一位负责打仗的部长,站在一家造火箭的公司里,宣布五角大楼要用马斯克的算法改造自己。
2026 年 2 月,皮特·赫格塞斯(Pete Hegseth)在得州星舰基地(Starbase)发表讲话。台下坐着 SpaceX 的工程师,讲的内容跟火箭没多大关系:设立唯一的技术总管,废掉三个委员会,重组两个创新机构,给数据共享定死期限,并且当场宣布 xAI 的 Grok 本月晚些时候接入军方的 GenAI.mil 平台。
这场讲话信息密度很高,值得拆开看。它同时回答了两个问题:美国国防采购体系烂在哪儿,以及它打算怎么修。
讲话地点本身就带着态度
赫格塞斯开场没有讲政策,先讲感受。
美国没有第二个这样的地方,世界上也没有。你们已经建成和将要建成的东西,是美国创造力与美国发明能力的证明。(皮特·赫格塞斯)
选择在星舰基地宣布五角大楼的创新改革,等于把 SpaceX 当成了对照组:一家没有几十年历史、没有庞大的分包层级、没有一堆协调委员会的公司,做到了传统承包商做不到的事。
这种对比在讲话里反复出现。他说五角大楼过去那套流程,高流动、低进展、几乎没有产出,然后停了一下:
这听起来跟 SpaceX 正好是反的。(皮特·赫格塞斯)
更狠的一句在后面。他提到 SpaceX 和 Palantir 当年为了拿到国防部合同的竞标资格,不得不把国防部告上法庭。
这句话的信息量比听起来更大。一家已经证明自己能回收火箭、能把卫星送上天、能把成本压到对手三分之一的公司,要打官司才能进入采购流程。所谓的"流程繁琐",落到现实里是律师费和诉讼周期,是拿融资去请律师的时间。
三十年前那顿"最后的晚餐"
赫格塞斯把病因追溯到一个具体的饭局上。
冷战结束后,一位前任国防部长在一次对工业界的晚宴演讲中,主张整合国防工业基础。那场演讲后来被叫做"最后的晚餐"(Last Supper)。整合的结果,是美国的国防工业基础收缩成少数几家主承包商主导的封闭生态。
他给出的后果清单很短:成本飞涨、交付迟缓、创新停滞。
这个结果,使得新的技术创新创造者几乎不可能在我们的部门拿到业务。(皮特·赫格塞斯)
这段话背后的机制值得说清楚。主承包商数量减少,意味着采购官员的选项减少。选项减少,意味着任何一次失败的采购都会变成不可承受的丑闻。为了避免丑闻,流程只能不断加码,评审、审计、合规文件层层叠加。流程越重,能承受这套流程的公司越少。能承受的公司越少,主承包商越集中。
这个循环自我强化,转了三十年。赫格塞斯用了一个词形容它的产物:风险厌恶文化。
他把特朗普政府关于国防工业基础和国防公司的行政令拿出来当依据,说法很直接:老牌主承包商的优先级应该是国家安全,不是下一场财报电话会。少一点股票回购,多一点工厂车间里的投入;少一点股东分红,多一点基础设施、厂房和设备。
八步传送带,和三十年的赌注
讲话里有一段对旧体系的技术性描述,写得很清楚。
传统的技术开发模式,假设技术沿着一条可预测的线性传送带移动:实验室、设计、开发、原型、测试鉴定、正式项目、交付。中间还有"八步"的说法,从第一次发现到变成武器,三十年。
这套方法的核心是一种傲慢,它假设你能轻易预测未来,假设你能预见到一项发明如何在八个简单步骤后变成武器,从首次发现算起要花三十年。(皮特·赫格塞斯)
这段话点出的是采购体系最深的一个假设:未来可以被规划。
在冷战年代,这个假设勉强成立。对手明确,技术路线清晰,大规模量产决定胜负。代价是慢,但慢不要紧,因为对手也慢。问题在于,今天的对手不按这条时间线走。无人机、商用卫星、开源模型,这些东西的迭代周期以月计。用三十年周期的体系去应对,等产品交付时,需求本身已经变了。
他补了一句更尖锐的:
我们不能再承受等十年,让老牌主承包商交付下一个完美系统,结果发现它晚了好几年,花了本该花的十倍价钱。(皮特·赫格塞斯)
他引用了马斯克的算法,但只引了一半
讲话中段,赫格塞斯把马斯克那套著名的工程方法论搬了出来,说是"马斯克用他的算法写下了新剧本"。
质疑每一条需求,删掉愚蠢的那些,然后拼命加速。(皮特·赫格塞斯)
这里有个细节值得拿出来说。马斯克原版的算法有五步:让需求变得不那么蠢、删掉能删的零件或流程、简化和优化、缩短循环时间、自动化。赫格塞斯引用的是第一、第二和第四步,跳过了第三步"简化和优化",直接从"删掉"跳到"加速"。
说这个细节,是因为跳过简化直接加速,恰好是这套方法论最常见的失败方式。旧流程删了,新流程没设计,速度又提上去,结果往往是混乱以更快的速度发生。当然,讲话归讲话,真正的执行细节要看后面那些具体的组织动作。
第一条规则:速度
赫格塞斯把新规则归纳成六条,第一条就是速度。
他的判断很干脆:军事 AI 在可预见的未来会是一场竞赛,在这场竞赛里,动得太慢对美国国家安全造成的风险,超过对齐不完美带来的影响。
这句话把价值排序摆到了台面上。过去十年,国防 AI 讨论的重心一直是"能不能用"和"怎么才算合规"。现在优先级倒过来了:先跑起来,边跑边修。
配套的动作很具体。新任首席数字与人工智能官(CDAO)卡梅隆·斯坦利(Cameron Stanley)及其团队,要在 30 天内为所有标杆项目定义"AI 部署速度指标",此后至少每月汇报一次。这些指标将成为整个部门的新基准。
用"部署速度"当核心 KPI,会改变激励结构。过去衡量一个项目,看的是它有没有按流程走完、文件齐不齐。一旦速度变成主指标,那些靠流程复杂度生存的岗位就要重新证明自己的价值。
第二条规则:挡路的可以另谋高就
第二条讲官僚障碍,措辞是全场最不客气的。
如果你想挡路,你可以去别的地方工作。(皮特·赫格塞斯)
他把数据共享的障碍、运行授权(ATO)、测试鉴定、合同签订这几类问题,从"官僚层面的不便"重新定性为作战风险。前者可以慢慢协调,后者必须立刻清除。这个转向让后面所有强硬的机制都有了正当性。
落地手段是一个"障碍清除特设小组",挂在研究与工程副部长权限下,有权豁免非强制性要求,并且可以把任何拖慢 AI 能力推进的事,直接上交给副部长史蒂夫·范伯格(Steve Feinberg)。
后面还有一句更硬的:对于持续阻碍数据访问的情况,可以上报到副部长层级处理,副部长有权在法定范围内重新分配岗位、终止人员职务,或者扣留不配合活动的经费。
在一套以共识和流程为默认操作系统的体系里,这种高压设计很少见。它能不能真跑起来,取决于副部长愿意用多少次这个权限。第一次用是信号,一直不用就是摆设。
第三条规则:算力
算力这条讲得比较简略,方向却很明确:从数据中心一直铺到战术边缘,并且撬动流向美国 AI 产业的数千亿美元私人资本。
具体的政策抓手是特朗普的 AI 行政令:在军事用地上建数据中心,与能源部合作扩大供电和配套资源。合作名单点了 Google、AWS、Oracle、SpaceX、微软。
把数据中心建在军事用地上,绕开的是美国本土数据中心最难的一关:选址、供电、审批。这三件事在民用市场通常要花好几年。军事用地加能源部配合,等于开出了一条快车道。
第四条规则:人
人才这条,赫格塞斯强调要用尽所有招聘和薪酬权限,把最好的技术人才弄进来,并且重奖那些真正推动了 AI 转型的员工。抓手是特朗普政府的"Tech Force"计划,从产业界和学术界引才。
他点了几个已经在政府里的名字:马斯克、大卫·萨克斯、埃米尔·迈克尔。
新任 CDAO 斯坦利的团队构成也值得一提。他提到团队成员来自 AWS、Databricks、Palantir、Meta,很多人放弃了高薪职位来干这件事。
这段的潜台词是:政府不可能在薪酬上跟科技公司竞争,能拿出来的是任务本身的分量。这个逻辑对一部分人有效,对另一部分人无效。它决定了这支队伍的上限。
第五条规则:AI 不许 woke
第五条是全场政治色彩最浓的一段,也是唯一直接动了"责任 AI"定义的。
过去那种"公平 AI"以及其他多元、平等、社会正义掺沙子、束缚和混淆我们使用这项技术的日子,结束了。(皮特·赫格塞斯)
他给"责任 AI"下了一个新定义:客观真实的 AI 能力,安全使用,并且符合管辖本部门活动的法律。
我们不会使用那些不让你打仗的 AI 模型。我们只按这一个标准评判 AI 模型:事实准确、与任务相关、不带限制合法军事应用的思想约束。(皮特·赫格塞斯)
然后是那句被媒体反复引用的:
战争部的 AI 不会是 woke 的。它会为我们干活。我们在造的是随时能打仗的武器和系统,不是给常春藤教工休息室用的聊天机器人。(皮特·赫格塞斯)
把政治修辞剥掉,这段话里立起了一道实质性的采购门槛:模型的军事用途条款,现在成了能不能进五角大楼的硬指标。
过去两年,几家主流模型厂商的使用政策里都有明确的军事用途限制条款。这些条款在商业客户那里几乎不构成障碍,放到国防采购里就是致命问题:一个模型如果不能合法用于目标识别或作战规划,它对军方就没有核心价值。
这也解释了为什么 xAI 能这么快进来。Grok 的使用政策相对宽松,这是产品策略,也是市场策略。
第六条规则:数据囤积算国家安全风险
第六条讲数据,是六条里最有制度设计感的一条。
赫格塞斯的判断:美国军队有二十年的军事和情报行动积累,这个数据优势没有第二支军队能复制。但眼下这些数据大量搁浅在各自的专用项目数据库里,锁在 Title 10 和 Title 50 的烟囱后面,一线的操作员、工程师和产业界根本看不到。
他给出的动作和时间表非常具体:
- CDAO 行使全部权限,执行战争部数据法令,把所有合适的数据在联邦化 IT 系统中开放给 AI 使用,包括各军种和各部门的任务系统
- 各军种部长和部门主管,30 天内向 CDAO 提交现有数据资产目录
- 数据访问请求被拒绝的,7 天内上报给 CTO,并且必须有充分理由
- 情报体系的数据由负责情报与安全的副部长布拉德·汉塞尔(Brad Hansel)同步处理
- 持续阻碍数据访问的,上报副部长,可重新分配岗位、终止职务或扣留经费
最后一句定性:
数据囤积现在是国家安全风险,我们会按这个来对待它。(皮特·赫格塞斯)
这套设计的巧妙之处在于,它没有试图消灭烟囱,而是给烟囱装了倒计时。拒绝共享依然允许,但必须在 7 天内说明理由,理由还要能过 CTO 那一关。把默认状态从"不共享"改成"共享,拒绝需举证",是这类改革里最有效的一招。
一个人管技术
组织层面的动作,是这场讲话最实在的部分。
赫格塞斯宣布,研究与工程副部长埃米尔·迈克尔(Emil Michael)成为整个战争部唯一的首席技术官。他用的措辞带着讽刺:"一个机构一个 CTO,多么新颖的概念。"
这位 CTO 的权限包括:设定技术方向,领导创新生态,每天当面向部长汇报美国在技术和创新竞赛中是领先还是落后,拥有决策权,通过严格评估推动可量化的结果。
与之配套的是三刀砍下去。国防创新指导小组、国防创新工作组、CTO 委员会,全部撤销。取而代之的是一个"CTO 行动小组",职能写得很直白:帮助 CTO 做决定、清除官僚障碍、追究领导责任、快速把新技术送到战士手里。
赫格塞斯对旧委员会的评价,是全场最好的一句吐槽:
我们做的事就是一层一层地加委员会和理事会,它们负责协调,但从来不做决定。(皮特·赫格塞斯)
以及:
我们创造了无数个项目,却没有一个能被问责的负责人。(皮特·赫格塞斯)
"协调但从不决定",这句话值得所有在大组织里待过的人抄下来。委员会的存在往往是为了分摊决策风险,风险分摊掉了,决策也就没了。
两个机构被重新布线
国防创新单元(DIU)和战略能力办公室(SCO)都被重新指定为战争部的外派机构,统一归到 CTO 之下。
DIU 的定位是技术侦察、快速合同和其他共性服务,按商业节奏运转。新主任是欧文·韦斯特(Owen West),3 月上任,海军陆战队出身,有实战经验,也有私人资本运作经验。他此前在战争部领导政府效率部(DOGE)的工作,据赫格塞斯的说法,为部门省下了数百亿美元。
选一个有战斗经历又有投资背景的人管 DIU,指向很清楚。DIU 的核心任务是把商业技术转化成战斗力,这活儿需要同时理解战场和资本。
SCO 继续负责识别和原型化颠覆性应用、现有系统的非常规用法、以及能产生战略效果的近期技术。它保留对副部长的法定直接汇报关系,但在运营上归 CTO 指挥。
赫格塞斯还宣布了一套双通道设计,用来解决初创公司最头疼的问题:找不到门。
任务工程与集成活动(MIA)负责告诉产业界"我们要解决什么问题",DIU 负责帮项目办公室采用产业界已经造好的东西。前一个通道给需求信号,后一个通道给采购路径。
他的原话描述了这个痛点:创始人绕着五角大楼跑圈,找对办公室、找对项目、找对有意愿的赞助人,通常最后被官僚体系磨死。找到了一个地方,地方说不是我们这儿;换一个地方,人家一开始就不想要你;然后继续跑圈。
解决方案是"更快的同意,更快的拒绝"。被吊在一堆没有下文的"也许"里,是创业者最难承受的状态。
六十多个实验室,从"发现"转向"交付"
联邦层面的改革只占三分之一。赫格塞斯给各军种下了同样的指令。
90 天内,各军种部长要向 CTO 汇报本军种的创新计划:怎么整合、精简、重新聚焦各自的实验室、研究力量、实验单位和快速能力办公室,围绕三类成果转。
我们有 60 多个实验室,这是国家资产。但早就该把它们组织起来去交付,而不只是去发现。(皮特·赫格塞斯)
另一个机制叫"创新插入增量"(Innovation Insertion Increment,简称 Triple I)。从 2028 财年预算开始,每一位组合采购执行官都要在自己的项目组合里留出这笔钱,专门用于最后一步:集成、测试,以及把已验证的方案快速塞进已经列装的系统。
这是整场讲话里最实用的一个设计。长期以来,国防创新最缺的经费不在原型阶段,而在最后一公里。原型做出来了,测试通过了,却没有预算科目可以把它装进现役装备,于是原型就停在演示阶段,等下一次预算周期。
Triple I 就是给这一公里修的路。
他还讲了一个具体的人。上个周末,一位叫德雷南·格林(Drenan Green)的陆军上尉发来一封邮件,详细写了他和所在部队打算怎么部署 AI。
创新可以也应该来自任何地方、任何人,来自想法最好的那个人所在的位置。(皮特·赫格塞斯)
他说要把这位上尉请来,听听这些工具怎么在他所在的部队和其他部队自下而上地用起来。这一段放在满篇改革动作里,说的是同一件事的另一种可能:需求从下往上冒,比从上往下压更接近真实。
1.5 万亿,和 30 天的"特朗普时间"
钱的部分,赫格塞斯提到特朗普几天前在社交平台上提出的方案:2027 财年给战争部 1.5 万亿美元预算。他称之为美国历史上最高的一次,是一次世代级投资。
他的逻辑是四件事叠加:资源增加、流程精简、向新进入者开放、拥抱 AI 和前沿防务技术。四件事合起来,让部队更灵活、更致命、更有威慑力。
私营资本这一段,有几个具体数字值得记。
战略资本办公室(OSC)在大卫·洛尔奇(David Lorch)的带领下,过去五个月完成了六笔关键矿产交易,部署超过 45 亿美元的资本承诺,其中近 50 亿来自私营资本伙伴。多数交易在 30 天内完成。
赫格塞斯对 30 天这个数字的评价很有意思:
这在华盛顿算是光速了。(皮特·赫格塞斯)
他还自嘲了一句:这个速度对 SpaceX 来说可能算慢,但在华盛顿已经是极限。
关键矿产这条线的背景是稀土和关键材料的供应链依赖。这套做法里,政府出的钱只是引子,真正的杠杆是用信用和订单把私人资本拉进来。政府不必自己建矿、自己炼厂,只需要把风险的结构改一改,让资本愿意先动。
六个被逐个点名的官僚病灶
赫格塞斯在讲旧体系的问题时,列了六个症状,每一条都能在大组织里找到对应物。
- 委员会和理事会一层层叠加,只协调不决策
- 项目无数,没有一个能被问责的负责人
- 人员流动率高,进展很小,产出很少
- 把创新当成一个要打勾的项目,不是一个要交付的结果
- 组织按阶段和烟囱切分:实验室在这边,快速能力单位在那边,商业拓展在另一个楼甚至另一个海岸,战士在最末端,几乎是事后才想到的
- 为了绕开系统本身的问题,又建了一个"生态"来绕路
最后一条最讽刺,也最普遍。他原话是"我们造了一个旧生态来绕开真正的系统"。任何在大型组织里干过的人,都见过这种套娃:主流程不好用,于是批一个副流程;副流程也不好用,再批一个特批通道。三年后,公司里有五套并行流程,没人说得清哪个是正式的。
他给所有创新项目的验收标准只有三条:改变游戏规则的技术、可规模化的产品、或者全新的作战方式。三条都不占的,要么重新定位,要么取消。
七个标杆项目,每周问一句"你干了什么"
AI 加速战略的抓手是七个"标杆项目",覆盖作战、情报和机关事务三条主线。每个项目一个负责人,激进的时间表,可衡量的结果。
衡量标准是一句话,赫格塞斯直接对着马斯克说的:
埃隆,你这周完成了什么?(皮特·赫格塞斯)
这七个项目不许在真空里跑,必须跟一线作战人员和转化伙伴直接合作,把实时作战反馈纳进来。预期是快速迭代,并且让失败加速学习曲线。
这些项目不做科研演示,也不设治理委员会。它们就是整个部门的执行标准。(皮特·赫格塞斯)
配套的是每季度一次的持续实验:人对人对抗实验室,AI 协同的蜂群,基于智能体的网络防御,分布式指挥控制。
关于失败,他讲了一段值得单独拎出来的话。
我们的部门过去不接受失败,所以我们从不失败,这意味着我们从没学到东西。我们要把这个逻辑翻转过来。(皮特·赫格塞斯)
这是国防采购文化最深的病灶。在一个把任何失败都当成问责事件的组织里,理性的选择是永远不做可能失败的事。于是所有项目都变成低风险的渐进改良,没有一个是真正的突破。
赫格塞斯还有一句关于 AI 应用方式的判断,比喻很准:
把 AI 撒在旧战术上,就像撒数字仙尘,赢不了未来。我们要靠发现全新的作战方式来赢。(皮特·赫格塞斯)
Gemini 已经进了 300 万人,Grok 本月上线
具体落地层面,赫格塞斯给了两个已经发生的事实。
上个月,战争部与 Google 合作上线 GenAI.mil,Gemini 应用在部门内部已经覆盖约 300 万用户。
这场讲话上,他宣布第二家前沿模型公司加入:xAI 的 Grok,本月晚些时候上线。
他的目标是,很快让全世界最强的 AI 模型,跑遍战争部所有非密和涉密网络。
把两个事实放在一起看,能读出一些东西。300 万用户意味着这套系统早已脱离试点阶段,进入了全面铺开。而引入第二家模型供应商,说明这套系统在设计上就要避免单一供应商锁定。对做模型的公司来说,GenAI.mil 正在变成一个必须争取的渠道。
被焊在一起的操作系统
讲话后段,赫格塞斯把散落的机构串成了一条线。
他说这是同一场战争的第三条腿。前两条是需求改革和采购改革:废掉了只看流程的联合能力集成与开发系统(JCIDS),把联合需求监督委员会(JROC)改成给真实联合作战问题排序的机构;建立了 MIA 来跑实验战役;废掉旧的国防采购系统,建立以速度、风险和问责为核心的作战采购系统。
第三条腿就是创新生态。他用了一个词:焊接。
现在的分工是这样的:
- DARPA 交付改变游戏规则的技术创新和战略突袭
- DIU 交付可规模化的产品
- SCO 交付全新的作战方式
- CDAO 和 OSC 提供数据、测试和资本,让整个体系按战时速度运转
这条链路的描述值得完整记下来:问题驱动实验,实验流向原型,原型流向项目执行官,项目执行官推向生产,生产送到战士手里,循环不停。
赫格塞斯把对手的优势逐条列出来做对比:对手没有美国的企业家,没有资本市场,没有二十年军事和情报行动积累的实战数据,没有那些砸了无数钱换来的保密技术,也没有那种拒绝认输的韧性。
然后他补了一句具体的能力展示:一支能在 37 小时内进出德黑兰市中心或加拉加斯市中心而不被发现的军队。
紧接着是转折:如果这些优势被官僚体系闷死,以上统统不作数。
这场讲话真正的看点
把整场讲话拆完,几个判断可以留下。
第一,这次的改革抓的是人。设一个唯一的 CTO,废掉三个委员会,把 DIU 和 SCO 归到一条线下,这些动作的共同点是压缩决策节点。流程改起来容易被反弹消化掉,动人才可能改掉激励。
第二,所有关键动作都带时间戳。30 天定义速度指标,30 天提交数据资产目录,7 天上报数据拒绝,90 天提交军种创新计划,2028 财年开始执行 Triple I。带时间戳的改革才可能被追责,不带时间戳的改革通常活不过一次人事变动。
第三,模型的使用政策成了采购门槛。赫格塞斯那句"我们不会用不让你打仗的模型",把商业公司的 AI 伦理条款变成了国防市场的准入条件。这会倒逼模型厂商做产品分层,也会催生专门面向国防的模型供应商。
第四,钱和通道都摆出来了。1.5 万亿的预算提案、Triple I 的专项科目、MIA 的需求信号通道、DIU 的采购通道、OSC 的资本通道。对一个想进入这个市场的团队来说,路径比以前清晰得多。
至于能不能落地,还要看两件事:副部长范伯格愿意用多少次那个"重新分配岗位、扣留经费"的权限;以及一年之后,还有几个标杆项目在按周汇报进度。改革最终取决于一个动作:第一次有人挡路时,他是不是真的被挪开了。
如果你在这个行业里,可以做的几件事
- 盯住 GenAI.mil 的模型准入清单。Gemini 已经铺到 300 万用户,Grok 本月上线,第三家、第四家会陆续出现,这个渠道正在变成事实上的分发入口
- 如果你是硬件或无人系统团队,从 2028 财年开始,每个项目组合里都会有一笔 Triple I 专项经费,专门用于把已验证方案集成进现役系统。这是过去最缺的一环,现在有了正式科目
- 搞清楚两个通道的分工:MIA 发布"我们要解决什么问题",DIU 负责"采用产业界已经造好的东西"。前者给你需求信号,后者给你采购路径。别再去五角大楼里绕圈找门
- 如果你是模型厂商,把军事用途条款当成产品设计问题来对待。它现在直接决定你能不能进入这个市场
- 如果你在体制内,记住三个数字:拒绝数据共享要在 7 天内上报,数据资产目录 30 天内提交,军种创新计划 90 天内汇报。数据囤积已经被定性为国家安全风险,它构成了问责依据
- 关注 3 月欧文·韦斯特正式接手 DIU 之后的头几个动作。一个既打过仗又管过资本的人来管技术转化,他的第一批合同会说明这套体系真的转了,还是只是换了个说法
- 如果你是创业者或投资人,把"更快的同意、更快的拒绝"当成一个信号:这套体系正在试图降低决策的不确定性。愿意给你明确拒绝的机构,比一直说"再看看"的机构更值得投入时间
0:00Wow, what a tour. What an opportunity to be here at Starbase, Texas with Elon and the Space X team. There's nothing like this in America. There's nothing like this in the world. And what you have built and what you will build here is a testament to the strength of American ingenuity and American invention. So, I want to thank all of you, all the folks out here for having us today. Elon, thank you so much for hosting us, for what you built, for the vision you have for this company, the vision you have for our country, the vision you have for American innovation. I could not think of a more fitting venue to continue our arsenal of freedom tour and to outline
0:43today the future of technological innovation at the War Department. Those of you here at SpaceX will appreciate this, knowing that as World War II was ending, the Secretary of War and Secretary of the Navy wrote to the National Academy of Sciences and declared that scientific research was essential to our national security.
1:05To ensure continued preparedness, they wrote, the research scientists of the United States must be called upon to continue in peace time some substantial portion that of which they have made so effectively during the stress of the present war. The competitive time element in developing those weapons and tactics may be decisive in future conflicts. You see, those secretaries of war and navy many decades ago recognized the importance that innovation and readiness holds for our national security. They knew what was at stake, the very freedoms of the country we hold dear.
1:47All across the United States today, extraordinary innovation is unlocking new possibilities for freedom, prosperity, and security. Now the question before us is not whether or not the most powerful technologies of this century will reinforce free societies. Is it going to reinforce our free societies or will that technology be shaped and twisted by malign regimes that seek to use those technologies for control and coercion? You see, over the past several months, I've talked at length about the challenges we face in transforming the war department to address current and future missions. all in service of meeting the needs of the 21st century warf fighter. I'd like to
2:31think we've already made dramatic progress in the Pentagon's culture by reviving the warrior ethos and we're moving out quickly in transforming our acquisition ecosystem as well. But today is about how we supercharge innovation at the war department for the era ahead. Innovation is happening at a pace we can't even foresee. And we need the entire enterprise, our enterprise to embrace the urgency required for this moment.
3:01Since the end of the Cold War, the defense industrial base in our country has consolidated. This makes it difficult, if not impossible, for new creators of technical innovations to win business at our department. The result is a risk averse culture that prevents us from providing our war fighters with the best resources that America has to offer.
3:27That ends today. Simply put, the United States must win the strategic competition for 21st century technological supremacy. Artificial intelligence, autonomous systems, quantum, hypersonics, and long range drones. If you talk to Elon Musk long enough, he will tell you how important hypersonics and long range drones are, and he's 100% correct. Space capabilities, directed energy, and biotechnology are the new areas of global competition.
4:01The challenge is that our legacy approach to technological development assumes that technology moves in a predictable linear conveyor belt from lab to design to development to prototype to test and qualify to program of record and can only be provided by a handful of companies that have consolidated dramatically.
4:24Now, while this system provided us with the weapons that won the Cold War, it is archaic and inconsistent with the novel threat environment that we face today. At its core, this old approach has the hubris to assume that you can easily predict the future, that you can foresee how an invention becomes a weapon in eight easy steps, three decades from first discovery. Our system cannot keep treating innovation as a decadesl long one-way march that dramatically reduces who and what is able to run the gauntlet at our department to get the capability to the war fighter.
5:06Until this administration, the Trump administration, the department's process for fielding new capabilities had become just one more postcold war peace dividend relic that has not kept up with the times. Worse than that, we've done nothing but add layer upon layer of committees and councils that coordinated but never decided.
5:30We created endless projects with no accountable owners. We have high churn with little progress and few outputs. That sounds about like the exact opposite of SpaceX. We treated innovation as a box to check, not an outcome to deliver. In short, when it comes to our current threat environment, we are playing a dangerous game with potentially fatal consequences.
5:57We need innovation to come from anywhere and evolve with speed and purpose. You see, America's open scientific community is an advantage authoritarian regimes cannot replicate. They can read our scientific papers and copy our invention or distill our ai models, but they cannot replicate a culture like this one of opened and distributed research.
6:23We need to be blunt here. We can no longer afford to wait a decade for our legacy prime contractors to deliver the next perfect system only to find that it's delivered years behind schedule and cost 10 times what it should. Winning requires a new playbook. Elon wrote it with his algorithm.
6:46Question every requirement, delete the dumb ones, and accelerate like hell. That's why I want to make clear across the war department and for our partners in the private sector that our under secretary of war for research and engineering Emil Michael right here in the front row is the war department's single chief technology officer. One CTO
7:10for the entire enterprise. Novel concept. As the sole CTO, Emil will set the technical direction, lead the innovation ecosystem that will welcome progress from anywhere it resides. And he'll tell me face to face every day and frankly whether we are gaining or losing the technology and innovation competition. He'll have the decision authority and will lead through rigorous evaluation with a focus on real measurable outcomes. Now, Congress, and we have member, many members of Congress with us here today, will be our partner in this. By affirming the under secretary's role as the department CTO, we're carrying out the core mission for R& that Congress assigned in statute.
7:55The last defense bill expanded the CTO's authorities, including the power to direct military departments and other components to align around clear, innovative outcomes. We we will use them to deliver those outcomes with speed and urgency. An empowered CTO will inject a disruptive mindset directly into our systems, directing the power of America's worldleading scientists and entrepreneurs, our cutting edge labs, the tech ecosystem, and our capital markets to build what the war fighter needs, but to do so better, faster, and cheaper. And Emil is the right man to do it. You see, this isn't about military structure. This is about building an innovation pipeline that cuts through
8:42the overgrown bureaucratic underbrush and clears away the debris Elon style, preferably with a chainsaw, and to do so at speed and urgency that meets the moment. As I've said repeatedly to every audience, the president of the United States and I have the backs of our war fighters who have to make split-second life and death decisions on the battlefield.
9:10And I want this audience to know that we also have the backs of innovators who share that very same urgency. American taxpayers, especially those parents whose sons and daughters have answered the call to serve. I swore in 10 more today. Demand and I demand and we demand that we arm our war fighters with overwhelming and lethal technology right now, not a decade from now. In modern warfare, the fastest innovator and iterator will be the winner. And no one can out innovate an American entrepreneur who has been liberated from the constraints of stifling bureaucracy.
9:55That old era ends today. We are done running a peacetime science fair while our potential adversaries are running a wartime arms race. Right now, from garage startups to factory floors to boardrooms across our nation, the question of how to harness innovation coming out of America's AI ecosystem is front and center. and rightfully so because it has the potential to disrupt and transform every area of human endeavor. The same is true
10:31in the Department of War. President Trump's AI executive order spells out our approach succinctly. It is the policy of the United States to sustain and enhance America's global AI dominance in defense of human flourishing, economic competitiveness, and national security. That, ladies and gentlemen, is the War Department's mission. We must ensure that America's military AI dominance so that no adversary can exploit that same technology to hold our national security interests or our citizens at risk.
11:08America first in every domain. Last month, I took the first step toward changing how the department does business with Frontier AI technologies when we announced the rollout of Gen AI with our partners from Google. And I want to thank the Google team for leaning forward and making the investment to get their Gemini app to about 3 million users in the war department.
11:33But today, we're excited to announce the next frontier AI model company to join Genai. Mill and that is Grock from X AI which will go live later this month. So I want to thank you Elon and your incredible team for leaning forward with us on this as well. Very soon we will have the world's leading AI models on every unclassified and classified network throughout our department. Long overdue.
12:04To further that, today at my direction, we're executing an AI acceleration strategy that will extend our lead in military AI established during President Trump's first term. This strategy will unleash experimentation, eliminate bureaucratic barriers, focus on investments, and demonstrate the execution approach needed to ensure we lead in military AI and that it grows more dominant into the future. In short, we will win this race by becoming an AI first warf fighting force across all domains from the back offices of the Pentagon to the tactical edge on the front lines.
12:43The catalyst for this acceleration will be seven pace setting projects focused on mission threads across war fighting, intelligence, and enterprise missions. each with a single accountable leader, aggressive timelines, and measurable outcomes that answer a familiar question. Elon, what have you accomplished this week?
13:07This is the execution standard for AI first transformation. Each of the seven pace setting projects will use the following model. One owner who reports monthly on their progress. These projects will not be run in a vacuum, but will work directly with war fighters and transition partners to ensure we incorporate real time operational feedback. We expect rapid iterations with failure accelerating the learning curve. These are not science projects. They are not governance boards. They are the execution standard for the entire department.
13:45We will not win the future by sprinkling AI onto old tactics like digital pixie dust. We will win by discovering entirely new ways of fighting. That's why we will run continuous experimentation campaigns, quarterly for forceonforce combat labs with AI coordinated swarms, agent-based cyber defense, and distributed command and control. pushing the envelope, learning from failure at every stop, which is exactly what this place does. You see, our department doesn't accept failure in the past, and so we never fail, which means we never learn. We're flipping that dynamic.
14:27Before talking about the new rules of the game, let me talk about a little bit more about our new team. Because no game can be won without the right team. And we're proud to announce that Mr. Cameron Stanley has been appointed the new chief digital and artificial intelligence officer, CDAO, of our war department.
14:44Cam will be leading a new team, many of whom have foregone, thank you, or left lucrative careers at pioneer companies such as AWS, Data Bricks, Palanteer, and Meta to join the fight. This team will not only provide a catalyst for change in this department, but will also act, we believe, as a magnet for other talented members of the tech community who want to join us in doing the mission focused work to protect our great republic.
15:14So, let's talk about the new rules. First, speed. Speed wins. Speed dominates. Our enterprise currently operates on staffing and committee cycles measured in months and years, and that's unacceptable. Military AI is going to be a race for the foreseeable future, where the risks to US national security of moving too slowly outweigh the impacts of imperfect
15:42alignment. To do this, Cam and his team at TD at TDAO will define AI deployment velocity metrics for all the pace setting projects in the next 30 days and report at least monthly after that. These will become the new benchmarks for programs across the department. Second, bureaucratic blockers. If you work with Elon, he you know he finds the blockers and you remove them. We will take a wartime approach to people and policies that block this progress.
16:18You want to block, you can work somewhere else. Barriers to data sharing, authority to operate or atto test and evaluation and contracting are now treated as operational risks, not simply bureaucratic inconveniences. We are blowing up these barriers. That's why today at my direction, I'm establishing a barrier removal SWAT team under R& authority to wave non-stutory requirements and escalate to our great deputy secretary Steve Fineberg anything that slows down the acceleration of AI capabilities.
16:57Third, compute resource. We will invest heavily in expanding our access to AI compute from data centers to the tactical edge and will tap into hundreds of billions of dollars in private capital flowing into American AI. President Trump's executive order has directed us to build data centers on military land and to work with the Department of Energy to ensure that we dramatically increase the number and breadth of resources needed to power this computing infrastructure.
17:28We will work together with our partners at Google and AWS and Oracle and SpaceX, Microsoft and others on these initiatives. Fourth pillar, especially in this room, you'll understand it, is talent. We will use every hiring and pay authority available to us to bring the best American technical talent and reward effective AI transformations by our workforce. We're going to heavily leverage President Trump's Tech Force initiative to bring in the best and brightest from industry and academia.
18:02With people like Elon, David Saxs, Emil, Mike, and others from the entrepreneurial and business world already in government, we have shown that we can and that we must enlist the world's leading talent in this cause. Fifth, responsible AI. Today I want to clarify what responsible AI means at the Department of War. Gone are the days of equitable AI and other DEI and social justice infusions that constrain and confuse our employment of this technology.
18:41Effective immediately responsible AI at the War Department means objectively truthful AI capabilities employed securely and within the laws governing the activities of the department. We will not employ AI models that won't allow you to fight wars. We will judge AI models on this standard alone. Factually accurate, mission relevant, without ideological constraints that limit lawful military applications.
19:11Department of War AI will not be woke. It will work for us. We're building war ready weapons and systems, not chat bots for an Ivy League faculty lounge. Sixth, and finally, data. AI is only as good as the data that feeds it. And the US military has an asymmetric data advantage from two decades of military and intelligence operations that no other military in the world can replicate.
19:41But right now, we are underutilizing this advantage. Too much of our data is stranded. It's stuck in bespoke program databases, locked behind Title 10 or Title 50 stove pipes, invisible to operators, engineers, and industry who could help us exploit it with winning speed and scale. And that's why today at my direction, CDAO will exercise its full authority to enforce the DOW data decrees and make all appropriate data available across federated IT systems for AI exploitation, including mission systems across every service and component.
20:22Each service secretary and component head will submit cataloges of their current data assets to the CDAO within 30 days. Denials of data access requests will be reported to the CTO within 7 days and they better have a good justification. Today, I'm also directing the under secretary for intelligence and security, Brad Hansel, to ensure appropriate data from across our intelligence enterprise receives the same treatment and can be fully leveraged to warfighting capability development and operational advantage. AI is only as good as the data that it receives. and we're going to make sure that it's there.
21:02Persistent barriers to data access will be escalated to the deputy secretary of war for resolution with authority to reassign or terminate personnel or withhold funding from non-compliant activities within the statutory limits. We'll be clear here, as I said, data hoarding is now a national security risk and we will treat it that way.
21:27AI is an important part of the future. But here's another truth that we've ignored for too long. Beyond AI, we've treated every other kind of innovation as if they're the same. As you know, they're not. We need to break down unnecessary barriers to rapid technological development, adoption, and transition. Some of you will remember this. A generation ago, one of my predecessors in a dinner speech to industry, now infamously known as the Last Supper, advocated for the consolidation of our defense industrial base. This consol consolidation created a closed innovation ecosystem dominated by just a handful of prime contractors.
22:13The results have been characterized by soaring costs, sluggish delivery, and stagnant innovation. That's what President Trump's recent executive order on the defense industrial base and defense companies seeks to address. It makes crystal clear that the priority of the legacy prime contractors must be our nation's national security, not the next earnings call. That means less focus on stock buybacks and more investment on the men and women on the factory floor. It means less stockholder dividends and more investment in infrastructure, plant and equipment.
22:57Today, that old era comes to an end. The Department of War is reopening to the disruptive energy and agile creativity of our nation's tech startups funded by our world's leading capital markets. I'm directing my chief technology officer to lead this charge and his wingman as always in this effort under secretary for acquisitions and sustainment Mike Duffy.
23:24For too long we organized our ecosystem around stages and silos labs over here so-called rapid units over there commercial outreach in a different building or on another coast altogether and war fighters somewhere at the end almost an afterthought. The result is duplication, drift and confusion. And like the acquisition process, we are already fixing the creation of organizations to work around the problems in the innovation ecosystem rather than taking the bold steps needed to transform it. We created an old ecosystem to get around the actual system.
24:05No more. Every dollar of innovation, whether it may be in a lab or a startup or a classified shop, must exist to deliver one of three things. Gamechanging technology, scalable products, or new ways of fighting. If it's not doing one of those three things at speed, it will be realigned or it will go away.
24:29And that's why I'm entrusting Emil as the CTO to ensure that this directive across the department is rapidly carried out. Emil, you're going to be busy. Again, what we're talking about today is a transformation in the way we think about innovation. In requirements reform led by Mike Duffy, we killed an old model, a sclerotic model, and rewired the department so that problems, money, and experimentation live in one system.
25:01In acquisitions reform, we killed the defense acquisition system and created accountable portfolio acquisition executives, making speed to delivery, speed to delivery our organizing principle. We're going to do the same for technological innovation. That's why today at my direction, we are ending the alphabet soup of councils that meet and brief and write memos and schedule meetings but never decide and rarely if ever accelerate outcomes effective immediately. The defense innovation steering group, the defense innovation working group and the CTO council are disestablished and abolished.
25:44In their place, the CTO will convene a CTO action group to assist him in making decisions, clearing bureaucratic blockers, holding leaders accountable, and most importantly, quickly delivering new technologies to our war fighters. Every organization in this ecosystem must earn its place by delivering warfighting advantages faster than our adversaries can adapt. No sacred cows,
26:14no exceptions. To back up the CTO, today at my direction, we are realigning two pillars of the War Department's Innovation Ecosystem. First, the Defense Innovation Unit or DIU. Since its establishment 10 years ago, DIU has lived through shifting reporting structures and uneven administrative support, doing great things. But at times its portfolio overlap with other parts of the department. Effective today, DIU is is designated a Department of War field activity, providing exceptional tech scouting, rapid contracting, and other common services to the department executed at commercial tempo.
27:00The CTO will provide support to DIU for administrative and resource matters and ensure that the unit's efforts fit into the department's innovation priorities. The the director of DIU will also continue to report directly to me as a principal staff assistant and carry out its statutory duties.
27:20And I'm appointing Mr. Owen West, executive of my drone dominance initiative, as the director of DIU starting in March when the FY27 budget cycle firms up. As a Marine with lots of combat experience, Owen will bring a war fighters mentality to DIU's core mission of transitioning technology to our troops.
27:44Owen also has the private capital experience needed to ensure DIU remains working handin glove with the venture and investor communities and continues to on-ramp new entrance into the war department. Owen led Doge at DOW, the most effective Doge effort across the administration, saving tens of billions of dollars for our department.
28:07And now he will lead the O.Lading Worldleading defense tech startups have attracted billions of dollars in capital. They're reshaping warfare through the proliferation of high-tech, lowcost technologies. DIU's mission is to accelerate the adoption of this commercial technology to help convert entrepreneurial products into tangible combat power.
28:31Second, the strategic capabilities office or SCO is also being designated a department of war field activity aligned under the CTO. SCO will maintain focus on its core mission of identifying and prototyping disruptive applications of new systems, the unconventional uses of existing systems, and near-term technologies to create strategic effects.
28:57SCO will continue to maintain its statutory direct reporting relationship to the deputy secretary, but will be operationally realigned under the CTO to eliminate duplication and ensure the relentless daily focus on delivering near and medium-term capabilities to our war fighters. Relentless urgency and focus is our focus.
29:24Today's defense innovation ecosystem is too fragmented resulted in insufficient technology transition to the war fighter. We address some of this in the transformation we are making in the acquisitions ecosystem. But it also needs to enter the innovation ecosystem. For too long and I know a lot of you have experienced this and others we met with in Los Angeles recently. The experience of founders and entrepreneurs has been running endless laps around the Pentagon looking for the right office, the right program, and the right motivated sponsor. And too often, new entrance are ultimately stymied by the bureaucracy. We hear it time and time again. They don't know where to go, and
30:05then when they go to that place, it's not the right place to go. Then they go somewhere else that didn't want them in the first place, and the lap continues. For example, SpaceX and Palanteer had to sue the Department of War just to get a shot at competing for department contracts.
30:23The bottom line is that new entrance need both a shot on goal, but also faster yeses and faster nos from the department rather than being strung along with a neverending stream of rudderless may. At the Secretary of War level, we will replace the existing maze into two clear channels.
30:47The mission engineering and integration activity MIA will tell industry what problems we're trying to solve and DIU will help program offices adopt what industry has already built. This will help get to faster yeses or faster nos. Clear guidance, clear guide rails, clear demand signal, which is what industry and capital expects.
31:14Now, this isn't just an office of the secretary reform. The services need to transform their innovation ecosystems as well. The Army, the Navy, the Space Force, the Marine Corps, and the Air Force. That's why today at my direction, our military services will take the following actions. Within 90 days, the secretaries of the military departments will brief the CTO on service innovation plans, how they will consolidate, streamline, and refocus their labs, research expertise, experimental units, and rapid capability offices around three innovation outcomes.
31:50Our 60 plus labs are a national asset, but it's past time they were organized to deliver and not just to discover. And beginning with the fiscal year 2028 budget, every portfolio acquisition executive will fund an innovation insertion increment, triple I, dedicated money for the last mile, integration test and rapid insertion of validation solutions into fielded systems.
32:19Innovation cannot be centralized and it should not be. In fact, hope hope it's okay I name him, but just this past weekend, I received an email, a proposal from an army captain named Drenan Green, who I've known for a while. He had a detailed plan about how he and his unit wants to deploy AI.
32:40Innovation can and should come from anywhere and anyone wherever those best ideas reside. We're going to bring that good captain in and hear from him. How can we apply those tools in his unit and other units bottom up, not just top down? The military services and program offices own the last mile. And to that end, I'm directing the services to establish that triple I innovation insertion increment within the budgets of all program portfolios.
33:14This will ensure that they have funds set aside to quickly integrate innovations during weapon system development. This will be a tool to free new innovative cutting edge weapons and improvements from the constraints of yearly budget cycles. Here you iterate in terms of hours and days, maybe weeks.
33:36In Washington, we talk in terms of months, years, and oftentimes multi-year. It's too slow. The CTO ecosystem will give them the supply. Services are contributors to this ecosystem, not bystandards. They must deliver outcomes and they must deliver overmatch. And speaking of budgets and budget cycles, you may have seen President Trump's truth post of a few days ago.
34:03He's proposing a $1.5 trillion budget for the War Department in fiscal year 27. This is by far the most ever in our history a historic and generational investment in American security. We will not squander this once-in-a-lifetime opportunity to rebuild our military. Think of what this means as our efforts to transform the department come fully online. You add resources and you streamline process speed to deliver those capabilities to the war fighter. an invitation to new entrance into the defense industry, which we need so badly, and an embrace of AI and other cutting edge defense tech. All of this will make our forces more agile, more lethal, and more ready
34:53to deter, and if necessary, win a future fight. This is what President Trump demands, and this is what we will deliver. This feeds into a next area we need to fix. You see, reorganizing and a new attitude are not enough to truly unlock innovation for the war fighter. Our department needs to capitalize on a key advantage America has over her adversaries, the broadest and deepest capital markets, and the best entrepreneurial talent in the world. You know this capital is the lifeblood of American innovation. And therefore to be successful we need to embrace the role in our department of private capital which we have not done for far too long.
35:40We need to be a better partner for private capital so we can help accelerate capital formation in key areas and lower the risk for the department. Private capital is already helping to solve a major problem that President Trump has directed us to confront and solve across this administration.
35:57ending our reliance on competitors for access to rare earth and critical minerals. The Office of Strategic Capital or OSC under the leadership of David Lorch is spearheading those efforts and delivering as we speak. In the past 5 months alone, the OSC team has deployed over $4.5 billion in capital commitments as part of closing six critical mineral deals, all of which will help free the United States from market manipulation.
36:30Crucially, OSC's strategy relies on crowding in private sector investors from this VC community to our country's largest financial institutions. The six OSE deals to date include nearly 5 billion from our private capital partners. OSE has closed most of those deals in less than 30 days, which is Trump time for complex transactions.
36:56That may seem maybe slow for SpaceX in Washington. That's as fast as it gets. I mean, that is warp speed in Washington. Our all of our investments serve one purpose. deliver faster for the war fighter. Now, some will want to cynically posit, is this just another innovation reorganization, another set of strategies that brief well but don't actualize?
37:24No, this is the third leg of a single war plan in fundamentally transforming our acquisition ecosystem. We killed JIDS that focused only on process and turned the J-Rock into a body that ranks realworld joint operational problems, not paper requirements. We created the MIA to run experiment experimentation campaigns to solve these problems. We killed the old defense acquisition system and created the warfighting acquisition system to focus on speed, risk, and accountability.
38:00With the AI strategy and innovation ecosystem transformation that we have just outlined, we are welding that third piece into place. The CTO, DIU, SCO, DARPA, CDAO, and OSC are no longer a loose federation. They are the office of the secretary of wars innovation operating system. DARPA delivers gamechanging technology innovation and strategic surprise. DIU delivers scalable products. SCO delivers new ways of fighting. And CDAO and OSC provide the data, test, and capital to move at wartime speed.
38:40As of today, they are an ecosystem, an arsenal of ideas and action for the arsenal of freedom wired directly into requirements, portfolios, and production. Gamechanging technology, scalable products, new ways of fighting. All three moving at wartime speed. Problems drive experimentation.
39:05Experimentation flows to prototypes. Prototypes flow to our program executives. Program exe flow these to production. Production flows to the war fighter. And the cycle never stops. Always iterating. One system, one purpose, speed to the fight. We are preparing to win the future because we know and you know, for the free world, for the west, the stakes could not be higher.
39:38Before this administration, our adversaries may have thought they finally broke American power. They're wrong. They do not have our entrepreneurs. They do not have our capital markets. They do not have our combat proven operational data from two decades of military and intelligence operations. They do not have our hard one classified technologies. They do not have the ingenuity and tenacity of American war fighters who refuse to lose.
40:11They don't have a military that can go 37 hours to downtown Tyrron or downtown Caracus without being seen in the process. You see, but none of that matters, however, if we suffocate those advantages under a stifling bureaucracy. That's why we're unifying the innovation ecosystem, making this an AI first department and holding every lab, every program accountable, all while pushing every officer to deliver warfighting advantages faster than others can adapt.
40:48We will not stop. We will not back down. We will forge the new arsenal of freedom with our partners in industry and the private sector. We believe the future will be shaped by those who lead in technology and innovation. We don't believe we know those who fervently believe in freedom and the western tradition like we do must be those leaders. If not us, if not America, if
41:18not the west, then who? And if not now, it will be too late to maintain that advantage. This is not reform for the sake of reform. It never has been. This is about whether our warriors fight with yesterday's tools or they fight overmatching our adversaries using tomorrow's technologies. We know the threat. We know the opportunity. We know what must be done. We share the urgency.
41:48Now, we will do it and we must do it at wartime speed. Thank you. God bless you. God bless this company that you've built and may God bless our great republic. Thank you very much. Appreciate it. Thank you. [applause] Thank you.
资源与出处
- 视频在 YouTube 收看
- 档案silicon-valley.tech 档案页本页视频与文稿数据来自 Elon Musk Archive
- 精读本篇精读 Markdown 版