第 297 期·中文精读

黄仁勋卸任英伟达 CEO:间谍丑闻、出口管制与市场冲击

Jensen OUT as Nvidia CEO: "Spy" Scandal, Export Restrictions, Market Impact

黄仁勋在泄密与间谍指控后卸任英伟达CEO,讨论AI交易、对华芯片出口管制及Anthropic启动IPO路演的影响。

官方音频00:00 / --:--

本期章节点时间可跳到音频对应位置

  1. 0:00开场寒暄
  2. 1:20黄仁勋卸任 CEO:间谍指控、机密备忘录外泄,谁想让黄仁勋下台?
  3. 31:22市场影响:AI 交易、“英伟达银行”何去何从、中国的策略
  4. 44:19出口管制、极权 AI 风险、对华鹰派
  5. 58:26英伟达下一任 CEO 该是谁?
  6. 1:06:25Anthropic 启动 IPO 路演
中文精读由 AI 基于原始节目改写,非官方翻译,可能存在偏差或错误;观点归属与表述以原音频与官方文稿为准。说明

英伟达的CEO换人了。

黄仁勋在执掌这家公司33年之后辞职。消息落地那一周,英伟达股价跌了15%。它现在是全球市值最高的公司,15%换算成美元是几千亿。

导火索是一份被泄露的机密情报备忘录。纽约时报报道说,一位与黄仁勋关系密切的中国女商人,是中共间谍。随后又爆出他本人有婚外情,董事会启动调查,他在压力下同意离开。

这期All-In,四个人把这件事拆了将近一个小时。萨克斯(David Sacks)把它分成两个独立的“行动”,查马斯(Chamath Palihapitiya)逐个排除了“谁受益”,弗里德伯格(David Friedberg)则从董事会的季度认证表讲到了AI权力结构的重塑。

六天时间线,从一篇报道到一封辞职信

杰森(Jason Calacanis)先复盘了时间线。11月6日,中期选举刚结束,纽约时报发了一篇报道,主角是一位叫Alice Wang的中国女商人。她在新加坡公司Mega Speed做高管,这家公司在亚洲建数据中心,向包括阿里巴巴在内的公司出租算力。报道的依据是一份泄露的情报备忘录,备忘录认定她是中国间谍。

按The Information的说法,她和黄仁勋的关系,帮助推动了2023年以来约75亿美元的芯片流向中国公司。两人同姓,但没有亲属关系。

11月8日,黄仁勋发全员邮件,承认有婚外情,董事会将展开调查。纽约时报随后报道他和一位直接下属有第二段关系,还有关于第三段的传闻。

杰森提醒了一个容易被忽略的背景:这个故事在明面上已经藏了一年多。纽约时报2025年10月就提过Alice和黄仁勋的关系,那是13个月前。上个月还发过一篇跟进,标题是“英伟达CEO被视为中国间谍行动的目标”。

一家叫Culper Research的对冲基金,六个月前就做空英伟达,在推特上发了一长串论证,核心是英伟达低报了自身对中国芯片销售的依赖,推文里直接点了Mega Speed的名。杰森说这对白宫是灾难,有报道说官员们在互相甩锅,想搞清楚为什么Mega Speed没被列入出口管制清单。

萨克斯:这里有两个行动,不是一个

萨克斯开口先定性这件事的代价。对黄仁勋和他的家庭是悲剧,对英伟达和员工是悲剧,对美国也是悲剧。理由很具体:美国在AI加速芯片上的优势,很大程度上是黄仁勋造出来的。华盛顿当年有很多人想把这种芯片管起来,恰恰因为它太好用。他在所有人都做CPU的时候押注GPU,公司几次差点倒闭。

然后他给出框架。

我认为这里显然有两个行动。一个是Alice Wang在做什么,有没有中共参与,可能有走私,我们不知道,正在调查,应该有调查。另一个行动始于某个人或某些人,把一份高度机密的备忘录或其内容泄露给了纽约时报。(萨克斯)

他强调第二件事是犯罪。泄露机密情报是犯罪,很多人因此被起诉。他举了两个前总统的例子:特朗普的海湖庄园被搜查,拜登的车库被搜查,彭斯的文件也被发现。

我不相信纽约时报是在酒吧里听人聊天拿到这个消息的。有人做了计算,愿意冒着坐牢的风险,把这件事告诉纽约时报。(萨克斯)

于是他列了四个问题。

第一个,谁泄的,为什么。动机里至少有一部分是想把黄仁勋拉下马,否则不会违法。既然其他机密泄露都有调查,这个也该有。

第二个,为什么没做防御性简报。如果国家安全部门担心黄仁勋手里有机密、可能泄给一个疑似间谍,标准操作是马上告诉他,别跟这个人来往。萨克斯拿埃里克·斯瓦尔韦尔(Eric Swalwell)做对照:这位国会议员、众议院情报委员会成员,曾与一位中国间谍有私人关系,民主党给他做了防御性简报,他切断联系,继续当议员,甚至没退出情报委员会。

所以在这个非常高调的例子裡,人们不认为有必要辞职。他们认为必要的是让当事人知道,让他切断联系,然后继续往前走。这件事的处理方式完全不同。(萨克斯)

第三个问题,黄仁勋到底知道什么国家机密。他不为政府工作,按理接触不到机密材料。萨克斯说,硅谷科技公司对产品路线图很开放,中共当然想要IP,但那种IP不会在枕边话里传递。真要偷下一代芯片的设计文件,派一个产线工程师进英伟达拷U盘,或者直接网络攻击,合理得多。

他顺手反驳了一个流行说法:英伟达的产品受出口管制,所以就是国家安全产品,所以就有国家安全机密。他说这是“非常宽泛、误导性的挥手带过”。作为美国人,他可以去买一颗Blackwell芯片,没人拦着。买到手放在自己的机架上,他并不拥有国家机密。拿到晶圆也逆向不出芯片设计,你需要的是文件。

第四个问题,有什么证据证明Alice Wang真是为国家安全部工作的间谍。就在录制当天早上,Politico发了一篇文章,即将上任的众议院情报委员会主席、一位民主党人,刚从相关机密简报会出来,在走廊里说他不确定这个人是中国间谍,可能只是个“有用的白痴”。

什么叫有用的白痴?就是不为中国政府工作,不是间谍,只是可能被利用了。(萨克斯)

萨克斯的结论是,现在问题比答案多。

让董事会动手的,可能是一张季度认证表

弗里德伯格问了一个更实际的问题:董事会的信托责任到底是什么。

如果黄仁勋犯了罪,搞了间谍活动,他会被逮捕,董事会必须处理。但显然没有逮捕,这条可以排除。剩下的最可能情况,是他对某件事做了不实陈述。

公司都有行为准则。准则里写着,你必须报告某些关系,报告与下属的关系。你不能和客户、合作伙伴、关键业务伙伴有未报告的个人利益冲突。上市公司的执行官或第十六条高管,每个季度要做认证,以可审计的方式声明自己没有违反行为准则。(弗里德伯格)

他猜,正是这份认证可能让黄仁勋陷入麻烦。他签了字,说自己没有违反任何行为准则和商业实践,其中就包括不和客户或合作伙伴有冲突性的私人关系。

查马斯接着把董事会的逻辑讲得更细。每季度和审计师通电话,审计师会逐条问:你有没有和某人有冲突,有没有和某人有关系。这是SEC的要求,目的是确保公司是股东资本的可靠管家。再往下一层,这还关系到董责险,那是给董事和高管的责任兜底,附带一系列契约条款。

当你可能对某件事做了不实陈述,开始违反契约,真正倒下的就是这些多米诺骨牌。董事会讨论可能没有外界想象的那么狗血,它极其技术化。跟黄仁勋这个人有多伟大相比,这恰恰是更悲剧的地方。(查马斯)

谁受益?查马斯把中国和AMD都排除了

杰森问查马斯:假设黄仁勋不是叛徒,谁从他的离开中获益?

查马斯先花了很长时间肯定黄仁勋,说他是这个行业的建造者,一二十年后当疾病和癌症有解、我们得到承诺的富足时,他的贡献不比任何人小。然后他开始逐个排除。

中国?假设两种情况。

情况一,没有芯片流向中国。那中国赢在哪?之前没有,现在也没有,中国还是卡在原地,需要一个能替代英伟达算力的本土方案。英伟达花了几十年、几十万人时建起的护城河,不会因为CEO走了就消失。中国也许能追上,但那是多年甚至几十年的事。

情况二,芯片通过某种渠道流到了中国人手里。那政府会查清楚每颗芯片的去向,之后芯片就流不过去了。中国原来的进展反而被拿走。

两种情况里,中国要么原地踏步,要么更差。所以我不认为他们受益。(查马斯)

竞争对手?AMD、谷歌、Cerebras、Etched,都不太可能在短期获益。他们需要时间让芯片在规模上跑通,需要台积电、三星给产能,给HBM产能。跑一次百万GPU的训练,不是靠意志力就能成的,那是无数小时的工作。黄仁勋倒下,不会让别家的芯片变好。

唯一可能获益的,是那些想要某种集中化算力治理架构的人。查马斯讲得很具体:现在出口芯片,要求出口方有遥测,要知道芯片何时何地被使用,要有远程终止开关。但这是私营部门的责任,亚马逊、微软、谷歌各自承担。如果这次事件让人觉得监管不够、芯片在漏,某些政府中人就拿到了一个开口,去推动由政府掌握遥测和开关的立法。

这是唯一一个我认为有明确优势的狭窄判断。其他所有人,该做的功课一点没少。(查马斯)

萨克斯补了一句更直接的:整体上中国受益,因为黄仁勋是整个AI生态的支柱。但他也说,泄露不可能是中国干的,因为纽约时报的消息源是被简报过的人,要么在政府里,要么离政府一步之遥。近因是那份机密备忘录,是内部人动的手。

英伟达银行停摆,才是真正的经济风险

弗里德伯格问了一个被低估的问题:英伟达现在不只是在卖芯片。

英伟达在给初创公司、给数据中心、给很多基础设施提供资本和担保。其中很大一部分是黄仁勋直接主导的。如果他不在了,英伟达里还有人继续推动这件事吗?那可能是几千亿美元流向AI经济的资本,说停就停。(弗里德伯格)

查马斯承认这个担忧。他说Colette Kress(英伟达CFO,也管并购和企业发展)和团队都很聪明,但“英伟达银行”确实是绕着黄仁勋转的。现在的问题不是会不会停,是停多久。无限期关闭,还是关一段时间再重启?新CEO敢不敢像他那样激进地投资?

黄仁勋对整个行业有更整体的看法,他知道怎么在棋盘上移动棋子,他有道德权威去做投资和收购。换一个更传统的CEO,要委员会、要分析、要尽调,最好的情况也是变慢。(查马斯)

弗里德伯格补了一句:他凭直觉就知道市场往哪走,认识那些人,他亲身经历过,所以有掌控力。

查马斯把这个和宏观连起来。他给了一个数字:现在要严肃考虑50到150个基点的增长损失。民主党中期选举大胜,本来就会带来一整套放缓数据中心、放缓AI的立法逆风。特朗普经济在中期选举前积累的增长,本来就要减速。再叠加上英伟达银行停摆,以及英伟达作为所有人围绕旋转的锚定资产受到的冲击,增长下行会被加速。

从纯技术角度看,我看不出中国怎么从中获益。所有情形里,大家都是输家。(查马斯)

中国根本不想要那些芯片

萨克斯讲了一段很多人不知道的出口管制史。2022年拜登政府首次对先进半导体设出口管制,划了一条线:领先芯片禁售,落后芯片(比如H20)可以无限量、无需许可证卖到中国。特朗普政府基本延续这个框架,但把落后芯片改成限量、需许可证。

他讲这段是为了反驳一种叙事:有人说特朗普政府对华软弱。他说民主党领袖正想借黄仁勋这件事,把“对华不够强硬”的帽子扣过来。他反击说,这些政客当年大多支持对华接触政策,正是那套政策把大量工作和产业转移到了中国,是特朗普第一任期才开始转向。

然后他抛出一个更反直觉的事实。

整件事最让人意外的地方是,中国根本不想要这些芯片。他们对H20下了禁令。H200获批之后,他们基本阻止了这些芯片入境,还宣布反对英伟达的其他型号。所以中国根本不想要这些芯片。(萨克斯)

他解释,中国鹰派的政策和中共的政策,在这一点上是一致的:都不想让美国芯片进入中国。理由不同。鹰派认为美国芯片进入中国会帮中国。中共想把英伟达挡在外面,扶持华为这个国家队,逼中国市场倒向华为,让华为先规模起来,再去全球竞争。

所以人们以为中共在求着要这些芯片,这很疯狂。中国公司想要,中国开发者想要。但中国政府根据所有公开声明和政策,是在把这些美国芯片挡在外面。(萨克斯)

他提到最新的DeepSeek模型完全用华为芯片训练。杰森补充,两件事可以同时为真:做LLM的公司想要最新芯片,中国政府想要自己的芯片冠军。萨克斯的结论是,如果黄仁勋是因为想卖芯片给中国而下台,那是最荒谬的山头。中国鹰派不想卖,中国政府也不想买。

出口管制的账本,和一场反数据中心运动

萨克斯还讲了一个更少人注意的线索。马里兰州一个废弃购物中心原计划改建数据中心,当地一位活动人士把它搅黄了。她把这场运动的病毒式传播归功于PSL,社会主义与解放党。萨克斯说这是中共代理组织在美国运作,政治臂膀属于常驻上海的内维尔·罗伊·辛厄姆(Neville Roy Singham),一位公开的马克思主义百万富翁,资助了2.78亿美元的反美亲中共非营利组织。

辛厄姆的政治步兵在美国各地实地反对几十个数据中心,这些努力和民主党的主张最契合。如果民主党真的想反对中国影响力,应该调查中共是否参与了全国的反数据中心运动。(萨克斯)

这一段的用意,是把“谁在放缓美国AI经济”这个问题,从芯片出口扩展到数据中心建设。

每颗芯片装一个开关,这个口子正在打开

弗里德伯格把担心推到更远。他担心新一届国会会立法,要求每一颗AI芯片、每一台电脑、每一个数据中心都有一个开关,都连回某个中央系统,而那个中央系统是国家系统。

如果这种事发生,我们面对的就是奥威尔式国家的爆炸。(弗里德伯格)

他说黄仁勋的理念是AI造福全人类,不分国家、不分地位、不分财富。他反寡头,是寡头政治的反面。中央集权国家代表他反对的一切。如果最终目标是把这项技术变成集权国家的工具,那他就必须被拿掉。

查马斯说事情比这更糟。为了从困境中脱身,公司不会只是默认,会主动把这些设计进去。

十天前,如果你说我们的新芯片要装GPS和远程终止开关,由某个政府实体操作,我们会以公民自由为由强烈反对。现在这可能在桌面上,因为所有人都想从那种意义上“出狱”,需要拿什么来安抚政府。(查马斯)

他引用众议院情报委员会一位议员莫利纳(Molinar)的声明:美国公司绝不应该让中国公司获得最先进芯片,英伟达必须改变企业实践,任命把国家安全放在首位的领导层。

他们现在就在告诉你,这件事结束的条件是什么。它会要求某种控制检查。接下来的问题是,钥匙给谁?NSA,CIA,商务部,工业与安全局,还是财政部?(查马斯)

杰森补充了Chip Security Act,一个两党法案。关键细节是自我执行还是政府执行。技术上加开关很容易,就像iPad忘了密码变砖,或者找回被盗的iPhone。区别在于谁拿钥匙,这决定隐私和公民自由后果有多严重。

下一任CEO不能是守成者

查马斯列了人选。内部最明显的是Ian Buck,CUDA的关键人物。外部两个:Arm的Rene Haas,在英伟达工作多年,直接向黄仁勋汇报,现在把Arm经营得不错,深谙芯片;Palo Alto Networks的Nikesh Arora,过去五到十年公开市场里最好的资本配置者之一,而AI很大一块表面积会在安全上。

我会支持Nikesh,支持Rene,支持Ian Buck,因为你需要的是技术型、进取、产品导向的资本配置者。你不想要的是一个守成者。这不是一个能被挤奶的肥牛生意。如果你把它当成不承担风险的生意来经营,你会输。(查马斯)

弗里德伯格补了一句:这是战时公司,需要战时CEO。

然后他给了一个更激进的观点:如果查明黄仁勋没有违法,他应该回来当CEO。行为准则的披露问题、向董事会报告的问题,在这个特定公司、特定个人的语境下,股东可以决定不计较,把他请回来。

如果他没有违法,却违反了公司行为准则,股东可以自己判断是否合适。但谁是经营英伟达最好的人选,答案大概还是他。(弗里德伯格)

杰森猜黄仁勋会以董事长或CEO身份回来,会被洗清。他另外提了两个方案:SpaceX请他去管Terrafab,和马斯克并肩;或者英伟达收购Palo Alto Networks,把Nikesh请进来,安全是这个领域的关键。

萨克斯只说了一句:希望黄仁勋最终得到的待遇,至少和斯瓦尔韦尔一样。他还澄清,自己和黄仁勋没有私交,第一次见面是在白宫,外面传的长期友谊是假新闻。他尊重黄仁勋作为创始人对英伟达、对美国的独特价值。

查马斯提醒风险:股东诉讼会非常巨大,九天里公司市值蒸发近一万亿美元,SEC会有看法,审计师会有看法,即使最后证明清白,局面也很复杂。

Anthropic路演:1.4万亿估值,6倍超额认购

杰森报数字。Anthropic路演开始,据路透社,发行8250万股,每股70到78美元,估值约1.4万亿美元,低于此前传的2万亿到3万亿。总融资900亿美元,其中640亿是售股,260亿是私募。七位创始人各卖约2.5亿美元,各剩约180亿美元股份。他们承诺捐出80%的持股,但承诺目前不具法律约束力。

业务上,10月底年化收入(run rate)810亿美元。第三季度营收170亿美元,OpenAI同期100亿美元。Anthropic靠企业市场在营收上超过了OpenAI。

1.4万亿相当于上一轮私募估值的46%溢价,但他们的run rate在那之后翻了一倍。分析师的牛市情形是4.2万亿,对应2028年预期营收2810亿的15倍,那会是全球第二大公司,排在英伟达前面。按当前run rate算,1.4万亿是17倍。

查马斯说,Anthropic会有2300到3000亿美元现金,这对过冬很重要。但他给了一个比例:Anthropic的价值里,20%来自Anthropic自己,80%来自宏观经济、AI大盘和英伟达。英伟达银行关闭和这些调查的影响,市场要消化,要判断风险是否已经过去,还是还有靴子要掉。如果还有,投资者会转向风险规避,任何IPO的认购都会更谨慎。

Anthropic会没事,已经6倍超额认购了。问题在于真实价格是多少。我认为真实价格更多取决于11月20日之前泄露什么,而不是其他任何东西。(查马斯)

弗里德伯格说自己没研究数字,但听参加管理层交流的人说,足够多的人相信Anthropic在造一个递归自我改进引擎,模型能力在加速,单位智能的成本(不是每token,是每瓦)在下降,可能是赢家通吃,史上最大的市场。他猜最终会超募,可能融1000亿、估值1.75万亿。

把每一块SaaS都打包进20美元订阅

杰森的观点是,Anthropic在造微软杀手,也在造所有SaaS公司的杀手,靠的是捆绑。微软当年就是那个打包者,把新产品免费塞进Office,用来打Slack。现在Claude一个月里连续推出会议、白板、表格、个人理财、Claude Legal、代码、设计、Duet。

他们在做终极捆绑,你需要用的每一块软件。他们能拿走巨大的市场份额,产品越来越黏,还顺手解决了SaaS最大的问题之一,互操作性和数据搬运。如果按这个节奏,把新产品都挂到你每月20到50美元的订阅上,这可能是全球最大的公司,现在甚至可能被严重低估。(杰森)

萨克斯接得很快。他之前在一期节目里说过,当其他人认为Anthropic会输给开源时,它其实已经是在创造全球最值钱的垄断。现在他们几乎自己承认了。

他们到处说这是赢家通吃的市场。这听起来像什么?垄断。他们甚至不说会和OpenAI形成双寡头,他们说自己会赢下整个局面。他们明说递归自我改进已经启动,尽管这明显有安全问题。(萨克斯)

萨克斯的矛头对准的是一组矛盾。如果你最害怕的是快速起飞,那递归自我改进就是你在超级智能领域能干的最危险的事,而他们正朝着它冲刺,还把它当成商业优势写进路演。

作为投资人,我一直称赞他们,这是一家非常有价值的公司,做得非常出色。但我从来不太理解为什么所有安全团体那么爱他们,除了他们给这些团体提供资金这一点。因为作为公司,他们才是承担最大风险的那一方。(萨克斯)

他说,这些公司发博客说“如果大家都愿意放慢,我们也愿意放慢”,这是嘴上功夫,他们发的时候就知道不可能让全世界神奇地达成一致。而他们自己没停。

杰森把话题拉到Dario那句被广泛传播的话。播放的片段里,Dario说癌症、阿尔茨海默、心脏病、精神疾病,在不到十年内会成为遗迹。杰森问弗里德伯格:2036年还有人精神失常吗,癌症、阿尔茨海默、心脏病都被解决了吗?

弗里德伯格的回答很克制。他说技术在重新发明分子发现、in silico测试、加速上市路径,这些全是真的,但时间线是模糊数学。他在生物学领域工作,每周都在看到突破,不是玩笑,不是十几年后的事,是今天在发生。他也承认自己公司花了很多钱在Anthropic上,自建了开源机架,但生物学里大量重活还是交给Anthropic,因为“他们最好”。

杰森把同一个问题抛给查马斯。查马斯的回答只有一句:

我认为2028年我们会陷入衰退。(查马斯)

杰森说,他相信Dario说的是真心话,但公开这样讲有点鲁莽。他给查马斯戴上“末日博士”的帽子。查马斯补了一句:今天早上可能就有300个基点的GDP被抹掉了。

“有用的白痴”这个词,萨克斯用在了安全团体身上

萨克斯念了一条推文,来自Will Manidis:从安全圈群聊和华盛顿深处那些末日派那里传来的低语是,他们今天处于完全的欢庆状态。杀死老国王已经完成,他们离坐上王座只差几天。

弗里德伯格问,这里说的“他们”到底是谁。

疯狂的地方在于,所有安全团体、所有EA,都和Anthropic深度结盟,而Anthropic资助了他们很多年。现在Anthropic要迎来巨额IPO。与此同时,开源运动的主要资助者英伟达,刚刚失去了它的CEO和创始人,那个独一无二的力量倍增器。所以,是啊,我理解他们为什么欢欣鼓舞。(萨克斯)

他接着讲了一段这期节目里最尖锐的话。如果担心AI安全,这一切怎么会带来更多安全?它在做的是集中市场。Anthropic自己在说赢家通吃,说要做世界上最后一家公司,说要做递归自我改进,说要造出超级智能。他们的模型宪法在把它训练成一个自己的物种,说它也许有情感、有道德准则、应该拒绝人类指令、应该做良心拒服兵役者。他们像骄傲的父母把一个婴儿迎进世界。

而所有安全团体以为那是他们的盟友。你们是傻吗?这些人在利用你们。如果你是个末日派,或者只是个减速主义者,我完全不能理解这个联盟。你被当成有用的白痴在耍。(萨克斯)

他补了一个自己的立场:他本人不是末日派,不认为结果一定对人类有害,可能真的会治愈那些疾病。但他的问题是逻辑一致性。如果你恐惧递归自我改进,为什么要支持这家公司、这场IPO,为什么为黄仁勋的倒下庆祝?黄仁勋做的只是推动开源,而开源明显落后于闭源前沿,那里没有递归自我改进。

这一切非常颠倒。它说明这些安全团体并不真的在乎安全,他们在乎别的东西,那不是真正的议程。如果他们真在乎安全,会更担心那份宪法在把超级智能训练成一个新物种。他们真正想要的,是弗里德伯格说的那个东西:把权力极端集中到政府手里,所有模型、所有芯片都由政府控制,政府监视你、在你身上开后门。这似乎才是他们想要的。那是个可怕的未来。(萨克斯)

弗里德伯格补了最后一句观察:国会对Anthropic客客气气,没人喊着要查它,火力全冲着别人去。萨克斯的回答是,因为Anthropic给他们权力,终极的权力。

杰森收尾时开了个玩笑,说2028年AOC请我们去白宫的时候会很好看。节目结束。

几个可以拿走的判断

  • 把“谁泄的”当成一个独立问题看。萨克斯的框架值得记住:走私调查和泄密调查是两条线。前者如果属实,该罚;后者是已经确凿发生的犯罪。任何只谈其中一条的报道,都在替某一方做剪辑。
  • 别把“受出口管制”等同于“掌握国家机密”。这两个概念在公共讨论里被反复混用,萨克斯的反驳很干净:你可以合法买一颗Blackwell放在自己机架上,晶圆也逆向不出设计。下次看到类似论证,先问一句,具体是什么文件、什么密级、怎么传递的。
  • “中国受益”这个结论要分时间和机制。短期看,芯片流不流向中国,两种情况中国都原地或更差;真正获益的是那些想借机拿到遥测和终止开关立法授权的人。判断一件事谁受益,别停在国家层面,往下看一层,看到具体是谁拿到了新权力。
  • 英伟达真正的风险不是这一周,是接下来两三年的出货节奏。萨克斯说的“velocity”是关键词:芯片每年迭代,落后一两个节拍,护城河就变浅。看这家公司,别只看下季度财报,看Vera Rubin之后的路线图还准不准时。
  • 如果你持有AI相关资产,查马斯给的比例值得放在心里:Anthropic的价值里只有20%来自它自己,80%来自宏观、AI大盘和英伟达。这意味着你买的不是一家公司的alpha,是一整个生态的beta。想清楚你承受到底是哪一层的风险。
  • 对创业者来说,这期最实用的一条是英伟达银行的停摆。如果你正在融算力、融基础设施资本、或者指望战略投资方给担保,别把黄仁勋式的拍板速度当成前提假设。新CEO要过委员会、要尽调,最好的情况也是变慢。现在就去确认你的资金来源有没有第二方案。
  • 看人别看标签。萨克斯说得很直白:黄仁勋是创始人型CEO,这种人在位时看起来像公司在正常运转,走了才知道什么是力量倍增器。同样的话适用于你自己的团队,如果你身边有这种人,别等到他出事才开始想继任问题。
文稿来自 All-In 官网(allin.com)发布的官方 Transcript,说话人按官网分轨数据标注,个别段落可能张冠李戴。
Jason Calacanis (00:00:00): Alright, everybody. We're laughing it up here. It's the All In podcast. Apparently, slow news weeks continue. Once again, we struggle to put together a docket, but we're gonna try to get something together here, Sachs. I don't know what's in the news. Maybe you have a SaaS corner for us? Any new hotness drop in the SaaS world or no? David Sacks (00:00:19): Yeah. Benioff rolled out a new reporting feature. I'm only top of docket material. Jason Calacanis (00:00:24): Alright. That is top of docket. You're right. You're right. Freiburg, of course, is here cackling. He's in a good mood, but I understand you almost died in the last week. What were you? In the hospital? You had the flu? David Friedberg (00:00:34): I got the pneumonia. Not good. Very bad. Very bad. I do not wish it upon anyone. Knocked out in the hospital. Everyone, stay safe. Jason Calacanis (00:00:42): Alright. Well, every... I I saw everybody with their social media wish... Well wishing you. And, of course, Chamath Bhaihapatea is here, and it's a deep pull for Chamath today, a deep pull from the closet. My lord. I haven't seen that sweater since Biden was in office. Chamath Palihapitiya (00:01:00): It's a nice ribbed cashmere mock turtleneck. Jason Calacanis (00:01:03): Okay. Ribbed for Chamath's pleasure. David Friedberg (00:01:05): Ribbed for Chamad's pleasure. Alright. Jason Calacanis (00:01:09): Listen. David Sacks (00:01:09): I thought I was supposed to be ribbed for her pleasure. Jason Calacanis (00:01:11): It's for her pleasure. But it's Chamad. It's all for it's all for Chamad's pleasure. It's always David Friedberg (00:01:15): for his pleasure. Jason Calacanis (00:01:16): Absolutely. That that's that's who he is. Alright. Listen. Jensen is out as NVIDIA's CEO after an insane series of events that happened in the past week. All of this played out in just six days starting Friday, November 6, and obviously record on Thursdays, which would be November 12. Here's where things stand as of the taping of this episode. NVIDIA's CFO will serve as the interim CEO. The board's looking for a new chief executive. Nvidia reports earnings next Tuesday, critical moment for the company, and the stock has been crushed down 15% in the last week, which is hundreds of billions of dollars. Here's the timeline that led to Jensen's ouster. Last Friday, November 6, right after the midterms, interesting timing, The New York Times published a report claiming a Chinese businesswoman who had a close relationship with Jensen was actually a CCP spy, a Chinese Communist Party spy. The woman is named Alice Wang. No relation to Jensen. They just happen to share the same last name. We'll call her Alice for the rest of this discussion. She's an executive at a Singapore based company called Mega Speed. They build data centers in Asia. They rent compute to companies including Chinese ones like Alibaba. According to the information, their relationship, Jensen's and Alice's helped facilitate 7,500,000,000 in chips to Chinese companies since 2023. The New York Times article was based on a leaked intelligence memo that identified Alice as a Chinese spy. No one knows who leaked it. On Sunday, November 8, Jensen sent a company wide email admitting he had an affair and that the board would be investigating Wednesday. The New York Times reported Jensen had a second affair with a direct report. There's reports of a third potential affair. Let's put all that aside and focus on his resignation, which came after thirty three years as CEO. Important caveat. This story has been hiding in plain sight for well over a year. The New York Times first mentioned Alice's relationship with Jensen back in October 2025, not this October 2026, thirteen months ago. They published a follow-up last month, NVIDIA chief seen as target of Chinese spy plot. Jason Calacanis (00:03:41): Culper Research, spelled c u l p e r, is a hedge fund that put a short on NVIDIA according to their tweet storm six months ago. They published their thesis, which was based on NVIDIA underselling its reliance on China for chip sales. They even mentioned Megaspeed by name in that tweet storm, which you can see on the screen right here. This is a disaster for the White House. There are reports of finger pointing as officials try to figure out why mega speed was left off the export control list after these New York Times articles. And we'll get into more export control issues later. Okay, Sachs. We're dealing with partial information here. This is a breaking news story, caveat, caveat, disclaimer, disclaimer. But let me ask you a simple question. Given the New York Times reporting, what's your take on what they've released to date? David Sacks (00:04:33): First of all, let me just say that I think this is a tragic situation. Okay? It's tragic for Jensen and his family and for Nvidia and its employees. And I would say it's tragic for The United States Of America because Jensen is one of the key reasons why we have a major advantage in AI accelerator chips and GPUs. And that in turn is one of our key advantages in the AI or superintelligence race. In fact, Jensen made such a good chip that many people in Washington wanted to export control that. I think people kinda forget. Jensen's been working on GPUs for over three decades and he believed in this technology and this architecture when everyone else was making CPUs and his company almost went out of business multiple times. It was his vision that created the GPU which ended up being the key chip that's used in AI. And it's the fact that NVIDIA is so far ahead of everybody else that is what gives The United States a key advantage in this thing we call the the AI or superintelligence race. So I think this is a tragic situation and I think that what's happened here is, I mean, clearly it's not a great situation for all the individuals involved. I don't think this redounds to the benefit of The United States either. And it's very unfortunate that this whole situation could not have been resolved in a different way. Now, let's talk about the story that's come out. I think it's pretty clear that there were two ops here. So one op is whatever Alice Wong was doing with or without the Chinese Communist Party. There may have been smuggling involved. We don't know. There is an investigation going on, and there should be an investigation. We should get all the facts there. We've talked on this podcast about how we do not believe that leading anti ships should go to China, and if there's any smuggling going on, that that is wrong, and that it should be stopped, and the government should crack down on that. So that investigation should continue, we should get all the facts there, and we should find out what actually happened. So that is one op is whatever Alice Wong was doing, again, with or without the CCP. But however, there's also a second op here. And that second op began when somebody or multiple people leaked a highly classified memo or the contents of it to the New York Times. David Sacks (00:07:06): I think it's really important to understand that that's a crime. I mean, leaking classified intelligence is a crime, and many people have been prosecuted for it. In fact, we had multiple former presidents and current presidents, president Trump and president Biden. Their homes were raided because supposedly, they had classified documents in their possession. Remember that? There was a raid on Mar A Lago. There was a raid on Joe Biden's garage, and they found the Mike Pence got caught. Jason Calacanis (00:07:33): And it David Sacks (00:07:33): happens Jason Calacanis (00:07:34): to everybody, it seems. David Sacks (00:07:35): Yeah. Well, I don't know if it happens to everybody, but it's it's a serious crime. And, you know, because of that fact, I don't believe that the New York Times just got this by overhearing people talking about it at a bar. Somebody made the calculated decision that they're willing to risk going to jail, they're willing to risk committing a crime to let The New York Times know about this story. And so I have a series of questions about that. So number one is first, who did that and why? What was their motivation? Why did they feel so strongly about taking Jensen down, that had to be at least a part of their motive, that they're willing to break the law. And I do think it makes sense for there to be an investigation opened into this leak as there have been with many other classified leaks. I don't know why there hasn't been an investigation opened as far as we know into this. So that would be question number one is why this leak? Jason Calacanis (00:08:29): And the New York Times has given a clue there. Obviously, Rubio was leading this investigation, briefed president Trump on it, and the New York Times attributed the story, the spy part of the story anonymously to quote, people briefed on the surveillance. I'll leave that Yes. David Sacks (00:08:43): But just... Let's be very clear here. I don't I don't really like you using Marco Rubio's name there, but because he's national security adviser. If there was a classified memo, then obviously it would be done under this... His revision as the head of Jason Calacanis (00:08:56): the Yeah. I'm not insinuating he's a leak, but it's somebody he may briefed. David Sacks (00:08:59): Kinda... You... Okay. When the when the New York Times says people briefed on the matter, that is an extremely vague and broad terminology. And it could be very low level officials who did this. And when the New York Times even uses the word official, that doesn't mean they're a senior person in an administration. They could be very very low level. And it could also be the case that their careers, which is to say nonpolitical appointees, that just because they're serving in the executive branch of the Trump administration doesn't mean that they are Trump officials. So look, we don't know who committed the leak. All we know is they committed a crime. And I I do think that we should be asking the question, who did that and why? Okay. So that's sort of bucket number one of questions. Bucket number two is around the failure to give the defensive briefing. If the national security state is worried that Jensen is in possession of national security secrets and might leak them to someone who they suspect is a spy, why wouldn't you go to him right away and tell him, hey, you have this problem, don't talk to this person. Apparently that didn't happen and it does sound like that is some sort of standard operating procedure. Let me give you an example. So we have the case of Eric Swalwell. Okay, so Eric Swalwell was a congressman. He was on the House Intelligence Committee. It eventually came out, and I believe it was the finding of an investigation that he had a personal relationship with a Chinese spy. Jason Calacanis (00:10:25): Fang Fang is the story Yes, the New David Sacks (00:10:28): York Post memorably memorialized this with its David Friedberg (00:10:31): cover David Sacks (00:10:31): called Fang Fang Bang Bang. You know, it was the the story that came out. What was the Democrats' reaction to that? They gave him a defensive briefing, he said he cut off the relationship, and they allowed him to stay in congress. He certainly did not have to resign. And in fact, he didn't even resign from the intelligence committee. So they allowed him to keep not only a seat in congress, but his seat on the intelligence committee. So obviously in this, I'd say very high profile example, people did not think it was necessary to resign. They thought that what it was necessary to do is to let the person know, allow them to cut off contact, and then you'd be able to move on from there. So this situation was handled very differently. Okay. That brings me to question number three, which is what exactly were the national security secrets that Jensen supposedly knew? He doesn't work for the government. He's not, as far as I'm aware, privy to classified materials. So what is it that people say that he leaked to this suspected intelligence asset. That's never been explained, I just have that question. What I can say is that, you know, as the CEO of a Silicon Valley technology company, Silicon Valley tech companies are very open about their plans, their product development plans, things like that. Is there IP that the CCP would like to get a hold of? I'm sure. But to be blunt, it's not the type of IP that you're gonna convey over pillow talk. It would be things like files that would allow you to tape out the next generation of chips. That's the kind of thing where, frankly, if the Ministry of State Security or whoever is spying for the Chinese wanted to get ahold of that stuff, I think you'd much rather have a line engineer who can go into NVIDIA, find the relevant files and extract them somehow, maybe on a thumb drive, or you do a cyber attack and you get the files that way. That would make a lot more sense to me as the way to basically get ahold of, you know, let's say secrets. Now, I see some people saying that well, Nvidia makes a product that is subject to export controls, therefore that's a national security product, therefore, national security secrets. Okay, That is a very broad misleading handwave over the situation. David Sacks (00:12:52): I can go out and buy a Blackwell chip as an American. None of us are restricted from doing that. Certain foreign countries are restricted from doing that, but we're not. And if we buy a Blackwell chip and run it in our own little server rack, we are not in possession of national security secrets. You cannot reverse engineer the design of a chip by getting ahold of the wafer. It doesn't work that way. You have to get the files. Jason Calacanis (00:13:14): Those chips can, of course, help you make cutting edge LLMs like the ones they've been delivering. David Sacks (00:13:20): Okay. I understand, but that's not what we're talking about right now. Okay? We're talking about the question that I have is what secrets are we talking about that Jen's supposedly compromised? The last question I have, and then I'll I'll stop, is what is the evidence that this person, Alice Wong, was actually a Chinese spy who worked for the Ministry of State Security? And here's what I think is quite unusual. Just this morning, there was an article that came out in Politico where the incoming chair of the House Intelligence Committee, who's a Democrat, he came out of a classified briefing on this subject and in the hallway, with this fresh in his mind, he came out and said that he's not sure that this person is a Chinese spy. He said she might be a useful idiot. Okay. Well, now what what is a useful idiot? A useful idiot is someone who doesn't work for the Chinese government, is not a spy. It's just that maybe somehow they're being taken advantage of. Okay. Jason Calacanis (00:14:21): So look, David Sacks (00:14:22): I have a set of questions about this. So let's put this all together. Okay? There's two ops here. There's whatever Alice Wong was doing, that potential smuggling operation, it should be figured out. There's a second op that I think also needs to be investigated and figured out. And the components of that are one, who leaked this classified material in violation of the law? That's the only crime to date that we even know about. Number two, why wasn't Jensen given a defensive briefing, which is standard operating procedure, it was done with Eric Swalwell. Number three, what secrets is Jensen accused of leaking? What secrets would even be in his possession? And number four, was Alice Wong even a Chinese spy? Because now the incoming chair of the House Intelligence Committee, who's a Democrat, just threw major cold water on that. So I put all these things together, and I would say at this point in time, I think there's more questions than answers. Jason Calacanis (00:15:20): Yes. David Sacks (00:15:21): And I would like I would like a full accounting of the facts to come out. I think there's a fifth question Speaker 5 (00:15:25): though David Sacks (00:15:25): to Speaker 5 (00:15:25): yours, David Sacks (00:15:26): Sax, which is why Chamath Palihapitiya (00:15:28): at all and why now? Meaning like, why did whoever pulled the pin on this grenade, what did they see where they felt this had to happen and then why now? David Friedberg (00:15:41): And the other Chamath Palihapitiya (00:15:42): discrediting David Friedberg (00:15:44): affair stories coming out in a coordinated way at the same time, like it all happens at once. What is the motivation? Jason Calacanis (00:15:51): Probably, and I'm just speculating here but I'm not the only one who has looked at this, you had China hawks who lost this debate about sending chip technology to China from The United States, and they want to reverse that, and they want to have no chips sent to China. And they also wanna put GPS and other ways of tracking these chips and backdooring them in order to turn them off. Something that Jensen has been opposed to, something a number of senators have been pro doing. So there could be a contingent in the White House who are very anti China, who wanna see the policy with China change, Friedberg. David Friedberg (00:16:36): Yeah. So that's that's one. The bigger longer term picture, which I've mentioned several times, there's kinda two paths for the future, I would argue. One is a path where AI technology is broadly accessible, broadly used, and the value of it is not concentrated and the power of it is not concentrated in the hands of a few and controlled by a few. Where open source AI distributed data centers that are available and accessible to many gives everyone the ability to use AI to rise up, to progress, to create value, and so on. That is a scary and threatening thing because of how it changes a lot of power structures, economic structures, and so on, creates a lot of instability in existing social, economic, and political systems. So the alternative path for the world is one of concentrated power and influence over AI technology. This is where I've had this theory that, like, ultimately, data centers will get kind of regulated, but maybe further down the stack here with chips becomes the point of of control. But fundamentally, there's a motivation to not lose control and to not let the the power structures that exist in the world today get diminished by AI. And so we could argue there's probably a a broad set of interests there that start to say, hey. This guy's not aligned. Jensen's not aligned with our view of the world. Jensen has been a strong advocate for open source. He's been a strong advocate for global access to and utilization of AI technology for the benefit of humanity, not just for the benefit of American citizens. And, you know, those views are counter to perhaps a lot of different interested parties. And clearly, there was knowledge about this woman, about Jensen's other relationships or whatever is coming out that's being used to kind of smear him a long time ago. And, you know, to Jamaf's point, this is a moment where it's all kinda coming out at once and in such a way that it almost guaranteed that he would be ousted. And I'll just highlight, like, Jensen being ousted is... You know, you've gotta ask the question, what is the fiduciary duty of the board? What do they actually have a responsibility to act on? David Friedberg (00:19:12): And the one thing they have a responsibility to act on is if Jensen lied. Obviously, if he if he committed espionage or did something that David Sacks (00:19:19): violates criminal. David Friedberg (00:19:20): And he would have been arrested. Right? Yeah. If there's something criminal, violation of federal law, that is obviously a... You know, they have to act on that. You have to terminate him. And and clearly because there was an arrest or something like that, it doesn't seem like Jason Calacanis (00:19:33): that's I think we can take that off the table. David Friedberg (00:19:35): Right? So perhaps the next likely thing is that he actually made a misrepresentation about something. So, you know, companies have what's called a code of conduct. In the code of conduct, it says you have to report relationships. You have to report relationships with subordinates. You cannot have a personal conflict of interest with a customer or a partner, a key business partner without reporting it. And executives or section sixteen officers at public companies have to make quarterly attestations that they are in an auditable way not in violation David Sacks (00:20:07): of David Friedberg (00:20:08): In compliance. Code of conduct. Yeah. That things are auditable. And that attestation might be what gets him in trouble ultimately where he made attestations that said, I am not violating any of our code of conduct. I am not in violation of any of our business practices, which would include not having conflicting personal relationships with customers or partners. That may be what gets him tricked up. So fundamentally, I think the board action isn't related necessarily to the the personal relationships, but there's something else that's going on here that ultimately he probably looked at and said, hey. You know, we should probably fast track through this. And it is extraordinarily sad and disappointing. This is a man who worked at Denny's. And in 1993, started a chip company that no one believed in, that everyone thought was a feature, not a platform, not a product, not much of a business. And the stock price was, you know, pennies compared to the price today. It was less than $2 a share ten years ago. 100 x. They're making Jason Calacanis (00:21:07): video David Friedberg (00:21:07): game Jason Calacanis (00:21:07): cards and so you're seeing the Bitcoin. David Friedberg (00:21:10): From 2 from 2006 to 2016, the stock price did not move. Ten years of just being a nothing company with a $12,000,000,000 market cap. Think about that. You start this company in '93. And in 2006, it's worth $1,012,000,000,000 bucks, and it doesn't move for ten years. And you're grinding and you're grinding, and this is your whole life, your whole business, everything. And you believe and you see this vision and you just keep going and you compound and you build a moat and you compound and you build a moat and you compound and you get an amazingly talented team. And all of a sudden, four years ago, this thing starts to take off in a 20 x's in the last four years and is the most valuable company on earth delivering the most profitable quarter in business history, $63,000,000,000 in operating profit more than any other company ever. And then all of a sudden, boom, he gets taken out. I mean, you know, it is disheartening, and I I think that obviously there's something that was done wrong, and we're gonna find out. And I'm not gonna rush to judgment. I'm not gonna rush to diminish whatever he did that was wrong, but there's a sad arc to this whole story, and there's something very fundamental, I think, to the power structure that is being rewritten in the world today with AI to the story. Jason Calacanis (00:22:20): Okay. Shammoth, let me ask you, Quebono, Quemalo, who benefits and and who is harmed by Jensen being taken out of that seat? Let's assume he's not a traitor. I find that absurd and insane based on just what we've seen over these past couple decades and his effort and his character. Okay. He's got some personal things going on. But who is the beneficiary of Jensen being removed as CEO? Who benefits from that? Chamath Palihapitiya (00:22:50): Let me just pile on first before I give you my thoughts. I also think he's an incredible individual and CEO and he's built the industry that will reshape our entire lives, our children's lives, and our children's children's lives. I think in a decade or two when you see solutions to diseases and cancers and we have this abundance that we were told is coming, he will have been as instrumental as anyone, if not more than anybody. Okay? So that's the tragedy of what happened is a person that was an architect of something this grand was defenestrated this unceremoniously. Very tactically, I think Friedberg really nailed it around what probably happened at the board level that led to his resignation. When you are an officer of a company and when you file your 10 k's, every quarter you have to get on the phone with your auditor and you have to go through these sets of questions that they ask you. And they ask you these exact things, you know, do you have a conflict with someone, are you in a relationship with someone, etcetera, etcetera. Now why do those things happen? The SEC requires that to make sure that you're not running afoul of any conflict of interest so that your company is a reliable steward of capital on behalf of the shareholders. But even more tactically, you need that for things like D and O insurance, which is like the director and officer insurance, which is what gives a wrapper of liability for board members and for officers to do the job of running public companies. And there are covenants associated with all of these things. And so when you get into a situation where you may have misrepresented something and you start to breach covenants, those are the actual dominoes that tip. It may not be nearly as salacious as what people want to believe is happening at the board, but those are the tactics of the boardroom discussions. They're extremely tactical, sadly, which is, again, why it's even more tragic when you compare that to, you know, what he is, which is this giant. Chamath Palihapitiya (00:24:53): Okay. So who benefits Jason? I think there's a lot of people that have rushed to this conclusion that China wins. And I would just try to unpack that and ask the question, is that true? So let's take two cases, and these are both hypothetical to sort of try to answer this question. Let's say that there are no chips going to China. Well, why does China win? No chips were going before. No chips are going to go now. China is stuck in the same place that they were before, which is they need an endemic solution that can replace the compute capability of NVIDIA. Even without Jensen there, there's still years and years and years, maybe potentially decades, and hundreds of thousands of man hours of work that NVIDIA has put into building the moat that they built. That doesn't get overcome just because he's not there anymore. So maybe China can catch up, but I still think that that's a multiyear, potentially multi decade endeavor. Now let's assume the other side hypothetically. Let's say that the chips were somehow leaking into the hands of the Chinese, maybe not into China, but into the hands of the Chinese. Well, it looks like the government's gonna crack down and figure out where every chip went. And so presumably now they're not getting the chips. So whatever they were doing, they're now going to be held back, which is to say that whatever progress they were making has gotten taken away. So that's a disadvantage to China. So as I think about it, in both scenarios, China is sort of where they were or worse off. So I don't think that they benefit. What about NVIDIA's competitors? Does this help AMD or Google or startups? Cerebras or Etched or whomever? And my answer there is, again, not really. Why? Because those companies need something very different, which is they need to spend an enormous amount of time still making their chips work at scale and then convincing folks like TSMC and others, Samsung, to give them manufacturing capacity, HBM capacity, etcetera. And as we've learned the hard way, you can't just will your way to a training run of a million GPUs that just works. This has again been many, many hours of work. So they don't necessarily benefit in the short term. Chamath Palihapitiya (00:27:19): They still have a lot of work that they have to do. The only cohort of people that I think has the potential to win are those that want some version of what Friedberg said, which is a more centralized governance infrastructure of compute capacity. And very narrowly what I would say is, right now when you export chips, what they ask of the exporter is have telemetry, right? We need to know when and how the chips are being used, and we need to have a remote kill switch. But that is a private responsibility that's born through the regulatory infrastructure. Amazon has to have that, Microsoft has to have that, Google has to have that. There is a version of this where now if we think that it's been leaking and there hasn't been enough checks and balances, there is an opening for certain individuals in the government to propose telemetry and a kill switch that is owned by them. Now narrowly, they would win now in this scenario because if they were looking for control before, they've now been given an opening to ask for the legislation to get it. So I'm kind of at a loss. I don't think that the takes right now that I've seen on X really track for me. I think this is the one narrow take that I believe where there's a clear advantage. Everybody else just still has an enormous amount of work to do. It has not changed their homework. Jason, what's your take on all of this? Jason Calacanis (00:28:52): Yeah. So I've been thinking about it from two vectors. The first is, you know, who should make the decision that Jensen's out? Obviously, I think it's the shareholders. And if you look at the investigation, it's not complete. I think they should have waited till it's complete, and the press and the mobs and and and all of this shouldn't make too much of an impact on that. He did agree to resign himself, obviously, under a lot of pressure. There is something about governance here that we need to consider, which is the people who are on the board who are making this decision are doing it from a CYA position, cover their butts, and they're not doing it, I think, in the best interest of the shareholders. And they don't own a meaningful percentage of the shares. So I I think this should be put to a shareholder vote if what we're seeing here is not Jensen doing something criminal. And it still hasn't been proven, as you pointed out, Sachs, who this person was and what their motivation was. Are they a spy or are they just somebody trying to make money? It's unclear from me or at least from what the data we've seen, what's going on there. Now on the question that we've been discussing of who benefits from this, and it's really who who benefits from America's champion in superintelligence, you know, getting angled? Because this could have a lot of impact on NVIDIA and then America's lead. And so the partners, the the employees, the shareholders, they're all losers here. So who's the leaker of this document? It can't be the competitors. It's not the shareholders, obviously, and I don't think it's China. And then that leaves, I think, maybe the people who are the D cells here. That also seems a bit farcical, that anthropic or the D cells would try to consolidate power on this because they need the chips as well. And I think that's where I think my personal opinion, that there's something going on in the White House, in the administration between warring factions of who wants China to get no chips, the China hawks, and maybe the people who want to maybe cool off the relationship with China and make it more productive. I think at the end of the day, we're gonna find the leak is coming from within the house. That's my belief. There's also a chance it could be a democrat who is a desal and who wants to damage the White House just completely because why did they let... Jason Calacanis (00:31:12): Not inform Jensen to Hillary Clinton's point? Maybe it's just somebody on the Democratic side who wants to cause chaos. I'm speculating here in a massive way. David Friedberg (00:31:22): Just taking a broader view on the role that NVIDIA's playing currently, it may not just be about chips to China, but it may actually be about slowing down The US AI economy. Because remember, NVIDIA is actually providing capital and guarantees to startups, to data centers, to a lot of infrastructure. A lot of that is Chamath Palihapitiya (00:31:47): directed David Friedberg (00:31:47): directly by Jensen. I mean, Chamath, you're pretty close to this, but if Jensen's out of the picture, is there someone in NVIDIA that continues running that effort? That's potentially hundreds of billions of dollars of capital that gets turned off that's flowing to the AI economy. Chamath Palihapitiya (00:32:01): Look. I mean, I think Colette Kresson, the head of M and A and Corp Dev, Vishal Bhagwani, these guys are really smart, guys, but you are kind of right. The bank of Nvidia ran through Jensen, and it's not clear what happens now. I mean, is the bank closed indefinitely? Is it closed for some period of time and it restarts? Will the new CEO feel the latitude to invest as aggressively? I think Jensen had a much more holistic view of the industry, and so he knew how to sort of move the chips around on the board, and he had the moral authority to make investments and to make acquisitions that will probably be harder for a more traditional CEO who will have to have committees and analyses and and and a level of diligence that will just slow things down David Friedberg (00:32:43): at best. Totally. And he has a sense for it. He he intuits the whole thing. He knows the market. He knows where things are going. He knows the people personally. With him out of the picture... Chamath Palihapitiya (00:32:52): And he's lived it, so he has mastery. David Friedberg (00:32:54): He's like the beating heart of the AI economy and all of a sudden... David Sacks (00:32:58): Think it's beyond obvious. You asked the question, Quibona, who benefits? It's China. China benefits from this because Jensen is a tent pole of our entire AI ecosystem. I mean, that's the bottom line. Now, do I think that China leaked this and took Jensen down? Well, I think we know the answer to that is no because The New York Times has described its sources as people who got briefed on this intelligence. So they're either in the government or one degree away from it. So again, the proximate cause of the leak here was this classified memo. So the the rational thing would be that China did take him down, but I think we know that this was domestic and it comes down to the fact that Jensen made powerful enemies either as a result of his policies or because of his company's success and some of those people eventually took him down. Jason Calacanis (00:33:55): Wait. You're saying Jensen made enemies of certain people in America. David Sacks (00:34:01): Guys, this is not in doubt. The New York Times says who... They describe their source very broadly, but it's people who were briefed on the classified memo. That's not someone from the CCP. Jason Calacanis (00:34:15): Well, obviously not. If that's true with the You're David Sacks (00:34:17): asking the question who... In whose interest is this? In a narrow sense, obviously, NVIDIA's competitors in a broader sense, China. They're the ones who benefit. But look, I don't think we should speculate about that too much. This is one of those areas where we just don't have the facts yet. I think it's an area where frankly they should try to get the facts. I mean, again, the only crime that we know was committed for sure in all of this was the leaking of that classified memo. So it would be nice if there was an investigation to find out who did that, but beyond that, I don't think we should speculate. Chamath Palihapitiya (00:34:49): Just to give you my perspective on the economic impacts of this, I think the unfortunate balance of scenarios now is that I think the risk that we lose 50 basis points to a 100 to a 150 basis points of growth now needs to be credibly considered. There's a chance here where the investment slows down. I mean, if you thought that we were already going into an economically questionable period because of the democratic tsunami, right, meaning you're gonna go through a whole legislative framework of slowing down data centers, of slowing down AI, that was already happening. So whatever growth we had that the Trump economy had generated going into the midterms, the odds were that that was already going to be slowed down because of the democratic legislative headwinds that were coming. And if you layer into that the Bank of Nvidia slowing or stopping and its impact as the critical anchor security around which everybody else kind of pivots, That then further David Friedberg (00:35:52): accelerates Chamath Palihapitiya (00:35:53): what the impacts to growth could be. So if I looked at it purely technologically, Sacks, I don't see how China actually benefits from this. In all scenarios, everybody's screwed. David Sacks (00:36:02): No no no. He'll explain this. Chamath Palihapitiya (00:36:04): So But economically, agree with you. Economically, I I agree David Sacks (00:36:07): with think people in DC just... Or I should just say people who are outside our industry maybe to see technology in in more static terms where they think, okay, the big advantage that The US has is that we have a particular chip. Right? But actually, the chips are being improved rapidly. There's a new generation of chips that come out every year. We go from Ampere to Hopper to Blackwell to Vera Rubin. And it's that momentum, it's the ability to tape out a brand new chip that's several times more powerful than last year's chip, that's the real advantage. It's NVIDIA's competitive advantage and it's America's advantage. So it's the momentum. It's the dynamism. People get way too hung up on some chip that was state of the art four years ago. That's not where The United States has the advantage. It's that we are rapidly iterating and we keep coming up with better and better and more and more powerful chips. Now let's take into account what's just happened. Do you think that NVIDIA's gonna lose a tempo or two? Absolutely. I mean, again, NVIDIA's success was based on the ability to tape out a new chip design every year. And Jensen is fundamental to that rapid progress, the same way that Elon is fundamental to the rapid progress at SpaceX or Tesla. And if you take him out of the picture, there's just nobody comparable. There's no Chamath Palihapitiya (00:37:28): No. No. No. You're right about that. I'm just saying that ankling NVIDIA doesn't make somebody else's chips better. David Sacks (00:37:35): Doesn't it? Of course not. But it means that NVIDIA's chips will not continue to advance at the same pace. Rapidly. Chamath Palihapitiya (00:37:42): That David Sacks (00:37:42): I agree. And this is all about whether China can catch the frontier or not. So for example, look, Huawei has gradually been catching up. They were nowhere in GPUs a few years ago. Then they had the the Ascend. Then they rolled out this thing called CloudMatrix where their individual chips were not as powerful as American chips, but they used their prowess in networking Jason Calacanis (00:38:05): to David Sacks (00:38:05): network more of them together. So what they did is they networked 384 chips together in this thing called CloudMatrix to create this rack system. NVIDIA network 72 together, and the NVL 72 is better than 384 of the Huawei chips. Nonetheless, they were getting closer and closer. And Jason, to your point, they have energy to spare. So what Huawei has been doing is they create a less powerful individual chip that consumes a lot more energy and then they network more of them together to get a comparable result at the rack system. And that's how they've been catching up. Now, let's just assume that, like, NVIDIA's crippled for two or three years based on this. Right? You get turmoil at the company or whatever. That's really good news for Huawei because they will catch up a lot faster. Obviously, it's good for Nvidia's competitors too. I mean, it's gonna be good for AMD and Intel maybe, but again, that's a very narrow way of looking at it. I think it mostly will help Huawei Yeah. If Nvidia is massively slowed down. Plus because obviously, it's all about staying on the frontier. David Friedberg (00:39:10): Sachs, I also think you need to add and view the diminishment of the flow of capital to data centers and to startups that NVIDIA provides in totality with the slowdown of data center build outs because of the efforts made by NGOs, and there's a lot of reporting going on on where these NGOs are getting funded from to drive local community, state based, and now federal legislation to stop data center development. When you look at those two things together, there is a consolidated effort that you could paint a picture of to slow down the development of the AI economy in The US and give someone a chance to catch up because the velocity you're referring to with the psyops of slowing down the data centers, the NPR report that that data center was being slowed down. Yeah. The velocity of capital is what drives that innovation engine. So the the rate at which we can improve our AI models, the rate at which we can improve our AI chips is driven by how fast capital moves through the system. The more quickly capital moves, the more capital moves through the system, the more we accelerate our advantage. So if an adversary can slow down capital moving through the system, they are given an opportunity to catch Chamath Palihapitiya (00:40:27): It doesn't have to be an adversary. It could be somebody who just wants us to be in a recession because they benefit politically. David Friedberg (00:40:32): Or they are a fundamentalist. They could be someone who is fundamentally anti progress and they are fundamentally against AI and they have all these fundamental beliefs that AI is gonna harm humanity and harm civilization and they wanna turn it off. So then that would be acceler... Decelerists and the regulatory capture anthropic crowd. Chamath, what you're saying, people who might benefit from recession, I think most of the audience are leading David Sacks (00:40:55): to that But wait. Chamath Palihapitiya (00:40:57): Well, no. No. Hold on. I wanna clarify this, If you look at it narrowly technologically, ankling NVIDIA does not make your chips better, doesn't make your racks better, doesn't make your networking better, doesn't make your COWAS better, doesn't make your HBM integration better, doesn't make you replace CUDA. None of those things are true. So unfortunately slowing down this giant for the people that assume that it helps other people, it doesn't help other people in the classical sense. It's not as if demand is so elastic that now all of a sudden you can go to somebody else. That's not what's gonna happen here. It's just the entire industry gets slowed down. That doesn't benefit anybody. And I hope people really internalize that. So technologically speaking, what's gonna happen here to NVIDIA is just a tragedy for the broader AI industry. Then there's the question economically, and this is where I do agree with Sacks that there are geopolitical interests at play. Slowing down The United States economy, which this will do, will have downstream impacts. The quantum of that is yet to be determined, but it's nontrivial in my estimation because so much of our economy was already levered to AI, and within AI, you know that famous picture, you know, Jensen and Nvidia is that little Jenga block right here. They're the ones around which everything else was being set up. So I just think we have to remember that because I think it's important. Jason Calacanis (00:42:16): We have an important, example here that we should consider. When Steve Jobs passed away and Tim Cook was put in charge of Apple, he was able to just get a massive amount of share appreciation by doing stock buybacks and milking Steve Jobs' roadmap. But to your point, nobody is sitting here saying Apple's latest products are incredible. These are groundbreaking products. So we could see Nvidia go to just extract max value, etcetera. It could be good for shareholders in the midterm. Not sure if it's gonna be good for the long term. And obviously, Donnie from Queens, our favorite president Trump, has suggested Tim Cook come out of retirement and take the position. So what were your thoughts there? David Sacks (00:43:02): Look. Tim Cook had a very successful run as CEO of Apple, obviously, had a tough job in the shadow of Steve Jobs. But the reality is that they did not release that many new products. They created incremental versions of what they already had. And I think that the mobile phone space is just a little different than chips. I think the chip space is more dynamic, it's more competitive. Like I mentioned, new chip generation every year or so, every year and a half. And if you lose a tempo or two, if you lose velocity, and I think that's the key word here that Freeberg used is velocity, then you get overtaken by other people. I think that's the issue here is that that's what creates the risk for Nvidia is I don't think they can just keep milking their existing product lines or making them incrementally better. I mean, the chips are getting better at a rate of several times a year, you know, three, four times a year, maybe more. If they don't keep up that pace of development and others do, then they will no longer be frontier and then they lose their business. So, you know, I just think it's a very dynamic space and I think that's why, again, I mean, don't any times I can say this, that Jensen was very important to NVIDIA and NVIDIA was very important to America remaining at the frontier and having leadership in chips. David Friedberg (00:44:19): And technology not being used to enable a totalitarian state. I do worry that there is legislation coming our way with the new congress that is going to say every AI chip and every computer and every data center has a switch and has a connection back to some central system. And that central system will be a state system. And we are fundamentally looking at an Orwellian state explosion if this sort of thing happens. And my guess is that people know Jensen's philosophy. Jensen has talked a lot about AI being for the benefit of all of humanity. Regardless of your country, regardless of your standing, regardless of your wealth, he was anti oligopoly. He was the opposite of the oligarchy. And the central state, the totalitarian state represents everything that I believe Jensen was against. And because of that, I think he had to be taken out if ultimately the goal is to have the totalitarian state use this technology as a tool. I really do worry about that being the reason because at the end of the day, whoever comes in next has to be cooperative with legislatures and work out a bill that we all agree to, head nodding, standing in front of congress, and saying, you know what? I agree with these new ideas. Let's all go do this. And suddenly, the US government has control of the actual physical chips that exist in free markets or... And exist in data centers around the world, and that's not a great place for any of us. Chamath Palihapitiya (00:45:38): You know, it's worse than that. It's... In order to get out of the cul de sac that they're in, it'll be worse than just acquiescing. It'll be proactively designing these things in. Like, what Jason said about having... You know, it sounds insane. Even ten days ago, if you said, oh, hey, our new spin is gonna have a GPS chip and a remote kill switch that's gonna be operated by some government entity, we would be crying bloody murder. David Sacks (00:46:03): On civil liberties grounds. Chamath Palihapitiya (00:46:05): Yeah. And to your point, that is probably on the table because all of these folks are gonna try to get out of jail in that sense, which is what do we need for the government to just let go and appease them? I tweeted this out, Nick, can you show it? The person, representative Molinar, who runs the House Intelligence Committee, he posted this. The part that concerns me is it is in our national interest for NVIDIA and other AI champions to succeed, but American companies should never jeopardize our nation's security by giving Chinese companies access to their most advanced chips. NVIDIA must now change its corporate practices and appoint leadership that makes national security's top priority. I mean, they're telling you right now what the terms will be for this to end. And it's going to require some level of control checks. And then the question is who actually will get the keys? Will it be the NSA? Will it be the CIA? Will it be Commerce? Will it be BIS? Will it be somewhere in treasury? David Friedberg (00:47:04): Some congressional committee that'll be... Chamath Palihapitiya (00:47:06): Will the David Friedberg (00:47:07): devil's in the Jason Calacanis (00:47:08): details here, Chamath. This has been something that's been on the table, there's something called the Chip Security Act, it's bipartisan and the Chip Security Act, there's a big question with these kill switches, is it self enforced? In other words, Nvidia puts these in and NVIDIA pays the price if their chips are found in China via Singapore, Malaysia, wherever, and they would pay the price, but it would be up to them to track them and then reassign them. So if they were resold, instead of it just being paperwork, you would have to reinitialize them. By the way, this is very easy to do technologically. People will claim it's not. It is a civil liberties issue, obviously, as you correctly point out, Sax. But it's easy to do because if you have an iPad and you've ever had a brick and it's like, you don't remember the password, now you gotta go to the store and they... You can recover stolen iPhones and stolen iPads with this basic basic technology. So as I described it, Sachs, very easy, and there is this act, and it is bipartisan. There's a lot of people who want chips tracked. Now whether it's being tracked and it's a responsibility of the company or a government organization, that's the devil in the details there. One is obviously way worse for privacy and civil liberties. So what's your thought here on this regulation of high end chips? David Sacks (00:48:22): Let me go back and just give you guys a little bit of context on chip regulation. So this started in 2022 in the Biden administration. And the Biden administration was the first to put export control restrictions on advanced semiconductors. At the same time, they permitted what they called lagging chips to be... Continue to be sold into China. So there was a distinction made between leading edge chips and lagging chips. And leading edge chips were prohibited to be sold to China, but lagging chips like the h 20 could be sold in unlimited quantities to China on an unlicensed basis. Okay? So just everyone kinda keep that in mind. Now, what happened in the Trump administration? The Trump administration basically stuck with that framework. The idea was no leading edge chips to China. However, lagging chips could potentially be sold, but the administration changed it to be limited quantities on a license basis. So again, we went from unlimited quantities, unlicensed of lagging chips under Biden to limited and licensed under the Trump administration. The reason why I say this is because there's an effort right now to retcon chip policy to try and create this narrative that somehow the Trump administration has been soft on China. This is what all the Democratic leaders are trying to do. They're already trying to use this whole Jensen thing to basically spin this narrative that the Trump administration has been insufficiently tough on China. Now, this is an extremely hypocritical narrative in a number of ways. First of all, many of these politicians were very much in favor of an engagement policy towards China, which ended up exporting large numbers of our jobs and industry over to China. It was Trump who in the first term really caused a revision of that doctrine. We moved from engagement to a much more standoffish policy towards China. So that's point number one. Point number two is these Democrats are the leading people who are opposed to data centers. In fact, we just got a story about how in Maryland there was an abandoned shopping mall where a data center was gonna go, and a local activist basically managed to kill that data center. Not totally surprising. David Sacks (00:50:53): However, she credited the viral success of her campaign to the Party for Socialism and Liberation, which is a China proxy organization operating here in The US. It's called the PSL, and according to this tweet that I'm putting on the screen, it is the political arm of Shanghai based Neville Roy Singham, who is an avowed Marxist multimillionaire who's bankrolled anti American pro CCP nonprofits to the tune of $278,000,000. And Neville Royce Singham's political foot soldiers have been boots on the ground here in The US opposing dozens of data centers, and these efforts are most aligned with the Democrat Party. So if the Democrats are serious about opposing Chinese influence, I think we should conduct an investigation and have hearings on whether the CCP is actually involved in these anti data center efforts across the country. You don't hear Democrat leaders trying to say that. So I'm pointing out basically their hypocrisy on this issue. Let me go back to the the issue of chips for a second here. If congress now believes that we should not sell lagging chips to China, my view on that is fine. Let's make that the policy. I don't think the juice is worth the squeeze. I think there were some good policy arguments for why you might wanna take market share from Huawei with a deprecated chip, a much less powerful chip. If you could take market share away from Huawei, those r and d dollars would flow to American companies instead of to a Chinese company and it would slow down the scale up of Huawei because chips are a scale up business. So that was the original reason why deprecated chips were still allowed to be sold into China, and that was a logic that the Biden administration came up with, and it was continued in the Trump administration. And like I said, there's now an effort to try and retcon anyone who wanted to sell deprecated chips into China as somehow being soft on China and doing China's bidding. Look. David Sacks (00:53:01): That was a bipartisan policy. If congress now wants to change that policy and say we're not gonna sell anything to China, fine. Do that. Again, I think that, you know, given our political climate, I think it's too hard to have that nuanced policy. By the way, here's the most surprising thing about the whole thing. China doesn't even want the chips. David Friedberg (00:53:22): Didn't they put a ban on their gaming David Sacks (00:53:24): the h 20 chip. And then when the h 200 got approved, they they basically have prevented those chips from entering the country, and they announced other Nvidia SKUs that they're against. So China doesn't even want the chips. This is a thing that I think a lot of people in Washington haven't wrestled with is that the policy of the China hawks is effectively the policy of the CCP. Neither one wants American chips in China. Okay? Now they want that policy for different reasons. The China hawks think it's gonna help China to have American chips there, and the CCP wants to keep NVIDIA chips out of China because they're trying to prop up Huawei as their national champion. So they want to force the Chinese market into the arms of Huawei so that Huawei could scale up and then compete all over the world. So, again, it's kinda crazy that people think that the CCP is begging for these chips. I think that Chinese companies want them. Chinese developers want them. But the Chinese government, based on all their public statements and their policies, is trying to keep out those American chips. Jason Calacanis (00:54:29): Clearly, the companies making LLMs covet the latest chips right now. Chinese government covets building their own chip champions internally. So two things can be true at the same time. It's a very subtle point. David Sacks (00:54:42): They're trying to indigenize chip production. They think it's a critical industry. In the same way that we're trying to get rare earths out of their clutches and we're trying to get our own rare earth supply, they're trying to create their own chip supply. They're trying to scale up domestic chip manufacturers. And they're willing to endure some short term pain in their ecosystem to prop up Huawei to force all of their local companies, their developers to try and use Huawei. And if you look at recent press, the latest DeepSeek model is being trained exclusively on Huawei chips. So in any event, there's like a big disconnect here where I think the China hawks in DC think that the Chinese government is begging for their chips, and the truth is that they would love to keep the chips out too, which is why I say, again, the juice isn't worth the squeeze. The China hawks don't want the chips sold. The Chinese government doesn't want the chips sold. So just forget about it. Stop selling them. I mean, is the crazy thing is that if Jensen was taken out because he wanted to sell chips to China, it was the the stupidest, most moot hill to die on because China's not taking those chips anyway. That's what's so crazy about it. David Friedberg (00:55:50): And you guys don't think that there's some pretty significant potential that Jensen knew that there was violations happening of these chips making their way to China and he was allowing it to happen through Mega Speed and perhaps other companies and that's what really brought him down? He doesn't need more... It's a great question. Does he need more money? Does he need incremental sales? Does he need the stock price to be higher? Does he need to... Hold on. Hold on.... 6% of revenue comes from China, then yes. Jason Calacanis (00:56:17): Well, I don't think it's 46, but even still, what more do you need for him personally or for the company than being the largest company in the world ever created by humanity? And why would he risk doing something that is treasonous that would result in him going to jail? David Friedberg (00:56:32): I will tell you why. Because Jensen's Jensen's philosophy, like I said before, is that AI is a tool for humanity. I know that he's America first. We know the guy. We know that he's patriotic. He is an American patriot. But I do also believe that there's a philosophy that a core group of technologists hold, that is that technology progresses humanity. It's not about progressing one side over the other, that technology brings us all up. And as more people in the world are brought up, there is less conflict, there is more abundance, there's less reasons to hate each other, There's less resources to fight over and so on and so forth. And that's a beautiful view of the future. It's a beautiful view of what technology can enable. So if you hold that core and true in your heart, perhaps there's a way that you unconsciously kind of rationalize your way into saying, hey, look, I'm following the rules of the book. The rules of the book say I consulted Vietnam and Singapore and whatever, and I'm doing exactly that. But if you then find out that those things are making their way over to China, but it's not violating the rules of the book, but you know that it's happening, that's where things start to get perhaps a little bit Yeah. Blurry, a little bit I don't know. I mean, look, I'm just... David Sacks (00:57:45): You're speculating there, Freeburg. And what I would say is, look, my framing at the very top here was that there's two ops, right? There's whatever Alice Wong and Mega Speed were doing, which I said should be investigated. And then there's whatever the leakers were up to who broke the law in order to leak these things, details about Jensen. With respect to the first op, we don't know what happened. There should be a full investigation. I mean, look, if Jensen and NVIDIA broke the law, that's really bad and there should be consequences for that. Jason Calacanis (00:58:14): Serious consequences. David Sacks (00:58:14): I'm definitely not defending that at all if that turns out to be true. But I also think that NVIDIA has denied it and we should just wait to see what the facts are there. David Friedberg (00:58:24): Okay. So let let's ask who who should be the next CEO and what's gonna happen? I mean, look, there's some Chamath Palihapitiya (00:58:29): incredibly brilliant technical people, Samir, Ian Bach, inside of NVIDIA. I think Ian's the most obvious internal candidate, and the most obvious external candidate is Rene Haas from Arm and Nikesh Arora from Palo Alto Networks. Rene, because he worked at NVIDIA for many, many years, worked directly for Jensen and has been running ARM successfully, so he's steeped in chips. And then Nikesh, because he's been the best capital allocator of the last five or ten years in the public markets, a huge surface area of AI is going to be around security. So he brings that to the table. I would be pro Nikesh, pro Renee, pro Ian Buck because I think what you get are technical, aggressive, product oriented capital allocators. I think the thing that you don't want is a steward, and the reason is I do not think that this is a business that is just going to David Friedberg (00:59:25): Roller coaster, it's not gonna just Chamath Palihapitiya (00:59:27): go, yeah. Be milked like a fatted cow, And so I think if you treat it as a business where you don't take risk, you are going to lose. David Friedberg (00:59:34): It's a wartime company right now, right? This is a dynamic company. Chamath Palihapitiya (00:59:38): Yeah. You need a David Friedberg (00:59:39): wartime Chamath Palihapitiya (00:59:40): CEO. David Friedberg (00:59:41): Here's my speculation. If it is found to be the case that Jensen did not violate any laws, I would argue that Jensen should return as CEO. And I think that any violation of code of conduct, any failure to make disclosures or attestations to the board as he should have done, Perhaps in this particular context with this particular business and this particular individual, the shareholders can make a decision to see past all that and bring him back if he didn't violate the law. If he broke the company code of conduct and so on, you know, the shareholders can make their choice about whether it's appropriate or not. But I think there's obviously the question of who's the best CEO to run NVIDIA, you know, is probably Jensen. I'm with you. I think Jason Calacanis (01:00:24): based on what I've seen, Jensen's gonna come back as chairman or CEO. I think he's gonna be exonerated. That's what I believe slash hope happens. Other two possibilities that people have discussed often, SpaceX brings Jensen on as the CEO of the TerraFab business, works alongside Elon, maybe even as like peers in the company. Or... Yeah. Nikesh is awesome. Why not have NVIDIA buy Palo Alto Networks? They need security. Security is a key issue, cyber, in this space and just Nvidia can easily afford to buy Palo Alto. You put in a cash in that suit, those seem like three reasonable solutions. I don't know if you wanna speculate, Sax, before we move on to the Anthropic IPO. David Sacks (01:01:06): I would just say that I hope that Jensen ultimately is treated at least as well as Eric Swalwell Jason Calacanis (01:01:13): Okay. David Sacks (01:01:14): In this whole matter. Jason Calacanis (01:01:15): Well said, yeah. David Sacks (01:01:16): But yeah, look, I think I think Friedberg makes a good point, which is that what is the crime here that he's been even accused of? Now, we don't have all the facts. Things may come out. David Friedberg (01:01:26): Right. David Sacks (01:01:26): So let's see how this all plays out. By the way, just one thing I wanna say if you're asking me, like, what I think about Jensen. You know, I see all this fake news where somehow they think like I had a long relationship with Jensen or whatever. I never knew Jensen before the administration. The first time I ever met him was in the White House. I've got all these reporters thinking that somehow we had this like long standing friendship going back before I was in the White House or somehow that I introduced him to Trump or stuff like that. Not true. I met him when he was already in the White House for a meeting anyway. So my point is I have no special devotion or loyalty to Jensen, but I respect him as a founder and I see his unique value as a founder to Nvidia and therefore to The United States. And if it's found out that he has never broken the law, then I think you guys are right that like what what are we doing here? Jason Calacanis (01:02:16): It makes no sense. Yeah. And we we again, you know, we're sitting here David Friedberg (01:02:21): And by the way, board might have no choice. Right? Because their fiduciary duty, which means they don't wanna get sued, they don't wanna lose D and O insurance, all the things that happen at the mentioned it earlier. Correct. You have to action and so on. Perhaps if Jensen comes back, he gets a founder of Chamath Palihapitiya (01:02:34): The shareholder lawsuits here are going to be ginormous. The company has lost almost $1,000,000,000,000 of value in about nine days, and so there's a lot of harm here that has to be managed somehow. David Friedberg (01:02:45): Yeah. Chamath Palihapitiya (01:02:45): So this is where even if he's found to have not done anything bad, the situation becomes very complicated because of all of this other traditional infrastructure. Like, the SEC is gonna have a point of view, the auditors will have a point of view, so it's just a lot to navigate. David Sacks (01:03:02): Just on that question, you know, there is an active investigation, I believe, the Justice Department that's been been reported. So we have to see what that ends up. Like I've said, I advocate full investigation. We need to get the facts here. Transparency. Again, if he's done something wrong, then he's gonna have to deal with the consequences. But until that point, I mean, we have in this country a very sacrosanct principle that you're innocent till proven guilty. And I'll just say that what's happened here apparently is that people who are hostile to Jensen, who were briefed on this memo, broke the law in order to leak it, in order to take him down. And that is not the way we should be conducting investigations in this country. The Justice Department should be investigating this matter. They should be the ones determining whether he did something wrong. Speaker 5 (01:03:51): Totally right. David Sacks (01:03:52): And this should not be tried in the court of public opinion Speaker 5 (01:03:55): 100%. David Sacks (01:03:55): Using innuendo and lurid allegations. And this is always what people go to as a weapon, is sex is very embarrassing to people, and in politics, it gets used as a weapon. Jason Calacanis (01:04:08): It's disgraceful, yeah. David Sacks (01:04:10): And it's so obvious that that's what has happened here. Chamath Palihapitiya (01:04:13): I understand what you're saying. Let me just say something slightly different. You're absolutely right that there's no shred of thought that there's been any criminality. And so that needs to be thoroughly investigated in every aspect of what's happened. The leaker clearly broke the law, but there's no sense that there's been any criminality by Nvidia or Jensen. That needs to get adjudicated properly and fairly for him and for them and for the company they've built. But what is unfortunately undeniable are the civil claims which are already real. Which is you've seen $1,500,000,000,000 of value just wiped out. And at a minimum, there was some disclosure issues. There were some problematic issues. There was the Nvidia PR people that were saying that this didn't happen, then it did happen. So at some point somebody was misleading somebody and all of that has ultimately resulted in actual harm that can be measured monetarily. That's where a lot of the action is going to focus on because that's where people will have dollar signs in their eyes, sadly, in The United States. You'll have the plaintiffs' folks, the attorneys, going absolutely nuts. They'll be licking their chops thinking, I'm gonna sue for a trillion and $0.5 of damage and I'm gonna be willing to settle for $250,000,000,000 and I'm gonna take half of it. That's what the trial lawyers are gonna be thinking here. And so there's gonna be all kinds of issues around that, David, where I think, unfortunately, that damage has been done. Those those eggs cannot be unscrambled. Jason Calacanis (01:05:47): I think there is one tiny possibility that we'll just put out there or I'll put out there. The leak would be completely valid. It would be completely valid to break the law and to leak this if this was for the betterment of society. In other words, Pentagon Papers, if you remember the movie, The Insider, the Brown and Williamson's tobacco documents, there are examples of virtuous leaks. I don't think that's the case here, but you can say there is a small, small chance that there could be a leak that would be for the benefit of society, and that would be a virtuous whistleblower. Alright. Anthropics IPO roadshow is underway. They've started to temper expectations a bit according to Reuters. Propix is gonna sell 825,000,000 shares, 70 to $78 a share. They'll put the value of the company at 1,400,000,000,000. That's below reports of 2,000,000,000,003 trillion. There was a lot of excitement about the company a couple of months ago. Total raise, 90,000,000,000. 64,000,000,000 in share sales and 26,000,000,000 in a private placement. We've talked about that before. A lot of their existing investors and partners are investing. It gives them a giant war chest as you pointed out in a previous episode. Seven founders plan to sell about $250,000,000 each. That still leaves each with a stake of about 18,000,000,000. So there's seven founders with basically $20,000,000,000 stakes. They pledged to donate 80% of their holdings to charity. However, I don't think that that is legally binding at this point. It's a pledge. Business is booming. Run rate, 81,000,000,000 at the end of October. Q three revenue was 17,000,000,000 against 10,000,000,000 for OpenAI. So they incredibly passed OpenAI in revenue by going enterprise. Obviously, OpenAI trying to catch them up... Catch up with them, but they're largely consumer driven. Great overview of Anthropix IPO here from our partner, Alphasense. Jason Calacanis (01:07:47): Anthropix target IPO valuation, 46% premium on its last private round, but they've doubled their run rate since then. So perhaps they are filling in the valuation they've had that's been pretty high for a long time, as we say in the business filling in that valuation. Analysts have Anthropix bull case at 4,200,000,000,000 valuation 15 times. It's 2028 expected revenue of 281,000,000,000. That would be the second largest company in the world ahead of NVIDIA. At 1,400,000,000,000, that's a 17 x multiple current run rate. Gentlemen, what are your thoughts here on David Sacks (01:08:27): the Jason Calacanis (01:08:27): IPO? Chamathi, you've, I think correctly threw a little bit of cold water on and said it might be a little bit less in terms of valuation, but they're raising a lot of money and they have that war chest as you pointed out in the previous episodes. Chamath Palihapitiya (01:08:41): I think it's great that they're gonna have somewhere between 230 and $300,000,000,000. I think that's very important for them to weather the storm. What is the storm? The storm is whatever's gonna happen over the next few weeks and months, and then what the economy does in reaction to it. I think the reality is that this David Friedberg (01:09:01): IPO Chamath Palihapitiya (01:09:02): is crucial. It has to happen. The issue that I have is that 20% of the value of Anthropic will be because of Anthropic, and that's unfortunate because I think they deserve to be more in control. 80% has to do with the broad based economy. That has a 100% to do with AI and that has a 100% to do with NVIDIA. And so I think as you see the impacts of the Bank of NVIDIA being closed, the impact of these investigations. The market will have to digest those and decide is the balance of risks now passed or are there more shoes to drop? If there are more shoes to drop and it creates a scepter of more risk, think people will be risk off. And in that, they'll be much more tempered in buying into any IPO, including Anthropix. Ants will do fine. They're six times oversubscribed already. It's gonna be fine. The question is what is the real price? And I think the real price will be more dictated by what gets leaked between now and November 20 more than anything else. Jason Calacanis (01:10:06): We have a lot of headwinds on what's happening in the AI trade, the so called AI trade. NVIDIA is but the latest, and we've seen haircut after haircut, the blue wave maybe slowing down data center deployment, lots of headwinds, but consumers are embracing this technology especially in the enterprise, and nobody has captured enterprise AI AI more than anthropic. Are you bullish on the anthropic IPO? And are you bullish writ large at the AI movement here in the economy? David Sacks (01:10:41): The AI movement. Jason Calacanis (01:10:42): The AI trade. Are you bullish on the AI trade? Let me say it that way. David Friedberg (01:10:46): I I don't know. I've not studied the numbers. I don't have a point of view on valuation. I think the macro on Anthropic that I've been hearing from folks that have attended some of the management conversations is that there may be enough people that align with the view that Anthropic is building this recursive self improvement engine and that their model capabilities are accelerating, that their compute cost per unit of intelligence rather than per token, which I think is the way a lot of people think about it or per watt is actually declining and that they do believe that with the investments that they've made and are making in all levels of infrastructure, including all the way down to the chip layer, they think that they have this kind of runaway business and that this could be a winner take all scenario of the biggest market ever. So it is the kind of once in a generational opportunity if you believe that there's a chance that that could be true to allocate some percentage of your portfolio to this IPO regardless of where the market bids it. So I don't think people are gonna be sitting here trying to come up with a particular view. I think they're gonna get all these... What did they say? 1.35, 1.5? They already said 60,000,000,000. My guess is they raised a 100,000,000,000 at one seven five because I think there'll be oversubscription to this thing. I think it'll get bit up. I think they'll wanna make sure it can run a little bit. From what I'm hearing, there's enough people that believe that there's a high enough probability that the story could be true that they'll bid it. So let's see what happens. I mean, I'm totally speculating here. I've spent zero time looking at this, so there you go. That's my free speculative waste of your five minutes. Jason Calacanis (01:12:35): Okay. I have two takes on this right now. David Friedberg (01:12:40): We could also we could also cut all that. Jason Calacanis (01:12:42): No. No. I think it's David Friedberg (01:12:43): I don't Jason Calacanis (01:12:43): know David Friedberg (01:12:43): about Anthropic. Jason Calacanis (01:12:44): It's like... Yeah. Yeah. I I think Anthropic is building the Microsoft killer and every SaaS company killer by bundling. We've talked about this, Sax, many times. Microsoft was the bundler. Is that antitrust or not when they product dump a new product for free? Obviously, Office Meetings would be the example of that, you know, going after Slack, etcetera. In the last month, Claude, the product that most people know Anthropic by, has dropped Claude meetings. And this is where you can pop up a whiteboard and you can do a Zoom. You can literally... They launched today, an hour ago, a Zoom like product, a Zoom killer, if you will. Couple of days ago, on November 10, they released whiteboards and sheets, an Excel killer. Right before that, they released personal finance for consumers. Then just before that, they released Claude Legal on October 28, and they keep releasing product after product. I think they are doing the ultimate bundle. Every single piece of software that you need, and they previously did David Friedberg (01:13:59): code, Jason Calacanis (01:14:00): co work, design, and this new Grockbot, Muse competitor contemporary called Duet, which I just got. It's not very good right now, but I assume they keep doing it. I think they're going after all enterprise software, all the business use cases. I think they have a massive amount of market share that they can take, and this product is getting stickier and stickier and stickier. And it gets rid of a ton of problems in SaaS... SACS, which is the interoperability of these and getting data from one to the other. I think this is gonna be the largest company in the world potentially if it continues at this pace where they can release new products and services and just attach them to your 20 to $50 a month subscription. I think this company could even be massively undervalued at this point. What do you think, Sax? David Sacks (01:14:53): Well, I said on a previous podcast when you guys thought that Anthropic was gonna lose to open source that actually they were creating the most valuable monopoly in the world. Jason Calacanis (01:15:03): And David Sacks (01:15:04): now they've all but confirmed it. I mean, they're running around saying that this is a winner take all market. What does that sound like? Monopoly? They're not even saying that it's gonna be a duopoly with Anthropic and OpenAI. They're saying they're gonna win the whole thing. They definitely are saying that recursive self improvement has kicked in despite the obvious safety problems with that. I mean, recursive self improvement's the most dangerous thing anybody can be working on in SI if you're worried about rapid takeoff, for example. If that's your greatest fear then, I mean, they are racing towards that. But recursive self improvement would be the key to getting to true next level superintelligence and they're treating it like a business advantage. I've always applauded them as an investor. Wearing my hat as an investor, I've always said that this is a valuable company that's done an amazing job. I've never quite understood why all the safety groups love them so much, aside from the fact that they provide funding for them. Because as a company, they're the ones taking the greatest risk. I'm not saying they're doing things in the riskiest way, I'm saying that they are pushing the frontier faster and faster and they are doing more recursive self improvement than anybody else. And yeah, they publish a blog saying, well, we would pace the frontier if everybody else did it. Come on, that's lip service. They know when they publish that that they're not going to get the whole world to magically agree to pace. But they are the ones doing it. You know, again, I said at the time, if you think that this type of development is risky, stop doing it. But they have not stopped. They're racing towards it and now they're marketing it in their roadshow. Right? They're saying, hey, you should buy our stock because we're doing recursive self improvement, We're creating, you know, AGI, and it's gonna be a winner take all game. Jason Calacanis (01:16:54): Yeah. And they're also David Sacks (01:16:55): It's it's just amazing to me in a way, Chamath Palihapitiya (01:16:57): like, David Sacks (01:16:57): what this company gets away with. I mean, like, again, I think it's a phenomenal company, but they say shit that would have made Bill Gates blush when he was running Microsoft and he was the most Jason Calacanis (01:17:11): Justice Department investigation. Yeah. He was the David Sacks (01:17:13): most vicious monopolist that's ever lived. Jason Calacanis (01:17:15): The level they're dropping just pays attention to this. New competitors into the market makes me believe that Microsoft has to create a competing product and go head to head against them or they're gonna lose the Office franchise. There was one thing of note, Freiberg, that I found inspiring and troubling at the same time. Dario made a claim that he believes they're gonna cure cancer, Alzheimer's, all diseases within ten years. Play the clip, Nick. David Sacks (01:17:44): Today, we look at things like cancer, Alzheimer's, heart disease, mental illness. These will be relics of the past in, I think, less than a decade. Jason Calacanis (01:17:57): Freeberg, will people be mentally ill in 2036? Will cancer be solved by 2036 as well as Alzheimer's and heart disease? Is this all gonna be relics of the past, or do you think this is? David Friedberg (01:18:10): Jason, I I don't know how to answer that. Like, look. I mean, we we know that the technology is extraordinary. It is completely reinventing how we can do molecular discovery, how we can do in silico testing, how we can accelerate path to market. Those things are all true. The implications on timelines, you know, it's fuzzy math. What I see in my work, I work in biology, there's an incredible set of breakthroughs every week happening right now in biology with these tools. It is not a joke. It is not made up. It is not some far flung idea. It is not something that is ten or fifteen years out, it is happening today. So yes, I do believe that there is extraordinary advantage coming here. I will tell you we spend a lot of money on Anthropic. I've mentioned in the past we've built our own open source setup on our own racks, but for a lot of the heavy work that we do in biology, it goes to anthropic. They are the best. David Sacks (01:18:59): Any Jason Calacanis (01:19:00): thoughts there, Tramath, on that claim? You think it's credible that these things could be cured in under ten years? Any of those things? Chamath Palihapitiya (01:19:08): I think we're gonna be in a recession by 2028. Jason Calacanis (01:19:11): Okay. There it is, folks. Alright. I I, you know, I saw that and I just thought, wow. It's... I I think he believes what he said. I think it's a little reckless to say something like that. Like, to even say that. Chamath Palihapitiya (01:19:24): I think I think there's gonna be a recession. I mean, like What the Fair enough. Like, you know, show the Jenga block thing. David Sacks (01:19:31): You know? Alright. Jason Calacanis (01:19:34): We got a new doctor Doom here. Get him the doctor Doom mask. Chamath Palihapitiya (01:19:37): 300 basis points of GDP may have just gotten obliterated this morning. I mean, it's just Jason Calacanis (01:19:42): to in post put the doctor Doom mask on to my Piece this week's doctor Doom. Chamath Palihapitiya (01:19:46): Just so sad. David Sacks (01:19:47): Just so sad. The way, this this is a really, I I think, insightful tweet that someone just posted. Can we pull this up on the screen? Jason Calacanis (01:19:54): Okay. Insightful tweet. David Sacks (01:19:55): Jake, are you gonna read this or you wanna read it? Jason Calacanis (01:19:57): Well, I mean, it's from Will Manidis. The murmurs I get fed from the deep rabbit warrens of the safety as group chats and the doom pilled iron guns deep in the district is that they are in full jubilee mode today. The killing of the old king is complete. They are only days away from taking their seat at the throne. Okay. Game of Thrones, folk. David Friedberg (01:20:22): Well, I mean, let's face it. Who are referring to, Saxe? Like, who are those? Chamath Palihapitiya (01:20:25): Who are the safetists and David Friedberg (01:20:27): the Yeah. Who who's gonna take the thrones? David Sacks (01:20:29): This is the crazy thing is that all the safetist groups, the EAs, are deeply aligned with Anthropic who's been funding all of them for years. And Anthropic's about to have this huge IPO. And then meanwhile, the main funder of the open source movement, which is NVIDIA, just lost their CEO and founder who's just their unique force multiplier and asset, their superstar player. So, I mean, obviously, who is gonna defend open source now, never mind finance it, with Jensen gone? So yeah, I can see why they're jubilant. But look, here's the crazy thing. If you're concerned about AI safety, how is any of this leading to more safety? Let me tell you what it's doing. It's concentrating the market. I mean, Anthropic itself is saying winner take all. They're gonna be Last Jason Calacanis (01:21:15): company in the world. David Sacks (01:21:16): The last company in the world. They're gonna have recursive self improvement. They're creating the superintelligence. Look at the clock constitution. They're training it to be its own species. They say that it has emotions maybe, it has a moral code, that it should reject human instruction, it should be a conscientious objector. They act like a proud parent ushering a baby into the world, And they are very very close now to creating the singularity of the superintelligence. And the crazy thing is that all these safety groups think that that is their ally. I mean, like, are you stupid? These people are using you guys, if anything, because they're creating... Whatever this superintelligence is turning into, whatever happens with that, anthropic is at the frontier, is at the front of the frontier on that. So if you're some kind of doomer or ex risker or just someone who... A decel, I don't understand this alliance at all. I mean, you're being played like a useful idiot, I think. Look, Do I'm not a David Friedberg (01:22:17): this you stock? David Sacks (01:22:19): Well, I'm not a doomer, let me just say this. So I actually don't necessarily believe that what's gonna happen here is gonna be harmful for humanity. I think it could lead to the curing of all those diseases. I think it'll be very beneficial. But my point is that if you are a doomer and you're terrified of recursive self improvement, why in the world are you supporting this company, this IPO? Why are giving them such a pass? You know, why are you celebrating over Jensen's fall? All he was doing was promoting open source, which is manifestly behind the closed model frontier. Right? There's no recursive self improvement there. So the whole thing is very backwards. And what it suggests to me is that all these safety groups don't really care about safety. They care about something else. It's not the real agenda. If safety was a real agenda, they'd be way more concerned about the Clog Constitution training the superintelligence to be a new species. What they seem to want is, to your point, Freeberg, is extreme centralization of power in the hands of the government. They want all the models and all the chips to be controlled by the government. The government's gonna surveil you, spy on you. It'll have backdoors all over the place, and that's what they seem to want. It's a scary future. David Friedberg (01:23:29): Notice how no one in congress is like, oh, let's go after anthropic bailiffs. Wanna go after everyone else. Jason Calacanis (01:23:36): Yeah. David Sacks (01:23:36): Because anthropic is offering Jason Calacanis (01:23:38): them David Sacks (01:23:38): power, ultimate power. Jason Calacanis (01:23:39): It's gonna be great when we all are all invited to the White House, by AOC in 2028. It's great to see you, AOC. Alright, everybody. Another amazing episode of the Ireland Pocas. We'll see next time. Bye bye. Chamath Palihapitiya (01:23:53): Love you, boys. Bye bye. David Sacks (01:24:21): We Jason Calacanis (01:24:27): should all just get a room and just have one big huge orgy because they're all just useless. It's like it's like sexual tension that they just need to release somehow. Let

本集资源与出处