里坐着该法案的核心发起人、知名投资人 Brad Gerstner，随即邀请这位老友走上舞台，同台揭晓这笔史诗级慈善捐助背后的传奇历程。 --- ### [01:04:03](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3843s) 62.5 亿美元世纪捐赠：直击 2500 万儿童的种子基金 Brad Gerstner 登台后，Dell 回忆起这段合作渊源。早在 2021 年，Brad 就向他提出了“Invest America”构想——让每个美国孩子从出生起就拥有一笔投资账户，借助复利分享国家经济增长的红利。虽然最初与上一届政府的接洽未能取得进展，但随着《投资美国法案》（Invest America Act）最终通过，数千家私营企业纷纷宣布跟进，提供匹配资金。 就在法案落地之际，Michael Dell 与妻子 Susan Dell 宣布了一项震撼全球的慈善决定：个人出资 62.5 亿美元，直接为全美 2500 万名儿童每人注资 250 美元。 ```mermaid flowchart TD accTitle: Invest America 捐赠与账户运作架构 accDescr: 展现 Michael Dell 夫妇 62.5 亿美元捐赠的资格条件、申领流程及长期愿景。 subgraph Eligibility["资格筛选条件"] Age["年龄：2 至 10 岁儿童"] Income["地区中位数收入：≤ 15 万美元"] Target["覆盖群体：全美 2500 万儿童"] Age --> Target Income --> Target end subgraph Execution["资金注入与申领"] Donation["Dell 夫妇出资：62.5 亿美元\n（人均 250 美元）"] Claim["家庭主动注册申领\n（当前速度：10 万+ 人/日）"] Donation --> Claim end subgraph LongTerm["成长与提取"] Invest["长期复利投资\n（多方企业后续配套追加）"] Payout["18 岁成年后方可支取\n（用于高等教育、创业、置业）"] Invest --> Payout end Target --> Donation Claim --> Invest ``` 主持人感叹，世人见惯了天文数字，往往忽略了这一善举的真正分量——两位慈善家坐下来做出的这项决定，是整个人类历史上最大规模的个人慈善捐赠之一。Dell 补充说明，虽然账户已经为孩子们建立，但家长依然需要主动注册完成申领，目前全国每天有超过 10 万名孩子完成账户认证。 - **捐赠总金额**：62.5 亿美元（每位符合条件的儿童获赠 250 美元）。 - **受惠目标人群**：年龄在 2 至 10 岁之间、居住在家庭中位数收入不超过 15 万美元邮政编码区域的 2500 万名儿童。 - **当前申领进展**：每日新增注册申领人数已突破 10 万名。 --- ### [01:05:58](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3958s) 白宫博弈、两党共识与资本主义信仰保卫战 Brad Gerstner 动情地将 Michael Dell 称为“国家英雄”与“国宝级资产”，并坦言 Dell 极其低调地淡化了这件事情推进的凶险与艰难。为了推动这一构想，Brad 奔走呼吁了四年之久。2025 年 4 月，正值关税纷争最激烈的关口，团队意识到新总统任期前两年内唯一可能通关的立法载体就是预算和解法案（Reconciliation Bill）。 当时参议院虽然起草了具备两党支持的条款，但最后窗口期仅剩短短 5 天，必须直接拿到白宫椭圆形办公室的支持。Brad 第一时间致电 Dell：“Michael，我们必须立刻行动，我们得进椭圆形办公室。”随后，凭借商界领袖在白宫的信誉，他们直接与唐纳德·特朗普（DJT）取得联系并完成了关键沟通。Brad 指出，无论个人政治立场如何，必须承认该届政府非常积极地采纳商界领袖的建设性构想，并对 Dell 这类传奇企业家怀有极高敬意。 Brad 强调，Invest America 绝不是党派政治的工具，“它不是红党的理念，也不是蓝党的理念，而是红白蓝（美利坚）共同的理念”。他指出，当前全美 40 岁以下年轻人中，对资本主义持正面看法的人已经不足半数，70% 的人感到自己被抛弃、与“美国梦”彻底脱节。这项举措不仅是出于慈善道义，更是为了通过“所有权社会”（ownership society）捍卫让美国繁荣了 250 年的自由资本主义根基。 现场还开玩笑提议是不是该把这项目改叫“特朗普账户（Trump Accounts）”，引得全场哄堂大笑。当被问及该项目如何融入整体慈善版图时，Dell 介绍道，迈克尔与苏珊·戴尔基金会（Michael & Susan Dell Foundation）一直聚焦于城市贫困儿童，而 Invest America 宛如一个高效的“直达受众平台”。他们最初甚至打算先在得克萨斯州试点，但鉴于戴尔公司近来业绩蒸蒸日上，夫妻二人索性决定将规模直接推向全国。Dell 呼吁道：“如果未来再出现 10 个 Michael Dell 会怎样？世上能拿出同等甚至更大捐赠的人大有人在，众人拾柴火焰高，这才是改变国家未来的方式。” ### [01:09:45](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=4185s) Invest America 的长期愿景：全美儿童终身资本账户与 5 万亿美元转移 Brad Gerstner 与 Michael Dell 共同担任 Invest America 捐赠委员会（Giving Committee）的联合主席。现场谈到这一计划时，Brad 强调这已经远远超出一笔巨额慈善捐款的范畴，而是正在演变为一场深刻重塑社会的“全国性运动，而非昙花一现的瞬间”（"a movement instead of a moment"）。为了推动这项计划落地，他们不断四处奔走、敲门游说，并向公众澄清关于该法案的关键事实： 这项立法的覆盖范围和机制具有颠覆性： 1. **全员可认领**：全美所有 18 岁以下的未成年人，现在都有资格去认领属于自己的账户。 2. **制度永久化**：外界误传该计划只针对 2025 年至 2028 年出生的孩子，但事实上立法确立的是一项永久性制度。从 2027 年 1 月 1 日起，美国出生的每一个婴儿，在出生时都会自动获得一个与社保号（Social Security Number）直接绑定的资本账户（法案中被称为“Trump account”）。虽然 1,000 美元的政府注资每四年需要国会重新授权一次，但账户机制本身永久存续。 Brad 将其定义为“社会保障 2.0”（Social Security 2.0），称其为美国过去 50 年来社会契约发生的最大变革。全美每年约有 370 万新生儿，每个孩子都将从降生之日起拥有一个通过复利不断增值的“出生即享 401(k)”账户。未来注资的不仅是顶级富豪，还包括在 IPO 上市时捐赠股票的企业、各州政府、企业雇主以及普通工薪父母。据测算，未来 15 年内该计划将直接把 5 万亿美元的资产注入那些此前财富积累为零的普通美国家庭手中。 - **18 岁以下全员覆盖**：当前全美所有 18 岁以下儿童均具备账户认领资格。 - **永久制度保障**：自 2027 年 1 月 1 日起，每年约 370 万名新生儿出生即自动开户并绑定社保卡。 - **巨额财富转移**：预计在 15 年内向原本净资产为零的底层及工薪家庭输送高达 5 万亿美元资本。 - **多元资金生态**：资金来源整合了国会拨款、科技公司 IPO 股票捐赠、州政府配资与家庭小额储蓄。 --- ### [01:11:56](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=4316s) 绕过政府官僚机器与社会保障制度的反思 现场就资金分配效率展开了犀利的辩论。大家一致指出，如果这 5 万亿美元按照传统的联邦福利计划运作，在经过层层官僚体系与裙带结构的中转抽成后，资本很可能会被直接“焚化（incinerated）”。将资金直接注入以个人名义设立的独立投资账户，彻底斩断了中介寻租与低效损耗的可能。 紧接着，对话引申到对美国现行社会保障制度（Social Security）底层逻辑的反思。现场嘉宾甚至情绪激动地将其直斥为某种“虚假的庞氏骗局”（fake Ponzi scheme）：在现行工业时代遗留下来的社会契约下，政府强制每位劳动者将收入的约 10% 扣缴为薪酬税（payroll tax），但这些钱直接坠入了一个无人知晓具体去向的黑洞，既没有真实资产与之对应，普通人也无法直观看到它的增值。 对话提出了一种彻底重构社保的构想：未来完全可以将现行的现收现付制社保逐步替换为确定缴费型（Defined Contribution）的真实资产账户。工薪族每一笔薪酬税扣缴都不再被通胀和赤字稀释蒸发，而是真正买入代表国家生产力的优质企业股权，不仅账目透明、享有选择权，还能直观看到个人财富随着国家经济一同复利增长。 ```mermaid flowchart TD accTitle: Invest America 资本直达机制与传统官僚体系对比 accDescr: 对比传统政府福利渠道的高损耗与 Invest America 直接注资个人账户的增值路径 subgraph Traditional["传统政府福利路径"] A["财政拨款 / 薪酬税"] --> B["庞大联邦/地方官僚机构"] B --> C["层层中介与裙带损耗 / 资金被焚化"] C --> D["底层家庭实际获益寥寥"] end subgraph DirectAccount["Invest America 创新路径"] E["政府种子金 / IPO股票 / 慈善捐助"] --> F["儿童独立资本账户<br/>（绑定社保号）"] F --> G["配置标普500指数等核心资产"] G --> H["15年积累5万亿美元<br/>复利直达家庭财富底盘"] end ``` --- ### [01:12:54](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=4374s) 数字化体验与全民做股东：从生日红包到各州领养计划 谈及产品的具体落地，Brad 透露在当年的 7 月 4 日美国建国 250 周年之际，所有开户家庭将在手机端迎来一个体验极佳的应用程序，界面设计类似于 Robinhood： - 目前已有超过 450 万名儿童认领了账户，正以每天约 15 万人的惊人速度递增，预计到 7 月 4 日认领量将突破 1,000 万人。 - 打开 App 后，孩子们和父母会清晰看到自己收到的 1,000 美元或 250 美元，并看到这笔资金被直接拆解为标普 500 指数中的具体持仓——“拥有一点点英伟达（NVIDIA）、一点点沃尔玛（Walmart）、一点点戴尔（Dell）”。 这种可视化让普通孩子从小树立“我是美国未来增长的股东”的主人翁意识。App 还深度融入社交与日常生活场景：父母可以通过 Apple Pay 随手充值 5 到 10 美元；在生日、圣诞节或犹太成人礼（Bar Mitzvah）时，孩子可以向亲友分享专属二维码，大家赠送的不再是随手丢弃的塑料玩具，而是直接注入账户的优质资产。Brad 动情地表示，这不仅仅关乎金钱，更释放了巨大的心理潜能——“让原本觉得自己被边缘化的孩子感受到：我也在牌桌上了，我也有改变命运的机会（'I'm in the game. I have a shot'）”。 在推动法案的过程中，团队极其自律地克制了直接撼动传统社保神圣地位的冲动，以免法案中途夭折，但他们坚信这种直观可见、人人拥有的个人账户终将引导制度走向正确方向。在慈善配捐方面，典范效应正在快速扩散： - Brad Gerstner 个人出资“领养”了自己的家乡印第安纳州，为当地符合条件的孩子提供配套资金。 - 桥水基金创始人 Ray Dalio 及其妻子 Barbara Dalio 认领了他们的居住地康涅狄格州。 - 整个认领体系极其模块化，任何企业家或普通人都能够以极低的门槛认领自己社区的 100 个孩子、认领一个特定邮编区（Zip Code）、一个学区甚至一整座城镇。 全场观众爆发热烈掌声，向 Brad Gerstner 与 Michael Dell 在商业之外展现的深远慈善远见致敬，本期全景探讨人工智能、硬核科技与国家级资本创新的 All-In 播客现场直播在热烈的欢呼与标志性片尾音乐中圆满收官。 ## Notable quotes > [03:00](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=180s) 物理世界的数字化（digitizing the physical world） — Travis Kalanick > [05:47](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=347s) physical automation to transform industries and move the world — Travis Kalanick > [05:47](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=347s) 打造更高生产力的矿山，为地球工业提供动力 — Travis Kalanick > [10:00](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=600s) shoot your shot — Travis Kalanick > [11:00](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=660s) existence proof is there — Travis Kalanick > [22:44](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1364s) 真理与正义是人类社会的免疫系统；当免疫系统受到压制时，所有的社会病灶就会全面爆发。 — Travis Kalanick > [29:23](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=1763s) 在那个年代，无论你的产品 App 做得多好，只要孙正义给你的竞争对手砸进 10 亿美元，你第二天就会丢失 20% 的市场份额。 — Travis Kalanick > [36:03](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=2163s) 从小处做起，然后持续不断地往上累加，事情就是这么简单。 — Michael Dell > [43:49](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=2629s) 在被击败前，自己抢先变成那样一家公司 — Michael Dell > [52:14](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3134s) 每个人都爱公有云，直到他们收到账单的那一刻。 — Michael Dell > [57:07](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3427s) 如果我们当初别叫它人工智能，而是把它老老实实称作‘线性代数矩阵计算’或者‘矩阵乘法器’，大家或许就不会这么抵触了。 — Michael Dell > [01:05:58](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=3958s) 它不是红党的理念，也不是蓝党的理念，而是红白蓝（美利坚）共同的理念 — Brad Gerstner > [01:09:45](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=4185s) a movement instead of a moment — Brad Gerstner > [01:12:54](https://www.youtube.com/watch?v=9y7TTU1jIvk&t=4374s) 让原本觉得自己被边缘化的孩子感受到：我也在牌桌上了，我也有改变命运的机会（'I'm in the game. I have a shot'） — Brad Gerstner ## Mentioned Travis Kalanick, Michael Dell, Jason Calacanis, David Friedberg, Brad Gerstner, Susan Dell, Ray Dalio, Barbara Dalio, Donald Trump, Bernie Sanders, Chesa Boudin, George Gascón, Elon Musk, 孙正义 (Masayoshi Son), Patrick Collison, John Collison, All-In Podcast, Uber, Uber Eats, DoorDash, City Storage Systems (CSS), CloudKitchens, Kitchen Valley, Namah, Cocinas Ocultas, Atoms, Pronto, Dell Technologies, Dell AI Factory, Tesla, Waymo, The Boring Company, Google, Microsoft, OpenAI, Hugging Face, Palantir, Stripe, Robinhood, Apple, NVIDIA, Walmart, UT Austin, Invest America, Michael & Susan Dell Foundation ## Video description (0:00) Travis Kalanick: Officially exiting stealth mode, what he’s been working on (5:52) How to automate the physical world, markets to go after (11:00) Return to self-driving: Tesla, Waymo, and the autonomous race (16:17) Leaving Los Angeles for Austin, the decline of truth and justice in California (25:51) Actuators, robot hands, “Capital as a weapon,” Middle East SWF impacted by Iran War (36:00) Michael Dell: Dorm room to $140B in annual revenue, why Texas attracts founders (43:46) Dell's $50B AI infrastructure bet (1:03:50) Invest America: Michael Dell's $6.25B gift - A 401k from birth for 25M kids This podcast was recorded LIVE at Arena Hall in Austin, Texas. Thanks to our partners for making this event possible!: EY: Austin vibes meet AI innovation. Thanks to EY for co‑hosting with us at #SXSW. Discover what executives are saying about AI transformation in the latest AI Pulse Survey. https://ey.com/en_us/insights/emerging-technologies/pulse-ai-survey Forge Global: We’re proud to highlight our partners at Forge Global, who are helping the world’s most innovative private companies and their teams gain #liquidcourage on their terms. Learn more here: https://forgeglobal.com/who-we-serve/private-companies/ De'Longhi Athena Polymarket Follow the besties: https://x.com/chamath https://x.com/Jason https://x.com/DavidSacks https://x.com/friedberg Follow on X: https://x.com/theallinpod Follow on Instagram: https://www.instagram.com/theallinpod Follow on TikTok: https://www.tiktok.com/@theallinpod Follow on LinkedIn: https://www.linkedin.com/company/allinpod Intro Music Credit: https://rb.gy/tppkzl https://x.com/yung_spielburg Intro Video Credit: https://x.com/TheZachEffect #allin #tech #news [Source](https://www.youtube.com/watch?v=9y7TTU1jIvk)

[3:00] Travis Kalanick: 物理世界的数字化（digitizing the physical world） [5:47] Travis Kalanick: physical automation to transform industries and move the world [5:47] Travis Kalanick: 打造更高生产力的矿山，为地球工业提供动力 [10:00] Travis Kalanick: shoot your shot [11:00] Travis Kalanick: existence proof is there [22:44] Travis Kalanick: 真理与正义是人类社会的免疫系统；当免疫系统受到压制时，所有的社会病灶就会全面爆发。 [29:23] Travis Kalanick: 在那个年代，无论你的产品 App 做得多好，只要孙正义给你的竞争对手砸进 10 亿美元，你第二天就会丢失 20% 的市场份额。 [36:03] Michael Dell: 从小处做起，然后持续不断地往上累加，事情就是这么简单。 [43:49] Michael Dell: 在被击败前，自己抢先变成那样一家公司 [52:14] Michael Dell: 每个人都爱公有云，直到他们收到账单的那一刻。 [57:07] Michael Dell: 如果我们当初别叫它人工智能，而是把它老老实实称作‘线性代数矩阵计算’或者‘矩阵乘法器’，大家或许就不会这么抵触了。 [65:58] Brad Gerstner: 它不是红党的理念，也不是蓝党的理念，而是红白蓝（美利坚）共同的理念 [69:45] Brad Gerstner: a movement instead of a moment [72:54] Brad Gerstner: 让原本觉得自己被边缘化的孩子感受到：我也在牌桌上了，我也有改变命运的机会（'I'm in the game. I have a shot'）

[3:16] Travis Kalanick: physical automation to transform industries and move the world. [16:40] Travis Kalanick: Truth and justice are the immune system for society. When the immune system is suppressed, all the social ills flare up. [36:03] Michael Dell: it's been working out pretty good [47:24] Michael Dell: a quarter is like a year [51:46] Michael Dell: everybody loves the public cloud, right? Until they get the bill. [57:07] Michael Dell: linear algebra matrix calculations [57:07] Michael Dell: how do you stop total software? [62:51] Jason Calacanis: one of the greatest philanthropic gifts in the history of humanity. [66:43] Brad Gerstner: red, white, and blue idea

[0:00] I don't know if some of you knew I was [0:01] an angel investor in some companies. [0:05] On the count of three, what's my [0:06] favorite angel investment of all time? 1 [0:08] 2 3? [0:10] Thank you. Give it up, TRAVIS KALANICK. [0:13] >> [cheering] [0:13] [applause] [0:19] [applause] [0:20] >> APPRECIATE YOU. ALL RIGHT. WOW. Um on a [0:25] big news day Travis is here on a very [0:27] big news day. You spent [0:30] wow, guess like 7 years just in the lab [0:34] building. [0:35] Last year, every year I ask you, "Hey, [0:37] you want to come to the All-In Summit? [0:39] You want to" He said, "Nah, it's like [0:41] I'm going to just chill. I'm building." [0:43] Next year, "Hey, you know, just always [0:44] available to you." He said, "You don't [0:45] understand. I'm stealth." Yeah, he's [0:47] stealth. I'm stealth. Nobody knows where [0:49] I am. Nobody knows what I'm doing. The [0:50] employees are not allowed to put the [0:51] name of the company [0:53] on their LinkedIn. Thousands of [0:55] employees that weren't allowed to put [0:57] the company name on LinkedIn. I mean, [0:59] incredible. And I'm like, "Okay." And [1:02] >> Their parents thought they worked for [1:03] the CIA. Yeah, and then he's like, "And [1:04] by the way, Jason, you can invest. You [1:07] can't announce it, and you have to sign [1:08] an NDA. You can't mention you're an [1:10] investor." So, I'm like, "Okay, no [1:11] problem. I'm just happy to be on the cap [1:13] table." Is he like kind of like secret [1:15] saying what he wasn't supposed to say? [1:17] >> That's what I was Now he's talking about [1:19] the company. [1:19] >> What just [laughter] happened? You did [1:20] it. Well, you came Chamath, you're out [1:22] now. Okay, let's go. [1:23] >> You're out. It's out. You You came out [1:25] of stealth today. Um That's so funny. [1:28] Okay. [1:28] >> great. You came out of [1:30] >> Stealth Well, you you talked a little [1:31] bit. You came to All-In Summit last [1:32] year. [1:32] >> Is that fair? You say you're coming out [1:34] of stealth today? Is that right? [1:35] >> Well, look. Let's just start with what [1:37] that meant for our employees because [1:39] again imagine if you're at a [1:42] multi-thousand person company and every [1:45] single employee has stealth [1:48] on their LinkedIn. [1:49] Including sales people. [1:52] >> [laughter] [1:52] >> Okay? Including recruiters. [1:55] Like it was hard They They were they [1:58] were living life on hard mode. [2:00] It's kind of fun, too, right? I mean I [2:01] mean, yeah, they it was like [2:03] What's What is this? Why are there Why [2:05] is this massive density of stealth [2:08] >> Right. [2:09] startup people in Los Angeles? What is [2:12] happening over there? [2:13] >> Yeah. Yeah. Also, technically, the name [2:16] of the company in different countries [2:18] was very generic names of companies. I [2:23] mean, everything was designed to be [2:25] >> stealth. Right. So, [2:27] we operate in 30 countries. [2:29] In the US, [2:31] the kitchens product is known as Cloud [2:33] Kitchens. [2:35] In Korea, [2:37] it's Kitchen Valley. [2:40] In the Middle East, it's Namah. [2:44] In [2:46] Latin America, parts Latin America, it's [2:48] Cocinas Ocultas. [2:50] I mean, you get the idea. [2:51] >> You can't ever remember all the names or [2:52] all the code words. [2:54] >> about it, yeah. [2:54] >> The thing is through, um, but today [2:56] >> China. You know, it's like all over the [2:58] place, yeah. But things have gone really [3:00] well. And you've been a little [3:02] inquisitive. So, tell us about the [3:04] branding today that you're announcing, [3:06] and then maybe some of the acquisitions [3:08] and evolution of the company. You're not [3:10] just renting kitchen space. Those who [3:14] I mean, know how I thought about things [3:16] in the Uber day, a lot of this stuff's [3:18] not surprising. I I would often talk [3:20] about digitizing the physical world. I [3:22] think I even did it all at Summit. [3:24] The quick version of this, I'll try to [3:26] do it quickly, but it's like, we know [3:28] the bits world, the computer world, the [3:29] one that Michael Dell essentially [3:31] invented for us. [3:33] CPU, storage, network. These are the [3:35] three core computing resources. When you [3:37] go to computer science class your first [3:38] day, three core computer resources. CPU [3:41] manipulates the bits, storage stores the [3:43] bits, network moves bits from point A to [3:45] point B. [3:46] But if you're digitizing the physical [3:47] world, you're treating atoms like bits. [3:51] You're building an atoms-based computer, [3:53] and I'll explain what I mean to say. I [3:54] know this is a little little out there. [3:57] CPU manipulates bits, what manipulates [3:58] atoms? Manufacturing. Storage stores [4:01] bits, what stores atoms? Real estate. [4:04] Network moves bits from point A to point [4:05] B, what moves atoms? That's [4:07] transportation or logistics. [4:09] So, you have these three core computing [4:11] resources in an atoms-based computer. [4:14] The name of my company [4:16] was very obtuse and purposely designed [4:18] to be as boring as hell. Was called City [4:20] Storage Systems. [4:23] So, that's digitized real estate in an [4:27] atoms-based computer, our first computer [4:29] being a food computer. What does that [4:30] mean? Manufacturing, real estate, and [4:33] logistics for food. [4:36] And so, you start to get there, and the [4:38] idea the the mission was infrastructure [4:40] for better food. The idea was, can you [4:42] get a meal that's prepared and delivered [4:44] to you so efficient that it starts to [4:47] approach the [4:48] the the cost of going to grocery store. [4:51] If you can do that, you do to the [4:53] kitchen what Uber did to the car. [4:55] But in the Uber day, the roads were [4:57] there. The cars are unused. You just had [5:00] to put an app in the app store. Wasn't [5:02] that easy, but kind of that easy. [5:04] In this world, you can't do this on a [5:06] restaurant. Restaurant doesn't have When [5:08] I left Uber, 13% of all San Francisco [5:12] miles were Uber miles. [5:15] You can't get And that was 10 year 9 [5:17] years ago. You can't get there [5:20] on food, on restaurants. They They have [5:22] like 20% capacity. Uber Eats and [5:24] DoorDash fill it, but the infrastructure [5:26] to do high-capacity, high-scale sort of [5:31] industrial production is just not there. [5:33] And the logistics just not there. It [5:35] just doesn't work. That's why on [5:36] e-commerce, you go through Amazon [5:38] big-ass warehouses [5:41] with awesome logistics. You've got to do [5:43] the same thing when food when food goes [5:44] to e-commerce. That was a lot. Okay, so [5:47] bottom line is [5:49] it's awesome. [5:50] We do this food computation stuff. We're [5:53] doing more computers now. [5:55] And so the name of the company is called [5:57] Atoms. And it's let's say the mission [6:01] is [6:03] is physical automation to transform [6:06] industries and move the world. [6:09] And so we have our food computer we [6:10] talked about. Then we do we're doing [6:12] mining. Mining as in mining minerals. [6:17] >> We're talking about atoms, guys. Yeah. [6:18] So well of course you do some mining [6:20] data mine, too. But the point is is [6:23] physical mining. So automation of mines. [6:27] And [6:29] the mission there is [6:31] more productive [clears throat] mines to [6:33] power Earth's industries. [6:36] Right? So it's got this industrial atoms [6:38] vibe to it. And then on the transport [6:41] side, it's wheelbase for robots. Cuz if [6:45] you're doing specialized robots, not [6:47] humanoids, specialized robots, [6:51] you need to be able to move and act in [6:53] the physical world. [6:54] But the minute you're moving, [6:57] you got to have a wheelbase. So it's [6:58] just part of the equation. [7:00] And a lot of people go look at Tesla, [7:02] it's great. Look at [7:04] look at Waymo, awesome. They're cruising [7:06] around Austin, of course. [7:08] But there's so many things that move. [7:10] It's not just a ride-sharing thing. [7:13] And so obviously including mining [7:15] equipment that's doing its thing. So you [7:17] guys that's the general sort of idea and [7:20] we acquired a company on the mining [7:21] stuff. A company called Pronto. [7:25] Or [7:25] it's about to close. It's we're we're [7:28] inches from closing is the way to put [7:30] it. What were they doing? What was their [7:31] business, Pronto? Automating mining [7:33] equipment. [7:34] >> Where are they based? They're based in [7:35] San Francisco. So you and I were [7:37] starting to talk about this backstage, [7:38] but there's some folks I talked to in [7:40] the mining industry who mentioned, you [7:42] know, like the the the big issue with [7:44] mining, number one, is just surveying, [7:46] like finding the the the locations, [7:48] right? Is there an advantage to be [7:49] created there cuz I know there's a [7:50] couple startups that are trying to be [7:51] really smart about selecting locations [7:53] to get the targets out of the ground. [7:55] Yeah. [7:56] >> And then the other one is like, well, [7:57] can you go deep because pretty much [7:59] anywhere on Earth you can get whatever [8:00] you want if you're willing to go deep [8:02] enough, but the cost is is distance [8:05] squared, right? So, the energy cost is [8:06] like, how deep are you going to the to [8:09] to the second power? So, it becomes, you [8:11] know, geometrically more expensive to go [8:13] deeper, but the deeper you go, [8:15] you're the more you're able to kind of [8:17] not worry about getting the right [8:18] location. So, does automation unlock [8:21] that capacity that [8:22] >> Automation definitely does. I mean, I [8:24] mean, also it's like, man, does Boring [8:27] Company have some good stuff going? [8:30] Like, I hope we we're like we're doing [8:32] the mining thing, like, and Boring goes [8:34] makes, you know, some good tunnels for [8:37] for cars to do the thing, but like, [8:38] there's some kind of boring mechanism, [8:41] automated tunneling to do some of this. [8:44] But, to be honest, there's there's, you [8:46] know, they have this this thing is like [8:47] rare earths. I don't know why they put [8:49] plural. Rare earths, isn't it? Rares' [8:52] Earth? I don't know. But, the [8:55] but rare rare rare [8:57] rare earth, yeah. [8:59] >> Yeah, but the it's not rare. It's a very [9:01] common mineral. Guys, it's not rare. [9:03] It's what you have to do the land is [9:04] aggressive. And what's rare is the is is [9:09] where the where are the places they'll [9:11] let you do it that you can also sort of [9:13] get people to. [9:14] When you automate, you can go to a lot [9:17] of places. [9:18] Uh well, first is all the mines that [9:21] exist are way more productive. [9:23] Um and the second is you can then sort [9:26] of justify going to places you wouldn't [9:28] have been able to go before because [9:30] um you don't have as much of a labor [9:32] footprint or a safety issue or a whole [9:34] bunch of other things that then So, if [9:35] it's inhospitable, [clears throat] [9:37] if it's regulated, if it's like I don't [9:39] want to live there. It's the end of the [9:41] earth. [9:42] >> Yeah. You can send robots and have [9:45] people monitoring them remotely. [9:48] Yeah. And [9:49] this is like a future that feels like a [9:53] little bit like science fiction. Look, [9:54] we're here in Austin. You [9:56] you got to do the shout out to Tesla and [9:58] all the things cuz I like to sort of [10:00] break down the physical AI stack. [10:04] Includes not just like oh yeah, [10:06] computation and I've got to have [10:07] physical AI models and I got to all the [10:10] things you sort of think of. What about [10:12] land development? [10:14] That should be in that stack. What about [10:16] chemistry? That needs to be in the [10:17] stack. Manufacturing needs to be in the [10:19] stack. When you look at the stack, [10:20] you're like [10:21] damn, Tesla's got this. [10:24] They are they are the Google of this [10:27] era, which is what I mean by that is in [10:29] the 2000s, if you were doing a startup [10:31] in the 2000s, the first question you [10:33] would get [10:34] is uh why isn't why isn't Google going [10:37] to kill you? Or why isn't Google just [10:39] going to do it? Why isn't Google just [10:40] going [10:40] >> going to know that they killed you. And [10:42] before that, Microsoft. [10:43] >> And before that was Microsoft, the late [10:44] '90s. [10:45] Uber had the time, 2010s. [10:47] >> Yeah, but if Uber puts that in the app. [10:48] >> Come on. [10:49] >> [laughter] [10:49] >> It's like dude, this is Uber. I'm like [10:51] but you know, I think in the physical AI [10:53] space that's a that's sort of a Tesla [10:55] thing. [10:56] But there's so many things to do. [10:59] You got to shoot your shot. You got to [11:00] do these stuff. And rumors that [11:03] hey, you might not be done with [11:04] self-driving, something that you were [11:06] very early on. How do you think about [11:09] what you're seeing in the playing field [11:11] of self-driving because [11:12] my lord, you know, Waymo's making great [11:15] progress. Tesla's making great progress. [11:17] >> Like pick a winner between Tesla Tesla, [11:19] Waymo, Uber. Or like Uber Uber Uber Uber [11:23] Uber Uber seems to be building a network [11:25] of stuff. Yeah, I mean the number of [11:27] Pick a winner. [11:28] >> The number of players in this space is [11:30] crazy now, right? Like [11:31] >> Yeah, look, there's I think there's more [11:33] noise if there's more bark than there is [11:35] bite right now. [11:36] Um [11:38] look, I think Waymo obviously is ahead. [11:40] The existence proof is there. [11:42] Their issue is manufacturing and scale. [11:46] Um and urgency and fierceness. Like [11:49] let's [11:50] >> Come on. [11:51] >> Let's win. Let's go. Yeah. You know, [11:53] Uber had an autonomy project [11:56] back in the day. So and they have a [11:57] different strategy these days. I haven't [11:58] been there for a while. So but the but [12:00] the point is is that you So you got [12:02] Waymo, then you've got Tesla. [12:05] Fundamentals. [12:07] Science. [12:09] Hard mode times 100. [12:13] And the question is do they get there in [12:16] what time scale? [12:18] If if they And like honestly, [12:20] everybody's like, "Could happen [12:21] tomorrow. Could happen in 5 years." [12:24] And I think that it's like when does the [12:27] chat GPT moment happen for vision? Is [12:30] basically the thing. Let's call it [12:31] vision without other sensors. So super [12:34] inspiring, but like what's the timeline [12:37] on it? [12:38] Um those are the base This is basically [12:40] the And then there's a lot of other [12:41] little guys that don't really have the [12:43] stuff, I believe, [snorts] yet. [12:46] There's nobody standing out just yet of [12:49] of the others. Do you think we're at a [12:51] point now like obviously now that you're [12:53] getting into [12:55] more of these kind of autonomous systems [12:57] that move around, like do we have these [12:59] vision language action models [13:01] tuned and ready for prime time? There's [13:03] There's been a conversation like, "Who's [13:05] going to have the Android, the operating [13:07] system for vision language action, where [13:10] I can use my voice, tell it to do [13:12] something, and it knows what I'm saying, [13:13] and then it identifies the objects and [13:15] does the thing in the physical world?" [13:17] Do those models exist today, or are they [13:19] still work? And is that like a Google [13:22] OS, or like where does that OS come [13:23] from? [13:23] >> Look, I think there's a there This is an [13:25] area of a lot of energy, a mix of [13:29] research and implementation. I think [13:30] there's a lot of hope and interesting [13:32] stuff. I mean, the high level is [13:34] we all remember what happened when you [13:36] use chat GPT 3.5 [13:38] and you're like [13:40] holy [13:41] Yeah, it's legit. Whoa, and then it went [13:43] to four and you're like [13:45] okay, like some stuff just changed. The [13:47] world just changed and I can sort of [13:49] connect some dots [clears throat] and [13:50] getting real. [13:54] Is it about to happen? [13:56] Is it about to happen for physical AI? [13:58] And that's what this [clears throat] is [13:59] about. [13:59] >> Yeah. And the fun part about it is [14:02] machine learning, deep learning, this [14:04] kind of thing for many years, decades [14:06] was like inscrutable. I don't know what [14:07] the thing is thinking. It just spits out [14:09] an answer and I know it's correct. Well, [14:11] now you can have a conversation with it. [14:13] Right? Like imagine if it's driving your [14:15] car [14:17] and there's different agents and one's [14:20] just driving, the other's like, "Yo, [14:21] look out over there." Yeah. It's like, [14:23] "Oh." [14:24] Just like how we roll. [14:26] Like somebody does that, you're like [14:27] you're like [14:28] "Honey, that's like [14:30] 200 m away. We're going to be okay." [14:32] >> Yeah. [laughter] [14:33] Jason and I don't call each other honey, [14:34] but I got you. Yeah. [14:36] Like [14:37] sweetie. You know. Yeah. [14:39] >> [laughter] [14:39] >> Okay. So anyways, so that was odd, [14:41] wasn't it? Okay. [14:41] >> a little strange. [14:42] >> Okay. Okay. I didn't mean it that way. [14:44] >> does that. I didn't mean it. [14:45] >> Yeah. Yeah, you know, I meant it. Okay. [14:47] >> [laughter] [14:47] >> Language is a beautiful compression [14:50] mechanism [14:52] that humans use 100 W [14:55] of energy. [14:57] Like and you put that in the scheme of [14:59] things of like AI training, AI energy, [15:03] the power plants that are built to do [15:04] the thing that isn't even at human [15:07] strength yet. [15:08] Okay? The Waymo machine takes 100 times [15:13] more energy to drive a Waymo [15:16] than a human does to drive a Waymo. [15:19] So [15:21] so language we there are still things [15:23] that humans are great at and that [15:26] unbeat like the goat, we're still the [15:27] goat at certain things. Language is this [15:30] epic compression. [15:32] And um we need to find ways to compress [15:34] cuz like when you think about how how we [15:36] first started looking at the physical [15:37] world is we saw everything. And you know [15:41] what guys, and this is sort of obvious, [15:43] like it doesn't matter what the cloud is [15:44] doing if I'm driving. [15:47] But like the car doesn't know that. It's [15:49] pulling in every freaking data point and [15:52] processing everything and it's it's you [15:54] know, look, they've been about sort of [15:55] carving out the things that don't matter [15:57] and things like this, but there's ultra [15:59] awesome versions of this and you can [16:01] imagine how you can use language or [16:03] things that look like language to [16:05] communicate either amongst agents or [16:07] sort of safety systems with a driving [16:10] system to sort of [16:12] get very efficient answers and to [16:15] identify safety issues very efficiently. [16:18] People don't know that you moved to [16:21] Texas as of or most people don't know, [16:23] but it's it's out there. Yeah. You moved [16:25] here in December, so now you're a [16:27] resident of Austin. [16:29] >> Yeah, I was I'm about [16:30] Thank you. [16:33] It's very exciting for me. We've been [16:34] getting to play some backgammon. [16:35] >> Backgammon, cards, It's cards. We're [16:37] having a good time. So, I've had a place [16:40] on Lake Austin since 2021. [16:44] And I go there. I'm an avid water skier, [16:47] like [16:48] You're impressive at water ski king, I [16:50] have to say, like So, I've had a place [16:52] in Austin for 5 years. [16:54] Freaking love it. It's my weekend I [16:56] would go 15 weekends a year. [16:58] What do you think's going to happen in [16:59] California? [17:00] It's pretty messed up. Look, I I grew up [17:02] in Cali. Like I grew up in Los Angeles. [17:05] My parents were born and bred in Los [17:08] Angeles, which basically makes them the [17:10] founders of LA, okay? [17:12] But um so that I have a lot of heart, [17:16] like my whole family, everything, you [17:17] know? It's pretty [17:19] It's pretty It's I don't want to A lot A [17:21] lot of us feel that way. [17:22] >> want to get the violin out, but it's [17:23] just but It's heartbreaking. The place [17:25] It totally It's just the place you grew [17:27] up. It's your home, you know? [17:29] When you have to leave. Uh [17:31] but it's getting weird out there. [17:34] And uh [17:36] it feels like it's getting weirder. [17:38] And at some point that's it's just too [17:40] weird. It's too weird. Do you think [17:42] everyone's going to leave? [17:44] I mean, it started with Elon, and it was [17:46] like [17:47] >> Yeah, he was in the right. [17:47] >> We don't want Elon here, and then he's [17:49] like, "Message received." [17:50] >> Like, cool. Right. And then it kind of [17:52] worked its way down the tech industry, [17:54] and then the kind of it, you know, world [17:56] of people building businesses and [17:59] whatnot. And now it's kind of gotten so [18:01] broad in terms of the group of people [18:04] >> comedy, music, New Yorkers, restaurant [18:08] tours. I mean, this place is much I'm [18:10] not even talking about this. I'm just [18:11] talking about everyone leaving LA or [18:13] sorry, leaving California [18:15] is almost like [18:17] working down the path of Look, my the [18:20] rest of my team's like, "Where When are [18:21] we moving?" [18:23] You know? They're like [18:24] >> And how are you dealing with that? So, [18:24] that was the question with like [18:26] >> Got to buy a home on the lake. There are [18:27] literally dozens of startup CEOs of [18:31] call it successful or growing companies [18:33] that I talk to who are like [18:36] "Dude, I want to leave, but I got [18:37] employees here. I got an office here. I [18:38] got a facility here. I build stuff here. [18:40] How am I going to leave?" [18:42] >> yeah, I totally get it. It's a real [18:44] thing. So, look, I think like most [18:46] things [18:48] uh [18:48] sort of [18:50] when it's time [18:51] and it feels painful to do something, [18:53] sometimes it's actually not as bad as [18:55] you think. [18:57] And you just got to make the move and [18:58] lead and do it. [19:00] Um and so [19:02] uh that's kind of what that's kind of [19:04] the process that almost like a mourning [19:06] process I went through. And that's just [19:08] what it is. And you're setting up a team [19:09] here? Yeah, of course. [19:11] Uh and I got that office [19:14] right on the lake. Did you get that? Uh, [19:16] it's it's we are negotiating No, it's [19:18] all good. No, no, no, it's all good. [19:19] We're negotiating right now. Okay. But [19:21] I'm going to jet ski to work. No, [19:24] literally, we were Is it Is it true [19:26] story? Last year we're like driving up [19:28] the thing and I was like, "Wow, I wonder [19:30] who owns that?" He's like, "I will." And [19:32] I was like, "Did you look at it?" He's [19:34] like, [19:35] "I looked at that." [19:36] And I was like, "That's a That would be [19:37] a nice one." But the the truth is, you [19:40] know, I I and I had a couple people move [19:41] here a couple years ago and they all had [19:43] the same reaction. "Oh my god, I'm [19:46] living in a place that's twice as big [19:47] for half as much. The people here are [19:49] dope. The food is dope. Everybody here [19:52] is got this sense that we're building [19:55] the future and it's just fun and we're [19:57] all positive and, you know, for me, I [19:59] got to live New York, LA, San Francisco. [20:01] I I did three of the great cities in [20:03] this country. This one feels the most [20:05] like home to me, which is a very strange [20:08] feeling to me, but it feels like [20:09] everybody here wants to build the future [20:12] and it's very diverse, you know, like it [20:14] it all these different industries and [20:16] people pursuing stuff. I I think this is [20:17] the future. [20:18] >> Yeah, here's the thing. Like you go to [20:19] San Francisco and I still have a little [20:21] nostalgia when I go to San Francisco, [20:23] just [20:24] having built Uber there and the whole [20:26] thing. Um, I still get the [20:29] you know, the butterflies. Just I did, [20:31] you know, but it does have something [20:32] magical. You you just can't take it [20:34] away. And then you look at all of these [20:37] bike lanes and these bus lanes that [20:39] never have a bus or a bike in them. And [20:42] cost $400 million to build 1 mile. And [20:44] it's literally it's sort of like this [20:48] subconscious [20:50] desire to choke the city off. Now, [20:52] remember, I look at things through [20:54] roads. That's how I think. So, I'm just [20:55] like, obviously, the city is totally [20:58] busted. Yeah. [21:00] No, but they they literally took Market [21:02] Street and they're like, "What would be [21:03] the optimal way to this up and virtue [21:06] signal at the same time. And they're [21:08] like, "Yeah, [21:10] buses." And it's like, "Nobody's on the [21:11] bus. Nobody takes the bus. It's a [21:14] beautiful small town. San Francisco's [21:16] Austin." That whole street is empty. [21:18] It's painted red. Okay, so we Okay, so [21:20] we all and complain nonstop on our [21:22] group chat. When are you leaving San [21:24] Francisco? When are you leaving? I'm [21:25] number one, I get it. But Friedberg, [21:26] when are you leaving? Okay, well, so [21:28] >> chat, you're the best by far. Okay, so [21:30] let me just [21:31] A couple glasses of wine in. [21:32] public-facing Friedberg And he's like, [21:36] "You know, I think there's a better way [21:37] to do this." And then there's like Darth [21:39] Friedberg in group chat. He's like, [21:41] "These people, these morons are morons. [21:44] They're destroying society." He is like [21:46] Darth [cheering] Friedberg in group [21:48] chat. [21:49] Am I lying? [21:51] Am I lying? Is he the most Correct. [21:54] Especially after a couple glasses of [21:55] wine. When I start drinking, he's like [21:57] takes a photo of a cliff. And he takes [21:58] pictures. He's like, "Fourth beverage." [22:00] And we're like, "Oh, it's worth staying [22:02] up on the group chat." [22:04] >> [laughter] [22:04] >> And then I'm like, "I'll go and attack [22:05] this congressman on Twitter." which I [22:07] realized Yes, and then you got to delete [22:08] the tweet. Yeah. Don't delete the tweet. [22:11] >> Yeah, I delete the tweets. Okay, so [22:14] >> [laughter] [22:14] >> there's a group of people trying to [22:16] raise $500 million [22:17] to create like a tech/business [22:20] coalition to go to Sacramento, which [22:23] arguably is something that everyone's [22:24] left and avoided doing forever cuz no [22:27] one wants to spend time in freaking [22:28] Sacramento fighting politicians. But [22:30] it's almost like we're all falling off a [22:32] cliff, it's time to do something. Do you [22:34] think there's a realistic path back? Do [22:36] you think the people can actually get [22:37] their together? That even if $500 [22:39] million came in, there's a way to kind [22:41] of turn around the state, fix some of [22:42] the policies. Do you think it's too [22:44] late? [22:44] >> think that Look, I would go I Look, [22:46] anybody who's doing anything to fix [22:49] things, I'm like, "Hell yeah, let's do [22:51] something." The issue is we all grew up [22:54] in the tech world which was like a [22:56] libertarian place where you stay out of [22:59] politics and [23:01] that kind of It was that kind of vibe. [23:02] It was just everybody was like that. [23:05] Leave me alone. I want to make stuff. [23:06] Yeah, I just I I'm not I don't do that. [23:08] And that's obviously [23:10] There's not a thing anymore. In [23:12] California, I think the the ballot [23:14] initiatives are very powerful and [23:16] there's very clean ways to get something [23:18] on the ballot. Love that. I think that [23:21] your DAs who have decided we do not [23:24] enforce crime at all anymore, [23:27] that's like a sweet spot. Like I believe [23:30] that [23:30] I sort of have this aphorism. Truth and [23:33] justice [23:35] are the immune system for society. [23:38] >> [snorts] [23:38] >> When when the immune system is [23:42] suppressed, all the social ills flare [23:44] up. [23:47] So, look for the places where truth and [23:50] justice are being deteriorated, are [23:51] being degraded, and say how do we get at [23:55] that? Because if you get at that, [23:57] everything else downstream will be [23:58] better. [24:00] So, that's kind of how I look at things [24:02] and how I also determine whether the [24:05] world's getting better or worse. When I [24:06] say weird, I'm talking about truth and [24:08] justice. [24:10] That's what I mean when I say, "Oh, man, [24:11] it's getting weird. It's getting [24:12] weirder." Which means it's weird. I'm [24:14] just talking about truth and justice. [24:15] >> Well, I mean, and you look at the [24:17] homeless industrial complex, you look at [24:19] Chesa Boudin, which the all in pod, [24:22] Sachs, myself in the pod, like we we [24:24] literally led the recall of him. And [24:27] then you have the same thing going on in [24:28] LA where they were just like if somebody [24:30] Gascon. Yeah, Gascon. I mean, [24:33] we basically lost the script. You're [24:35] running the city for the criminals. It [24:37] literally is like a Batman movie. It's [24:39] like Bane. [24:40] I mean, here's here's the [24:41] >> to arrest the criminals. [24:44] Look, I was born in the darkness. I [24:46] mean, these guys are lunatics. [24:49] Yeah, look, I I know police officers in [24:51] Los Angeles who are no longer police [24:54] officers. [24:55] And these are lifelong guys who protect [24:57] and serve, that's in their blood or DNA. [25:00] They want to protect people. They want [25:01] the bad guys to be dealt with and they [25:05] they almost have PTSD from [25:08] what it is like to want to serve and see [25:12] bad things happening and not being [25:14] allowed to stop it. Yeah, nobody's got [25:16] their back and they're not allowed to do [25:17] their job. It's it's crazy and I it's [25:20] getting weird. [25:21] >> Okay, hey, I want to just go back to AI [25:23] for [25:23] >> Sorry for the darkness. [25:24] >> no, I think it's good. [25:25] >> I was trying to induce I'm trying to [25:26] induce [25:28] dark Freeburg. [25:29] >> brought it up. [25:30] I mean, someone bring me a tequila, I'll [25:31] get going. Yeah, let's do it. Can we get [25:34] a couple tequilas? [25:35] >> was funny. I went on this podcast [25:36] yesterday and the guy and I the first [25:37] the guy was like the first hour was [25:38] middle of the road. I was talking about [25:39] tech and science and then like politics [25:41] came up. He's like, so socialism. And he [25:43] said like you lost it and then you were [25:45] like he's like the energy went 10x and [25:47] he [25:48] Don't take yeah. So that'll come out in [25:50] a couple weeks, but I was like it got me [25:52] going. Okay, I want to talk about [25:53] physical AI one more time. Yeah. So one [25:55] of now that now that you're doing this, [25:56] I saw a presentation the other day. [25:58] Someone showed like a video of a [25:59] squirrel jumping from one tree to [26:01] another tree and they're like a tenth of [26:03] a watt or something. Like like the [26:05] biology is tuned and it's so perfect in [26:09] terms of its efficiency of energy [26:11] utilization to do physical things and [26:13] we're taking these like big things of [26:15] metal and motors and like actuators and [26:19] if you add up or you compound all of the [26:20] inefficiencies in the system, it's like [26:23] 1200 watts to get the robot to walk four [26:26] foot. Like [26:27] >> Yeah. [26:28] Like break apart not just the software, [26:30] but the hardware layer and where we at [26:33] in evolving things like actuators [26:36] and the materials and everything else [26:38] that's going to make physical AI work [26:41] and scale. [26:42] >> a lot with the questions you're asking [26:44] are going down humanoid lane, which is [26:47] like this thing and and everybody talks [26:49] about how do you do the hand? It's [26:50] almost like Terminator 2 type obsession [26:53] with the hand, which is fair. Like it's [26:55] a very critical part of it. I mean, look [26:57] at the I like to look at the Achilles [26:59] the quote-unquote Achilles tendon of any [27:01] of these machines and you're like that's [27:03] where the action is. This this is a [27:05] couple other places. Um [27:09] Look, [27:10] I'm in the non-humanoid space. I mean, [27:13] but in mechanical engineers have been [27:14] dealing with actuators and you know, all [27:17] the sort of electromechanical sort of [27:20] interactions that make machines do [27:22] certain things, but like I'm in the food [27:24] machine space. So, I can tell you how to [27:27] open a paper bag [27:29] and put a [27:31] put a a bowl in a paper bag without [27:33] tearing the paper bag, but I am less [27:36] into the [27:38] I I forget the name perio they're [27:40] they're the the senses to understand [27:43] awareness and touch. [27:45] Um [27:46] I'm not in that game. Um so, when you're [27:49] mining you're like [27:51] you're not [snorts and clears throat] [27:51] like, you know, [27:53] you know, [27:53] >> You're not threading a needle. You're [27:54] not playing tennis. Um [27:57] Certain things may be equivalent to [27:58] tennis. So, look, the bottom line is [28:00] we're seeing obviously all you have to [28:02] do is go online and look at where the [28:04] humanoids are going over time and how [28:07] much better they're getting. Um it's [28:10] wild and it's happening so freaking [28:13] fast, but any humanoid demo starts with [28:17] dancing and martial arts. [28:20] Yeah. And we're sort of down specialized [28:24] robot lane, which is gainfully employed [28:27] robots. [28:28] >> Yeah. So, I know I didn't totally answer [28:30] the question on like the technology [28:32] piece, but I just like do you agree that [28:34] there's probably like a big opportunity [28:36] for venture money and like research to [28:38] go into material science actually yeah. [28:40] >> For sure. Because if the physical AI [28:42] stack manipulation and all of the [28:46] related things around it [28:49] is massive And so if you get the [28:51] software working it's almost like the [28:52] hardware has to catch up. Yes. We got a [28:55] lot of [28:56] investment to be [28:57] >> I do. Well actually [28:59] it's good that you bring this up. You [29:00] know one of the things you pioneered [29:02] at Uber was [29:04] capital as a weapon and you were very [29:07] thoughtful about hey if we can take this [29:09] capital off the table then that's going [29:12] to let's call it what it is it's going [29:14] to be an advantage versus the [29:15] competitors and these other competitors [29:17] couldn't get that capital. That's now I [29:19] think people have seen that playbook and [29:22] they're like [29:23] someone was like that was smart let me [29:25] try. [29:26] And it's at a different scale. Now that [29:27] you've come out of stealth Yeah. Now [29:29] that you've got and and people are [29:30] starting to understand Yeah. starting [29:32] today Yeah. how big your vision is [29:35] capital as a weapon Yeah. This is I [29:37] guess in your plan yeah? Well I mean [29:39] here's the thing right so capital as a [29:42] strategic weapon for its own sake is not [29:45] a thing [29:46] when it is actually a strategic weapon [29:49] then it is a thing and what I mean by [29:50] that is like [29:51] in the Uber world early days if you [29:54] didn't have capital didn't matter how [29:56] good your app was because Masa's going [29:58] to put a billion dollars in your [29:59] competitor and you're going to lose 20% [30:01] market share tomorrow. So a critical [30:04] competency in fact your world class [30:06] competencies one of them has to be [30:08] raising capital and you need to do it [30:10] better than everybody else and if you [30:11] don't you are going to lose. Let me ask [30:13] one follow up to that. Sorry you go [30:14] ahead. Yeah. Um but the Middle East [30:16] Yeah. I've heard theories the last [30:19] couple days that big capital seekers are [30:21] kind of right now because of what's [30:23] going on in the Middle East with the [30:24] Iran war [30:25] Dubai Qatar [30:27] Saudis are kind of going to close up the [30:30] capital flowing to the US right now and [30:33] is that real? I mean do you think that's [30:34] a real threat? [30:35] >> So look our Middle East business was [30:36] supposed to go public in January [30:40] and the Saudi market was went down 20% [30:45] over like a 2-month period and that was [30:47] like a massive damper on the situation. [30:51] Now, part of it's part of that was [30:53] because the oil prices had gone down so [30:55] dramatically. [30:57] And if you went into KSA, you went to [30:59] the kingdom, everybody's like, "We need [31:01] We need oil prices to go up." [31:03] You know, that's the other side of the [31:05] equation. So, I don't know what happened [31:08] like I don't I'm not in the market [31:10] raising money right at this moment and [31:12] this is a 2-week old thing that I, you [31:14] know, look, I see the news just like [31:16] everybody else and I'm not out there [31:18] calling while the war is going on and [31:20] saying, "Hey guys, you got some money?" [31:22] Um [31:23] so, I don't know exactly what's going to [31:25] happen, but if you are an optimist and [31:27] you're like, "Okay, this isn't This is [31:29] not going on forever." just like the [31:31] tariffs [31:33] it was the end of the world and then it [31:35] wasn't very quickly. If you're an [31:37] optimist about this situation and it [31:39] won't be the end of the world [31:41] maybe even a better world then we get to [31:44] a better place and I think [31:46] progress, abundance, the golden age [31:49] happens and a lot of it is about all the [31:51] things that are happening in AI, in [31:53] physical AI and and and just the [31:56] productivity gains that are coming in [31:58] very massive ways. Yeah, I don't know it [32:01] it was shock shock and awe and then [32:05] hey, now we've got a steady state and [32:06] let's hope that's what happens in Iran [32:08] is that we can depose these evil [32:11] dictators and replace it with something [32:13] a little more stable. And related to [32:14] this before we wrap, they're going to [32:16] China, there's big trade deal being [32:17] negotiated. What do you hope comes out [32:19] of this Chinese thing? And what did you [32:21] learn in China? Yeah, what would you [32:23] learn and and what do you think would be [32:24] great for America like what would you [32:26] like to see and be like, "Man, that's [32:27] going to set us all up. No more Look, [32:29] here's the thing, if you go to China [32:30] right now and you go and just take a [32:32] tour of the manufacturing that's going [32:34] on there just the manufacturing base [32:38] The cities, especially if you've gone to [32:40] China for a couple decades in a row. [32:43] You're like, damn. Yeah. You so, let's [32:46] just do two things. You go to Shenzhen, [32:49] which before felt like Kansas City, but [32:52] 50 years ago and really humid, which I [32:55] guess Kansas City sometimes. But, [32:58] you go there now and it's like [33:00] one-upping Singapore. [33:02] Right? [33:04] Or so that's the city view. You're just [33:07] experiencing a very awesome, you're [33:09] like, [33:10] this is advanced. [33:12] And you just get the vibe and it's [33:13] everywhere. And then you go and you [33:16] start seeing the manufacturing base and [33:17] you see what like Xiaomi is doing or any [33:21] of the other there's so many scrappy [33:23] guys, badass guys everywhere and you're [33:25] like, [33:27] F. They're hungry. So, does anybody [33:30] remember the 2008 Olympics in Beijing? [33:33] Anybody? Does anybody this is a little [33:35] bit we're down rabbit hole. Does anybody [33:37] remember the opening ceremony? [33:40] And you're like, these mother [33:42] are taking over. At least that's what [33:44] they want to do. [33:46] That shit's happening. [33:48] So, [33:49] I don't have any issue or this is not [33:52] negativity for me. I'm like, these guys [33:53] are killing it. The best idea is [33:55] winning. They're fiercely they're [33:57] fiercely going after truth and progress [34:00] and they're making happen. Let's [34:02] step up our game, okay? But we can also [34:05] have a friendly game. Like we don't have [34:07] to like be like the Detroit Pistons in [34:09] the '90s, you know? [34:10] >> Yes. We can we there there's a way [34:13] >> the stands. Yeah. Yeah, you know. [34:15] There's a way to do this right and [34:17] there's a way to do it like adults. I [34:19] hope that's where we would end up. [34:21] Um I have a employee who cuz we we were [34:25] we for for a long time were the largest [34:28] built kitchen builders in China. I I an [34:30] employee in China has an American wife. [34:34] Okay? [34:35] They both live in China. [34:37] They're both from China originally, [34:38] okay? [34:39] But [34:40] one of one of them to It would be great [34:42] for him to work here on some things I'm [34:44] doing. It's very hard to make that [34:46] happen right now. Now, that's selfish [34:49] like I like maybe selfish like I'm like [34:51] there's a person I've been working with [34:52] for over a decade. I'd love to continue [34:54] here. Maybe there's other bigger picture [34:58] items that I'm not dealing with. I'm not [34:59] the geopolitical guy, but I'd love for [35:02] there to be sort of [35:05] good [35:07] relations and good like like if you have [35:11] a significant other in who is an [35:13] American citizen, like do we have to [35:16] make that hard? Is an example. [35:18] >> Some normalcy would Something, you know, [35:20] I'm just saying. Now, I agree like [35:23] there are ways to do immigration [35:24] properly. Like we effed it up super bad. [35:28] Don't even get me started, but there's [35:30] also there's good migration too. Like a [35:33] lot of great innovators [35:36] all over the place came from other [35:38] places for their own version of the [35:40] American dream. God bless. Free bird. [35:42] And we don't have to [35:44] that doesn't have to be a negative [35:46] thing. [35:47] And so I'd like to see more of that and [35:50] um [35:50] yeah, China's China's wild. So let's [35:52] let's keep our eye on the ball and let's [35:54] let's give them a run for their money, [35:55] too. [35:56] >> it up for TK. [35:57] >> [applause] [35:58] >> All right. Well done, brother. Thanks, [35:59] brother. [36:00] That's good. [36:01] Good to see you, brother. [36:03] Wow. Michael Dell. [36:06] >> [applause] [36:07] >> My lord. Texas native. [36:10] Michael [36:11] >> born in Houston. [36:12] >> Well, I missed the the opening. We we [36:14] jumped to the music, but you started [36:16] Dell computer here [36:18] in Austin with a thousand bucks. 42 [36:20] years ago in my dorm room at uh Doby at [36:24] UT. [36:26] Uh about a 10 days before I finished my [36:29] freshman semester. [36:31] Moving. [36:33] And it's been [36:34] working out pretty good. [36:36] >> [laughter] [36:37] >> Yeah, there's been some bumps in the [36:38] road, but yeah, it's generally generally [36:41] worked out okay. You know, we'll have [36:42] about 140 [36:44] billion in revenue this year, so [36:47] Yeah, it's okay. [36:48] >> Hey, it compounds over time, doesn't it? [36:50] >> Yeah, yeah, yeah. You know, you start [36:52] small and just keep adding and [36:54] there you go. That's how it goes. It's [36:56] just that easy. [36:57] >> [laughter] [36:59] >> But [gasps] why Texas? Like I think this [37:01] is an important thing. We're in Austin. [37:03] Jason lives here. David Sacks lives here [37:05] now. [37:06] More people are moving from California [37:09] to Austin. Why Austin? Why Texas? Why is [37:11] it work here? And is it getting better? [37:14] Or has it just always worked? [37:16] You know, I think Texas has had a [37:19] uh, you know, low-tax, [37:23] pro-growth [37:25] uh, environment for a long time and pro, [37:28] you know, sort of progressive business [37:31] climate. And you know, if if you sort of [37:33] look at the the growth of [37:35] the Texas economy relative to the rest [37:38] of the United States without Texas, [37:41] you know, Texas just kind of looks like [37:43] a better version of the US economy. [37:45] >> [snorts] [37:46] >> And uh, you know, you now you've got [37:49] uh, Austin is sort of just about in the [37:52] top 10 cities in in the United States. [37:55] So, you've got [37:57] uh, [37:58] when that happens, you'll have four of [37:59] the 10 largest cities in America in [38:01] Texas. [38:03] One out of 10 children born in the [38:05] United States born in Texas. [38:08] Uh, more New York Stock Exchange [38:10] companies in Texas than [38:12] in New York or anywhere else. [38:15] And uh, [38:16] you know, you've got University of Texas [38:18] here in Austin, which I would always [38:21] think of as kind of the wellspring for a [38:23] lot of the companies that are here, [38:25] certainly ours. [38:27] And [38:29] you know, long history of of uh [38:33] innovative, pioneering spirit, and [38:36] entrepreneurship. And [38:39] it's been a fantastic place for us. And [38:42] part of this, I think Freeberg and [38:44] Michael, is what's happened in the other [38:46] great cities, or what were once great [38:49] cities. My hometown, New York, got to [38:51] spend 10 years in LA and in the last 12 [38:54] in [38:54] the Bay Area. And what's happening there [38:57] is incredibly un-American, [39:00] uh and they're decelerating when [39:02] compared I think maybe the gap maybe in [39:05] the disparity from these [39:07] um two locations has gotten greater, [39:10] yeah? And you're seeing a lot more [39:12] people say [39:13] life there, here in Austin, seems a lot [39:15] better than the life I'm living in New [39:18] York, LA, [39:20] or in in the Bay Area. Yeah, well, I've [39:22] got a lot of new friends and neighbors, [39:24] you know, that that have that have that [39:25] have come. And certainly, I mean, if if [39:26] you look at the the the migration [39:29] statistics, Texas has attracted an [39:33] enormous number of people. And and look, [39:35] I mean, when you when you when you look [39:38] at the environment here and compare it [39:40] to the other kind of situations that are [39:43] going on, [39:44] uh it's it's it's very attractive. But, [39:47] you know, it's it's kind of been great [39:49] for a long time. So, [39:51] uh [39:52] it's not really new news to us that have [39:54] been here a while. [39:56] Yeah, Elon had a great experience when [39:58] he was building the the Gigafactory over [40:01] here. They let you do stuff here. Yeah, [40:03] basically, you know. They let him build [40:05] it. [40:06] Which he said it was like an incredible [40:08] experience for him, because in [40:09] California, they didn't let him build [40:12] you know, these factories. He In fact, [40:14] the Tesla factory that in Fremont was [40:16] just an old ancient factory that he was [40:18] able to retrofit. So, [40:20] uh there's something going on here as [40:21] well with the data centers. And then [40:23] that's actually I think very close to [40:24] what you're working on at Dell. Maybe [40:25] you could talk a little bit about the [40:26] data center boom that's going on in [40:28] Texas that maybe people aren't paying [40:30] attention to. Sure. Well, there's, you [40:32] know, obviously been enormous build-out [40:35] of AI infrastructure and that requires, [40:39] you know, lots of new data centers, lots [40:41] of power. Texas, [40:43] you know, [40:44] has an enormous uh advantage there [40:47] relative to other states. A lot of [40:49] power, a lot of land, and it's and and [40:53] you can build stuff, right? So, there's [40:55] there's been a massive build-out [40:57] particularly in some of the cities in [41:00] and towns in West Texas where there's [41:02] not a lot of population. And so, they're [41:04] not really too opposed to having data [41:06] centers [41:07] out in the middle of nowhere where [41:09] there's land and power. And so, [41:11] um [41:13] Yeah, I mean, the the [41:15] the demand for tokens is enormous. [41:18] You know, we've been building these AI [41:21] data centers not just here in Texas, but [41:24] around the world. And, you know, the [41:27] growth in that has been tremendous. [41:30] You know, we we introduced the the first [41:32] H100 server It was literally a couple [41:35] weeks before Chat GPT was announced. [41:38] >> [snorts] [41:38] >> And, you know, [41:40] uh the progression of our business in [41:42] that area sort of [41:44] gone from like 2 billion to 10 billion [41:47] to 25 billion to this year it'll be like [41:50] 50 billion. So, so tremendous growth. [41:53] And [41:54] when you when you think about what these [41:56] models are creating, [41:59] there's this face change that's happened [42:01] in computing, right? We we we had 60 [42:04] years of calculating and computing. Now [42:06] we have machines that are thinking and [42:08] helping us think. And so, the demand for [42:11] that kind of intelligence and you know, [42:13] the models are amazing, but they're also [42:16] the worst they'll ever be and they're [42:17] continuing to improve. And so [42:20] we just see [42:21] uh [42:22] a lot more [42:24] demand than supply and it's happening [42:27] not just in [42:28] the hyperscalers and the cloud service [42:31] providers, it's happening in 4,000 [42:34] enterprises where we building these [42:36] these Dell AI factories. [42:38] It's happening in sovereign AI, you [42:40] know, like Palantir and you know, people [42:43] want to protect their data, but also use [42:45] AI on it. They want to bring the AI to [42:47] where their data is. [42:49] And you know, when this kind of started [42:51] a few years ago, we had some [42:54] really sophisticated [42:56] uh large companies, think think of like [42:58] Fortune 100 and they started, you know, [43:01] buying these AI servers from us and [43:05] and uh they kind of knew what they were [43:06] doing, right? You know, and and uh we [43:08] said, "Well, what what are you doing?" [43:09] All right. Yeah. And [43:11] >> [clears throat] [43:11] >> they they were they were kind of taking [43:13] uh building their own models. They were [43:16] taking open source models. They were [43:17] running them. Some of them were were [43:19] algorithmic traders or, you know, [43:22] derivatives of machine learning. [43:25] And of course um [43:27] they needed a lot of help in in doing [43:29] that cuz [43:30] it it it was sort of a complicated [43:32] thing. So, about 2 years ago we we put [43:34] together this [43:36] product that we we we we we called the [43:39] Dell AI factory. And now we've got 4,000 [43:43] plus of these and it's kind of running [43:46] rampant across enterprises. How do you [43:49] think about [43:51] the payback time on the investment [43:54] that's being made? The administration [43:56] put in place this accelerated [43:58] depreciation rule or by the company [44:00] >> very helpful actually. Yeah. [44:01] >> So, just for folks to understand that a [44:03] little bit, like if you spend a hundred [44:06] billion dollars this year building and [44:08] data centers and buying infrastructure [44:10] for those data centers, you get to write [44:12] off a hundred percent of that this year. [44:14] Correct. [44:15] >> it, so you don't pay taxes. You pay way [44:17] much fewer taxes. [44:18] >> in place for 10 years, I think. [44:20] >> That's a 10-year deal, so it's [44:21] accelerating the investment. How much is [44:25] that helping [44:26] versus [44:27] how are you seeing folks rationalize the [44:29] investment relative to the return [44:31] they're going to make and over what time [44:33] scale? This is still the big question. [44:35] Is the money really there? The [44:36] hyperscalers, maybe they're starting to [44:37] come up, but end usage, end states, [44:41] are we kind of, hey, wait and see, we [44:43] don't know yet, or folks are getting 20% [44:46] ROIC starting in year one after they've [44:49] made the investment? You know, I I I can [44:52] tell you in in our business in in our [44:54] company, we definitely see plenty of use [44:57] cases where the ROI or the improvement [45:00] in productivity efficiency is [45:03] 20% or or greater. The right away gets [45:06] there. Yeah, I mean, it it's you know, [45:09] it's not like you just hit a button and [45:10] you get 20%, right? There There's [45:12] There's work required in thinking [45:15] through the processes. And you know, [45:17] it's worth a little bit describing that. [45:19] So, you know, when you have a a [45:22] any company, [45:24] its processes and tools and technology [45:26] are a function of what was available at [45:28] the time it created those things. [45:31] And so, what you sort of have to do is [45:33] step back and say, all right, what's the [45:35] trajectory of the improvement of the [45:37] tools? [45:38] What outcome are we trying to create? [45:42] And now let's simplify and standardize [45:45] the processes, [45:47] get all the tools together, get all the [45:48] data together, and then apply the [45:51] technology. [45:52] And this really has to be done in kind [45:55] of a tops-down way. [45:57] Uh you can't sort of do it spontane- in [46:00] you know, in in [46:01] silos are not going to spontaneously [46:03] improve themselves and often that means [46:05] that you're completely changing [46:08] the way the organization works. It's [46:10] like a wholesale re-architecture. [46:12] It's a it's a re-imagining of the way a [46:15] company works and you know, I mean the [46:17] way I described this to our team [46:19] about 3 years ago is [46:22] you know, we were going to have a new [46:24] competitor 5 years from now that would [46:26] be 2 years from now, you know, that was [46:28] in every business that we're in except [46:30] they were going to be faster and more [46:32] innovative and more successful and lower [46:35] cost and they were going to put us out [46:37] of business. [46:38] And the only way we were going to [46:39] prevent that is is we're going to become [46:40] that company and here's how we're going [46:42] to do it and [46:44] you know, it [46:45] excited some people, it scared some [46:47] people and but I actually believe that [46:49] that's [46:51] what's going to happen and and so we've [46:53] been [46:55] dramatically changing our business. I [46:57] would say the biggest benefit by far is [47:00] speed. We're much faster at being able [47:03] to apply innovations and so, you know, [47:06] you look at our at our infrastructure [47:08] business last quarter grew 73%. Well, [47:11] that's [47:12] kind of unusual for a business of this [47:15] size. [47:17] And [47:18] you know, this quarter we guided that it [47:21] would grow even faster like 100% so. [47:23] You've lived through a couple of [47:24] paradigm shifts here. The PC revolution [47:27] obviously you led that and then you of [47:29] course had you know, client server, the [47:32] network revolution, online, [47:35] internet, mobile. [47:37] >> mobile. [47:38] >> Yeah. So, each one of those we saw [47:40] massive disruption. We're talking in in [47:42] the green room about hey, we used to [47:43] have a typing pool, there was a mail [47:45] room. All these things got abstracted [47:47] away by the PC and networked PC [47:49] revolution. [47:51] But it took a decade or two. [47:54] And this one's happening a lot faster, [47:56] yeah? [47:56] >> Yeah, this one I think it's it's like um [47:59] you know, a quarter is like a year. [48:01] Maybe maybe it's five times faster or [48:03] something like that. But but back to [48:05] your question, I I I would say maybe 10 [48:08] or 15% of large companies have really [48:11] figured this out and the rest of them [48:13] are kind of fumbling around and you [48:15] know, there's a tendency when when you [48:17] hear about a new technology to like oh, [48:20] let's just let's just go do it, you [48:21] know, and show the boss, hey, we did AI, [48:23] you know. Yeah, the board said we got to [48:25] do AI. We got to do AI, guys. [48:27] >> We need AI. Are you Are you proud of me, [48:29] boss? You know? Yeah. Um [48:31] and Look at what I made. [48:33] Exactly. And I also think, you know, is [48:36] an important point about about uh this [48:39] which is [48:40] you know, the barrier to technology [48:42] adoption [48:44] is is not technology. [48:45] It's culture and leadership and courage, [48:49] right? And and so Willingness to change [48:52] and to adapt yourself. [48:53] >> and you know, if you if you're in a [48:55] business that you don't think it's [48:56] changing very much or you know, heart [48:58] change is really hard, right? And [49:00] you you have to it can be very [49:02] uncomfortable. You're like, well, we're [49:03] going to stop doing that. Well, maybe we [49:04] don't need this anymore. Particularly if [49:07] your bonus is dependent on not messing [49:09] things up. [49:10] But [49:11] let's go let's use the internet as an [49:13] analogy which you saw [49:15] up close. There were [49:17] businesses that were internet transition [49:21] successful. They made the transition. [49:23] Maybe Macy's dot com versus Sears [49:26] Roebuck, right? [49:27] Maybe Macy's did a better job of taking [49:30] advantage of the internet than Sears. [49:32] But then there was internet native [49:34] businesses that seemed to blow them all [49:35] out. [49:36] And uh maybe Amazon's a good example or [49:39] um CSN stores, whatever they be Wayfair, [49:42] etc. [49:43] What's the right way to think about this [49:45] evolution in industries generally? Are [49:48] we going to have [49:50] there businesses that are going to [49:51] transition successfully and those that [49:53] aren't and they're going to die. And is [49:55] this really going to Are we going to see [49:56] AI-native businesses in every industry [49:59] come in and just disrupt everything? I [50:01] believe we will and certainly, you know, [50:03] when you talk to the Collison brothers [50:06] at Stripe, they'll tell you that the [50:09] rate of growth of the 2025 cohort [50:12] companies is about four times faster [50:15] than the 2018 companies. And so, [50:18] every year the new batch of companies [50:20] are growing faster and faster because [50:22] they're starting with all these new [50:23] tools that, you know, Well, because they [50:24] see all the new companies on their [50:26] platform. Exactly. And so, when when you [50:29] think about [50:30] an incumbent company, okay, that already [50:33] exists, it has, let's say it's got [50:35] brands, it's got balance sheets, it's [50:37] got, you know, customer relationships, [50:40] whatever stuff, right? Okay? [50:42] Um but that's sort of like those are [50:45] expiring value assets. If it doesn't [50:47] change quickly and get on the other side [50:50] of this, I think it will go out of [50:52] business. And which is exactly the [50:54] speech I gave to our team, you know, [50:56] three three years ago. And [50:58] Uh [50:59] I think, [51:00] you know, [51:01] you you you you you have to be [51:04] uh bold and you got to go make those [51:07] changes to to [51:09] not only survive this but but to but to [51:11] thrive and and, you know, I think about [51:14] it is how do we prepare our company to [51:15] be ready for the 2030s? [51:17] >> Right, isn't it like it's much more [51:20] it's kind of the storyline. There's more [51:22] to do than there ever was. It's like [51:25] when the internet arrived kind of came [51:26] around, Sears doesn't just need to sell [51:28] locally, they can sell to the world. [51:30] Well, sure. I mean, [51:32] this is the point. [51:33] >> AI lets everyone do everything. [51:35] >> When when when we have better tools, we [51:36] can do way more things, right? And and [51:39] you know, when it when you know, when I [51:40] hear people say, oh, you know, [51:44] um maybe we're just going to have all [51:47] these great tools and we we won't do [51:49] more things. We'll just do the same [51:51] things with fewer people. [51:53] It doesn't sound right to me. I I mean, [51:55] there'll be some of that, but I think [51:56] most of it will be we're just going to [51:58] do a whole lot more things. We're going [51:59] to solve a lot more problems. We're [52:01] going to accelerate scientific [52:02] discovery. [52:04] We're we're going to invent all sorts of [52:06] new things. We're going to solve all [52:07] sorts of problems that haven't been [52:09] solved. [52:11] And, you know, that's that's super [52:13] exciting. What do we have wrong on [52:14] infrastructure? So, [52:17] the original build cycle looked a lot [52:19] like everything's in a data center. [52:21] Everything's got to sit there. That's [52:23] where all the intelligence, it'll all be [52:25] in these kind of hosted proprietary [52:27] cloud models. Do you think that it's [52:29] open source? Is it distributed on the [52:31] edge? Where does the intelligence, where [52:33] does the inference sit? And how does [52:35] that really change or kind of [52:36] re-architect the industry, do you think? [52:38] It's really all the above. I mean, it [52:41] it's not like there's one answer. I [52:44] mean, certainly if you go to [52:47] uh [52:47] any industrial company or natural [52:50] resources company, advanced [52:52] manufacturing, [52:54] retail, logistics, there's tons of [52:57] inference at the edge. [52:58] And that's growing very, very fast and [53:02] uh you know, we make a lot of that [53:04] embedded equipment. Certainly, you know, [53:06] telcos are doing that too. I mean, it's [53:07] pretty much every every industry. Think [53:10] about wherever data is being created, [53:12] you want the AI infrastructure and the [53:15] inference, you know, close to the data. [53:18] Um [53:20] you know, there there there has been [53:21] this sort of rebalancing as companies [53:24] have figured out, you know, sort of [53:26] everybody loves the public cloud, right? [53:29] Until they get the bill, right? When [53:31] they they get the bill, they're like, [53:32] wait, this is supposed to save us money, [53:33] but [53:34] cost quite a bit more. So, you know, the [53:37] the the lowest cost token is going to be [53:39] the one that's generated right where the [53:41] data is, on the device. You're going to [53:44] have, you know, tokens being generated [53:46] on your phone, on your PC, [53:49] in every embedded piece of equipment. [53:51] And And look, we have an interesting [53:53] perspective on this business cuz we have [53:54] 10,000 customers where they embed our [53:58] product in their product. This is, you [54:00] know, think medical devices, security, [54:03] all sorts of things in hospitals and [54:05] industrial plants and, you know, any any [54:08] kind of [54:10] uh [54:11] you know, [54:11] uh [54:13] data-driven activity, right, requires [54:16] some kind of computing network storage [54:18] infrastructure. Yeah. So, when you um [54:21] look at the desktop where you started, [54:23] it's coming full circle and this must be [54:26] at least very interesting or intriguing [54:28] to you that you see this open claw [54:30] movement, everybody trying to buy the [54:32] most powerful desktop they can, and all [54:35] these hobbyists who were your customers [54:37] who were calling you up and ordering [54:39] from, you know, Dell, their their [54:41] bespoke PC. [54:42] Now they're Dell.com. [54:44] >> Dell.com. [54:45] What did I say? Dell.com. Yeah. You said [54:47] ordering from Dell. Calling us up. They [54:49] They They order online usually. [54:50] >> They order online now, yes. They have [54:52] this thing called the internet, Jason. [54:54] >> yes. It works out pretty well. Um [54:56] But this is incredible that they're like [54:59] all stacking computers and and running, [55:03] you know, uh local models. [55:05] I was just thinking back to how much the [55:07] first couple of computers I owned cost, [55:08] $4,000 in 1980. [55:12] And then the prices came down, you could [55:14] buy a Dell for 500 bucks, 800 bucks, [55:16] like really nice laptops for that price. [55:19] Um [55:19] use the promo code all in. [55:21] Um [55:22] This is not a sponsor, it's a joke. Um [55:26] But do you think there's a world where [55:27] we're going to start to see [55:29] the desktop, because people want to [55:31] protect that data, they want to protect [55:33] the skills they're building, they don't [55:35] want to give it to Sam Altman, and it in [55:36] a cloud somewhere. They don't want to [55:38] give it to Google, whoever it happens to [55:40] be. [55:41] Um and that the desktop revolution comes [55:42] back, and everybody's got a $10,000 [55:44] desktop. Is that coming? [55:46] I don't know if everyone will have a [55:47] $10,000 desktop, but that would be [55:49] great. I mean, you know, uh [55:52] Um you know, so we have this [55:56] tel- uh portal on Hugging Face. And we [55:59] have all these open models, and we [56:01] qualified them on every kind of machine [56:04] we have. And you know, there's been [56:06] enormous progress in the open source [56:08] models. You know, Google has these Gemma [56:11] models, g e m m a, and they work really, [56:13] really well on small machines. You know, [56:16] Open AI has their open source models. [56:18] You've got the Nvidia NeMoTron models. [56:21] You've got, you know, enormous [56:23] uh ecosystem of open source that is, [56:28] you know, thriving, and certainly Open [56:30] Claw, and you know, there'll be some [56:33] good discussion about that. How many [56:35] people have set up Open Claw? Raise your [56:37] hand. [56:38] Oh my lord, that's about what, 20% of [56:41] the audience here? Yeah, so you know, [56:43] autonomous agents, [56:45] um big deal, and certainly inside [56:47] companies, [56:49] there's going to be a lot more [56:50] autonomous agents. [56:52] There are significant security [56:53] requirements that need to go with that. [56:54] We need to We need to be able to [56:56] authenticate and validate who these [56:59] agents are, and what they're doing, and [57:01] you know, have the right controls and [57:03] and and that sort of thing. Yeah. And uh [57:07] your take on uh AGI, and [57:11] when we're going to hit it. Do you [57:12] actually think about superintelligence [57:14] and AGI and the [57:16] the the two sets of problems they could [57:19] solve there. And do you have a personal [57:21] definition that you like to use for [57:22] those when you're talking internally [57:24] with your team of how things are moving? [57:28] I I I don't really know, Jason. Uh [57:31] um you know, I I think if it feels like [57:34] with the latest releases, we were [57:36] talking about this backstage, [57:38] you know, the Gemini 3.1, [57:42] the Opus 4.6, the OpenAI 5.4, it feels [57:46] like we've sort of [57:47] hit some kind of threshold where the [57:50] just the quality of the the models are [57:53] are just tremendous. And [57:55] when I listen to what our teams are able [57:57] to accomplish in a day or two weeks that [58:01] would have taken them, you know, a few [58:02] months or 9 months time, [58:05] you know, it's it's just amazing the [58:07] speed of innovation. Yeah. And so [58:11] it it seems to be continuing [58:14] and we get all the reinforcement [58:17] learning and there's also tons of [58:20] private dark data [58:22] that these models haven't been applied [58:24] to. And that's sort of what's happening [58:26] with these I think the auto research is [58:28] the that's the the key with auto [58:30] research, the capacity to take a [58:32] standard model and then retrain it on [58:35] your private data and keep it private [58:37] and build an advantage for your [58:39] organization based on the history of [58:41] your data that no one else has. That [58:43] seems to be what a lot of folks are [58:44] thinking about that have [58:46] the capacity. But if you were to start a [58:47] company today that was not in computing [58:50] and you were to build a business from [58:51] the ground up, how would you architect [58:54] your people and your organizational [58:55] principles [58:57] as [58:58] an AI kind of first knowing what you [59:00] know about computing and where things [59:02] are headed? Are you hiring people? Are [59:05] you hiring a bunch of people to run a [59:06] bunch of agents? How do you think about [59:08] architecting a new business today? It's [59:11] a great question. I don't really spend a [59:12] lot of time thinking about that. You [59:14] know, I'm thinking about how do how do I [59:16] What the rest of us are thinking about. [59:17] How do I run our company? I mean, that's [59:20] that's hard enough. So [59:22] Yeah. Everyone I talk to, that's the [59:24] question. They're like everyone goes to [59:25] these offsites and they're like [59:27] I'm actually doing this with my [59:28] management team on Monday. We're doing [59:30] like a teardown and be like, "Hey, how [59:31] would we build the business differently [59:32] today?" Yeah, I mean [59:34] what we've been thinking a lot about is [59:38] the sort of this this reimagining [59:40] question. [59:40] >> Yeah. [59:40] >> You know, sort of all right. [59:43] We know the trajectory of the tools. [59:45] What are the tools going to be in 27, [59:47] 28, 29? And how do we [59:52] accelerate, you know, our path to that? [59:55] How worried are you about [59:58] social issues? So, AI recently ranked as [60:01] the most unfavorable term of a list of [60:04] terms including president [60:06] >> saw that. [60:06] >> somewhere between ISIS, the Democrats [60:09] >> Ice ISIS and the Democrats. [60:11] >> Ice was better than AI. [60:13] People liked Ice, masked agents more [60:16] than they liked AI. [60:17] >> Yeah. [60:18] >> Well, I think I think part of the [60:19] problem is it's been it's been uh [60:22] you know maybe sold as [60:26] as you know, it it sort of presents [60:27] itself like a human would. Yeah. Right? [60:30] And you know, maybe if we called it [60:33] linear algebra matrix calculations [60:37] >> instead. [laughter] [60:38] >> Right. Matrix multiplier [60:39] >> that would be more friendly. I don't [60:41] know, you know. Yeah. But do you think [60:42] do you think we're going to have [60:44] >> No, I I think you're right. The [60:45] positioning is wrong and then we're not [60:47] communicating to people, "Hey, this [60:49] could help healthcare. This could make [60:51] you live longer. This could help your [60:52] kids get educated more. That this could [60:54] help with housing costs. This could help [60:56] with food costs." [60:57] >> messaging aside, I mean how much do you [60:59] actually worry about disruption or [61:02] dislocation and employment about [61:04] acceleration of earnings for some people [61:07] and deceleration for other people in [61:09] society that feel left behind? And that [61:13] starts to fuel more of the kind of [61:15] social concerns and politicians saying, [61:17] "Hey, we got to stop building all the [61:19] data centers." You know, like that kind [61:20] of stuff. And and how much are you [61:22] really [61:23] >> tend to be, you know, more optimistic [61:26] and and um [61:29] you know, I I do I do think that in all [61:31] technology cycles you get sort of these [61:34] network effects. [61:36] And [61:37] uh that's kind of inevitable. But I also [61:40] think, you know, [61:42] uh [61:43] we're going to do more with the tools. [61:45] You do have this acceleration of all [61:49] sorts of uh great things. I mean, [61:51] education can dramatically improve, [61:53] scientific discovery, health care, [61:55] energy, you know, all sort of the the [61:58] unsolved problems can can be [62:00] accelerated. Ultimately, I think it's [62:02] it's amplification of human potential [62:05] and capability. [62:06] >> And Extending the frontier, too. And And [62:09] And And by the way, we should also [62:10] remember that basically what we're [62:12] talking about here, beyond sort of some [62:15] of the advanced semiconductors in the [62:17] you know, big data centers, we're [62:18] talking about software. Right? [62:21] Yeah. Right? It's like software that [62:23] runs on your computer. [62:25] >> [snorts] [62:25] >> So, [62:27] you know, somebody says, "Well, we don't [62:28] we don't want that." It's like, how do [62:30] you stop total software? I mean, [62:31] >> [laughter] [62:32] >> Yeah, I know. It's How are you going to [62:33] not Are you going to stop someone [62:34] putting an open-source model on their [62:35] computer at home and asking it for [62:36] medical advice? You know, in New York [62:38] just passed a law saying AI models can [62:40] no longer give medical advice. [62:42] >> Yeah. Right? It's being proposed. [62:44] >> Yes. It's being proposed. Yes. You can't [62:46] give legal and health advice. We're [62:47] anti-software. It's like Yeah, well, [62:49] we're also anti- my books and advice. [62:51] So, if you were going to look it up in a [62:52] book, but were you dropping into your [62:54] Bernie Sanders right there? [62:56] >> That was my Bernie Sanders. [62:57] Michael Dowell Do Do you have a the 1% [62:59] of the 1% that you're enabling with your [63:01] data centers. [63:03] Why are you doing this to the people of [63:06] our great nation, while you give your [63:08] money to children in the Invest America [63:11] accounts? Yeah. This is a good one [63:13] you're doing. Yeah. Can we talk about [63:15] Invest America? [63:15] >> I think [63:16] >> might have even criticized that, but but [63:18] you know Well, that's [laughter] the [63:19] problem. The billionaires are giving our [63:21] children money and they're not asking us [63:22] permission and then those kids are going [63:24] to buy things that their parents never [63:26] asked for. [63:29] Well, they they don't they don't [63:30] actually get the money till they're 18 [63:32] years old. So, that's that's [63:34] But what gives you the right to give our [63:36] children an education? What is this [63:38] philanthropy? It makes no sense. [63:40] No, I mean honestly [63:42] >> Well, I see I see my great friend Brad [63:44] Gershner here. Brad's here? [63:47] There he is. Brad, come up for this [63:48] little segment here. Sit sit for a [63:50] second. Let's talk about Invest America. [63:51] We got 5 minutes left here. So, you [63:53] know, I heard about [63:54] >> Everybody in the fifth best, give him a [63:55] round. [63:56] I didn't know he was going to be here. [63:58] Here you go. [63:59] >> [applause] [64:00] >> All right. [64:01] I I [64:02] >> How did this go down, Michael? How I I [64:03] heard about this idea [64:05] in 2021 from Brad. [64:08] And I thought, you know, that's a that's [64:10] just a great idea. That's an awesome [64:12] idea. [64:13] And you know, I think there were some [64:15] discussions with the prior [64:16] administration, [64:18] but they didn't uh do do anything about [64:20] that, [64:21] unfortunately. [64:23] >> [snorts] [64:23] >> And uh you know, here we are, you know, [64:26] a miracle. You know, [64:28] uh the the the Invest America Act was [64:30] was passed and [64:32] um you know, now we have thousands of [64:34] companies that are joining in and [64:37] matching the government's contribution. [64:40] And uh you know, Susan and I made a big [64:43] announcement uh giving $250 to 25 [64:47] million children in [64:49] uh zip codes where the median income is [64:53] >> [applause] [64:53] >> I mean, Michael, let's just let's talk [64:56] for a second here. [64:58] This is [65:00] one of the greatest [65:01] >> What do you think, Bernie? Do you [65:02] approve? [65:03] >> [laughter] [65:04] >> I'm going to Jake out of this place. [65:06] I just want to pause on this because it [65:08] is one of the greatest philanthropic [65:10] gifts in the history of humanity. And I [65:12] I [65:13] people have just kind of glossed over to [65:15] it cuz there's a lot of big numbers in [65:16] the world, but we're talking about you [65:18] you personally, you and Susan, sat down [65:20] and said, "We're going to give a [65:21] number." And that number was 5 6 7 [65:25] billion dollars? Or this is [65:27] Well, it's it's $250 [65:30] to 25 million children ages 2 to 10 in [65:34] zip codes where the median income is [65:35] $150,000 or less. It's It's $6.25 [65:39] billion. I mean, [65:41] >> [applause] [65:41] >> I'm a And I just want to say something. [65:43] You know, we live at a time [65:46] where [65:48] But, they have to sign up to claim the [65:50] accounts. [65:50] >> Yes. [65:51] >> They They have accounts, but they have [65:52] to sign up to claim the accounts. You [65:54] know, I think we're getting 100,000 plus [65:56] kids now a day signing up. [65:58] >> Yeah. First, thanks for having me up. I [66:00] mean, what what a national hero and [66:03] national asset that my friend Michael [66:06] Dell is, but he understates this [66:08] because I've been working on this for 4 [66:10] years. [66:11] We've been talking about it on the [66:13] All-In pod. We had a lot of momentum, [66:15] but behind the scenes Trump gets [66:18] elected. [66:19] Um [66:21] and so it's April that we're in the [66:23] middle of the tariff strife, April 25. [66:26] We realize there's only going to be one [66:27] piece of legislation that gets passed [66:29] during Trump's first 2 years. It'll be [66:32] the you know, this big beautiful bill, [66:33] the reconciliation bill. [66:36] And so I I call up Michael and I said, [66:38] "Michael, we got to go. We've got 5 [66:40] days. [66:41] We have the It It's drafted in the [66:43] Senate. We have bipartisan support, but [66:46] we have a window. And like we have I [66:49] have to get in the Oval Office. We have [66:51] to get in the Oval Office." [66:53] And you know, Michael said, [66:55] you know, "What What should the text [66:57] say?" [66:58] And you and I had a conversation and you [67:00] you know, [67:01] uh But You know, the text to DJT. Yes. [67:04] I'm not talking out of school. Listen, [67:06] Biden [67:08] what wherever you sit on the political [67:09] divide, I will say I've said this, Trump [67:11] seeks out [67:13] ideas from business leaders. And he has [67:15] deep respect for business leaders like [67:17] Michael Dell. And like wherever your [67:18] politics are, that's just the truth. And [67:21] the last administration didn't. And you [67:23] know, If it was Hillary Clinton, she may [67:25] have done the same thing. [67:26] >> this and I just have to say this Invest [67:28] America, it's not a red idea or a blue [67:30] idea. It's a red, white, and blue idea, [67:32] right? It's it's [67:34] >> So, [67:35] And [applause] to and to the prior [67:36] conversation, [67:38] when Michael and I first talked about [67:39] it, [67:40] um [67:42] you know, [67:43] it was this is the right thing to do. [67:46] Right? Like we have to reconnect the 70% [67:48] of people who feel left out and left [67:50] behind to the American dream. [67:52] Right? But this is in our self-interest. [67:53] This is about defending [67:55] the ownership society and capitalism [67:58] that for 250 years created the greatest [68:01] experiment in the history of the world, [68:03] but that's at at risk. Less than half of [68:05] people under the age of 40 have a [68:07] favorable view of capitalism. So, when I [68:09] talked to you about it the first time, [68:10] Michael understood both sides of it. [68:12] It's the right thing to do and it's the [68:14] right thing for the country. And so, at [68:17] any rate, Michael Dell, tremendous [68:19] American. [68:21] I have one just one [68:24] one punch up. [68:26] The name Invest America, Trump accounts. [68:28] What do you think? [68:30] Were you Were you considering this in [68:32] the context of other philanthropy? I [68:34] mean, how do you How do you kind of put [68:35] this [68:36] together in the spectrum of how you [68:37] think about giving back? Yeah, great [68:39] question. So, so, you know, we have a [68:41] foundation that's very focused on [68:44] children in urban poverty. That's [68:47] basically the central [68:48] focus of the foundation. Although, folks [68:51] in central Texas would know that we do a [68:53] few other things here in our local [68:56] community. [68:57] Um and [68:58] >> [snorts] [68:59] >> you know, when I've heard about this [69:00] idea, one of my thoughts was wow, this [69:02] is like a platform for directly giving [69:06] to the people that we're targeting. [69:09] Right? And you know, we actually thought [69:12] about doing it just in Texas first. [69:15] And [69:17] um you know, [69:18] things have gone pretty well with the [69:19] company and all that. So, you know, [69:23] we thought we just go bigger. And and [69:26] what happens Brad if [69:28] you know, 10 more Michael Dells show up? [69:31] And and there are dozens of them. [69:33] >> a lot of Michael Dells, let's be honest, [69:35] but There are there is a number of folks [69:37] who could make an equal size or even [69:39] greater gift. Um there are people who, [69:42] you know, many hands makes for light [69:44] work. There are a thousand people who [69:45] could make a a gift of significance. [69:48] What if this actually becomes a movement [69:50] and we change the dial? [69:51] >> I I think I think it actually is [69:53] becoming a movement instead of a moment [69:56] and and and we've you know, we've got a [69:58] lot of that queued up, Brad. I want you [70:00] >> Wait, have you called anybody, Michael? [70:02] Have you called Did you call Michael [70:03] yourself? My my Michael and I chair the [70:05] Invest America Giving Committee and [70:07] we're ambitious guys. So, you're you're [70:10] knocking on doors. You've you've had a [70:11] few conversations. [70:12] >> people. [70:13] >> Yeah. [70:14] And so so so so so so just There was a [70:16] question earlier. First, it's really [70:17] important to understand and for you guys [70:19] to spread the word, every child under [70:21] the age of 18, [70:22] every child under the age of 18 is [70:24] eligible to claim their account. Number [70:26] one. [70:27] Number two, [70:29] you've heard this like, oh, kids born [70:31] between 25 and 28. No. This is [70:34] forevermore. The legislation creates [70:37] this account forevermore. Every child [70:39] born in America starting January 1st, [70:41] 2027 will automatically get an will get [70:44] a Trump account, right? At birth, [70:46] stapled to their social security card. [70:49] The thousand dollars has to be [70:50] reauthorized every four years. [70:53] Okay, but the accounts don't. So, every [70:55] kid, this is social security 2.0. This [70:58] is the biggest change to the social [71:00] contract in America in in 50 years. 3.7 [71:04] million kids a year will get an account [71:07] that can compound as a 401k from birth. [71:09] And yes, we're going to have a lot of [71:11] announcements, but it's not just [71:14] billionaires. It's going to be companies [71:16] that are donating stock on their IPOs [71:18] into these accounts. It's going to be [71:20] wealthy people. It's going to be states. [71:23] It's going to be moms and dads. It's [71:24] going to be corporations. Normal people. [71:26] >> And the estimate is over 15 years, we [71:29] can move $5 trillion into the [71:32] into the pockets of families that would [71:34] have otherwise had zero. [71:37] $5 trillion. [71:39] Right? And so to me, the leadership that [71:41] Michael showed not only in helping me [71:43] get the meeting that ultimately got this [71:45] passed into law, and it does take [71:47] people. Like those moments either happen [71:50] or they don't happen. And if they don't [71:52] happen, there's no law, and this doesn't [71:55] change kids' lives. By the way, two [71:56] things on this. [71:58] If this $5 trillion [72:00] moved through government programs, it [72:02] would get incinerated. [72:03] >> Exactly. [72:04] That's what we see happen. There's just [72:06] a million crony structures that take it [72:09] away and destroy it. So to give it [72:10] directly into the accounts [72:12] is the circumstance. The second thing is [72:14] it makes a lot of sense that [72:16] you guys can I'll I'll I'll be the lead. [72:19] But can we replace social security in [72:21] this country with a defined benefit or [72:23] defined contribution like this? And [72:25] eventually everyone have a Trump account [72:27] or whatever you call it. And we don't [72:30] have to have this fake Ponzi scheme that [72:32] we call social security. [72:33] >> we can do it. [72:34] >> Well, they they they have a defined [72:35] benefit program, but I'm saying like [72:37] everyone has an account and they all own [72:39] a piece of their future. And every time [72:40] you get a payroll tax deduction, instead [72:42] of it getting eviscerated and destroyed [72:44] and vaporized, [72:46] that money actually goes into an account [72:48] and you buy a piece of a company, and [72:49] maybe you can direct it. [72:50] >> Freeberg's getting angry. [72:51] >> on July 4th of this year. You're getting [72:53] me wound up. [72:54] >> So, for all the There are 4.5 million [72:56] kids who've claimed their account. [72:58] Almost $150,000 a day will have on the [73:01] trajectory we're on. 10 million by July [73:03] 4th or 250 at the anniversary of the [73:06] country. Every one of those kids' [73:08] accounts, the parents and the kids, on [73:10] July 4th, they'll see an app on their [73:12] phone that looks a lot like a Robinhood [73:14] app. They'll see them owning It'll say [73:17] you've received your $1,000 or your [73:19] $250. [73:20] You're And it will show a little bit of [73:22] Nvidia, a little bit of Walmart, a [73:24] little bit of Dell. We decompose the S&P [73:26] 500, which they own, into the [73:28] constituent parts so they can get [73:30] excited about being an owner in the [73:33] upside of America. And when moms and [73:35] dads double click an Apple Pay 5 10 [73:38] bucks into the account, right? When they [73:40] send their QR codes to their friends on [73:42] their birthday, and now their friends [73:44] all add to the account, or on Christmas, [73:46] or bar mitzvahs, and they add to the [73:48] accounts, [73:49] when companies add to the accounts, all [73:51] of this they see it growing, and it [73:53] unlocks the human potential. It's not [73:56] just the money. It's that I'm in the [73:59] game. I have a shot, which [74:02] to David's point, I think the biggest [74:04] crime of Social Security, and we made [74:06] very clear, Social Security is a sacred [74:08] promise. We We refused, and many people [74:11] tried to get us to to to take on the [74:14] broader struggle, and we didn't do it [74:16] because we knew it it would kill this [74:18] program. But But let's be clear about [74:20] this. Our government requires all of us [74:22] to give 10% of what we earn into Social [74:25] Security, right? It was the social [74:28] contract evolution in the Industrial [74:30] Revolution that kept kept the country [74:31] together. The only problem is it goes [74:33] into a black hole. Nobody sees it. [74:36] Nobody knows what's there. But it is [74:38] your savings. Now, imagine if that same [74:40] money was required to, you know, [74:42] government took it away, but it was in [74:44] an account with your name on it. You [74:46] could see it grow. You know exactly what [74:48] was there. You could get excited and [74:49] say, "Hey, I'm going to add a little bit [74:51] more to that." Right? And and and you [74:53] had a little bit of choice. That to me [74:55] is [74:56] is the possibility and I think we will [74:57] end up there. And and and Brad, thank [74:59] you for [75:00] >> Yeah, let's give it up for Brad [75:01] Gerstner. [75:02] >> [applause] [75:04] >> Finally! Yeah, I'm just going to say, [75:07] you know, Brad [applause] Brad also [75:08] adopted his home state of Indiana. We [75:11] have Ray and Barbara Dalio Dalio adopted [75:14] their home state of Connecticut and many [75:17] many more to come. And and and look, [75:19] it's going to be super easy for anybody [75:22] to add 100 kids in your neighborhood, [75:25] adopt a zip code, adopt a school [75:26] district, adopt a town. [75:28] >> It's going to be amazing. Give it up for [75:30] one of the great entrepreneurs of our [75:32] time and an incredible philanthropist. [75:35] >> [applause] [75:35] >> I'M DOING ALL IN. [75:42] >> [music] [75:48] [music] [75:51] >> I'M DOING [75:52] ALL IN.