Nvidia, Wall Street Launch $500B AI Plan as Musk Eyes Pay Loophole

英伟达联手华尔街启动5000亿美元AI计划，马斯克瞄准薪酬漏洞 英伟达价值5000亿美元的基础设施方案利用折旧保险将GPU债务转化为投资级证券；与此同时，特斯拉一项隐蔽的控制权变更条款可能让埃隆·马斯克通过SpaceX收购获得1万亿美元回报。 主持人 John Coogan 与 Jordi Hays 深入分析了英伟达在华尔街顶尖金融领袖支持下推出的里程碑式5000亿美元AI基础设施融资方案，以及黄仁勋的风险缓解策略。他们详细讨论了派拉蒙CEO David Ellison 因华纳兄弟合并拖延而威胁将总部迁出加州的最后通牒，并探讨了埃隆·马斯克薪酬方案中一项隐蔽的控制权变更条款如何可能通过SpaceX与特斯拉的合并带来1万亿美元收益。此外，节目还评估了马克·扎克伯格领导下Meta的AI进展、Google DeepMind 的 Demis Hassabis、Bending Spoons 激进的收购模式，以及部分简要的企业与文化动态。 英伟达召集了包括高盛、黑石、阿波罗和贝莱德高管在内的华尔街顶尖金融领袖，共同支持规模达5000亿美元的巨额AI算力基础设施方案。 (0:02) 阿波罗估计，全球公开与私募资本市场在AI基础设施上的总支出最终将超过8万亿美元。 (0:02) 为了向GPU部署开放传统债务市场，英伟达提供高达25%的折旧保险，并提供1吉瓦数据中心的参考设计，将相关资产标准化为投资级债务。 (0:02) 由于收购华纳兄弟探索遭遇反垄断异议，派拉蒙CEO David Ellison 向加州发出最后通牒，威胁在10月1日前将派拉蒙总部迁出该州。 (7:46) 若华纳兄弟探索合并案推迟至2027年5月审理，可能会使派拉蒙面临累计约12亿美元的按日计算“计时费”违约金。 (7:46)

TBPN outlines major corporate and financial maneuvers across technology and media, highlighted by Nvidia assembling a Wall Street-backed $500 billion compute financing vehicle and Elon Musk holding an acquisition shortcut to a $1 trillion Tesla payout. Across industries, companies are using structured debt, regulatory ultimatums, and contract clauses to bypass operational bottlenecks and legal hurdles. Wall Street Backs Nvidia's Compute Financing Framework (0:02) Jensen Huang gathered leaders from Goldman Sachs, Blackstone, Apollo, Brookfield, and BlackRock to assemble a $500 billion financing package for artificial intelligence infrastructure. Apollo's leadership forecasts that total AI infrastructure spending will ultimately exceed $8 trillion, requiring $50 billion to $60 billion per gigawatt. Huang reported that $500 billion entered AI startups over the preceding six months alone, driving intense compute demand despite ongoing constraints in chips, memory, packaging, power, and construction capacity. To overcome bank reluctance toward fast-depreciating GPU collateral, Nvidia is offering up to 25% depreciation insurance to underwrite marginal loans. Nvidia is also supplying standardized reference designs for 1-gigawatt Blackwell data centers. This standardization allows lenders to package compute debt into asset-backed securities, collateralized loan obligations, and collateralized debt obligations with investment-grade ratings, enabling offloading to institutional pension funds and insurance firms. Paramount's California Relocation Threat Over Antitrust Standoff (7:46) Paramount CEO David Ellison has threatened to move the company's headquarters and operations out of California by October 1st if state Attorney General Rob Bonta refuses settlement negotiations regarding the proposed Warner Bros. Discovery acquisition. California and 12 other state attorneys general are suing to block the transaction on antitrust grounds, with Bonta insisting on structural remedies like asset divestitures rather than behavioral promises. The relocation threat is fueled by escalating financial pressure. Under the merger terms, Paramount faces a daily ticking fee of $7 million paid to Warner Bros. Discovery shareholders starting October 1st if the deal remains unclosed. With the antitrust trial scheduled for May 2, 2027, Paramount could accumulate roughly $1.2 billion in penalties. The Paramount-Skydance board has formally authorized relocation plans, with Georgia, Texas, and Tennessee under review as potential replacement sites. Musk's Change-of-Control Shortcut to $1 Trillion (11:28) A report highlighted an obscure mechanism in Elon Musk's shareholder-approved 2025 Tesla compensation plan that could grant him up to 423 million shares—valued between $824 billion and $1 trillion—if SpaceX acquires Tesla. Under standard performance terms, unlocking all 12 tranches requires reaching an $8.5 trillion market capitalization alongside demanding operational milestones: delivering 20 million vehicles, logging 10 million active Full Self-Driving subscriptions, deploying 1 million Optimus robots, and putting 1 million Robotaxis into service. However, the compensation agreement includes a change-of-control clause that waives every operational performance target in the event of an acquisition. In that scenario, stock awards are evaluated strictly against Tesla's implied valuation or the transaction price. If SpaceX were to acquire Tesla at an $8.5 trillion valuation, Musk could collect the full 12-tranche equity package immediately, bypassing years of operational benchmarks. Meta's AI Positioning and Frontier Industry Dynamics (13:35) Mark Zuckerberg published a 6,000-word manifesto defending open-source AI models ahead of anticipated model launches, conducting interview calls using Meta Smart Glasses. Meta Superintelligence Lab is currently viewed by observers as holding third place among frontier AI labs, ahead of Elon Musk's xAI and aiming to surpass Google's Gemini models with subsequent iterations of Llama. The program contrasts Zuckerberg's shifting communications between open and closed models with Anthropic CEO Dario Amodei's consistent anti-open-source posture. Meanwhile, Google DeepMind CEO Demis Hassabis was characterized as a uniquely effective advocate for the AI sector, able to address policy concerns regarding energy and surveillance using his scientific credentials and Nobel Prize background without the social media controversies tied to consumer tech platforms. Software Acquisitions, Executive Exits, and Conservation Campaigns In additional corporate developments, Italian software firm Bending Spoons reached a valuation of $33 billion to $34 billion—trading at a 265x price-to-earnings ratio—after adding $11 billion to its market capitalization following its acquisition of Airtable. In the executive sector, former Zillow Chief Operating Officer Jun Choo posted an abrupt, one-line LinkedIn departure notice without further explanation of his exit. Separately, actor Leonardo DiCaprio mobilized his 60 million Instagram followers to lobby the Chilean government to halt a power transmission line project. The campaign aims to protect the critically endangered spiny chest frog, whose wild population is estimated at approximately 1,000 individuals. The source demonstrates how major corporate actors are leaning on structural financial instruments, contractual loopholes, and legal threats to manage regulatory scrutiny and the unprecedented capital requirements of modern technology and media consolidation.

本期节目梳理了近期科技与商业领域的关键博弈：Nvidia联手华尔街资本构建5000亿美元算力融资机制以抵御GPU折旧风险，Paramount以撤出加州向反垄断监管施压，以及Elon Musk或借助控制权变更条款通过SpaceX与Tesla合并绕过运营考核以实现万亿美元薪酬。 Nvidia联手华尔街启动5000亿美元算力资产证券化 (0:02) 据《金融时报》报道及CNBC节目披露，Nvidia首席执行官Jensen Huang联合Goldman Sachs、Blackstone、Apollo、Brookfield及BlackRock等华尔街顶级机构，推出了规模达5000亿美元的AI算力融资计划。Apollo总裁估计，全行业最终的AI基础设施投资将超过8万亿美元。按当前每吉瓦算力需消耗500亿至600亿美元建设成本计算，该笔5000亿美元资金预计可支撑约10吉瓦的算力规模。Huang同时指出，供应链从芯片、封装到电力与土地均面临限制，但过去6个月内全球已有约5000亿美元资金流入AI初创公司，市场对算力的需求极度迫切。 为解决传统银行业因GPU折旧速度过快而不愿将其作为抵押品的难题，Nvidia推出了最高25%的折旧保险机制，以促使银行承销边际交易。此外，Nvidia正为1吉瓦规模的Blackwell数据中心提供标准化参考设计。通过将算力中心重构为可标准化的资产，相关债务可以被打包并转化为资产支持证券（ABS）、贷款担保证券（CLO）和债务担保证券（CDO），从而获取投资级信用评级，最终转售给养老基金和保险机构等长期资本。 Paramount以迁址通告施压加州反垄断审查 (7:46) Paramount首席执行官David Ellison针对加州司法部发起最后通牒，威胁若加州总检察长Rob Bonta拒绝展开和解谈判，公司将在10月1日前将运营中心及历史总部搬离加州。目前加州正联同12个州的司法部长提起诉讼，以反垄断为由阻止Paramount对Warner Bros. Discovery的收购。《Variety》报道称，Paramount-Skydance董事会已正式批准若谈判破裂便启动迁出加州的计划，潜在迁址地点包括Georgia、Texas和Tennessee。 该举动背后存在严峻的财务压力。根据并购协议，若交易未能在10月1日前完成，Paramount每天需向Warner Bros. Discovery股东支付700万美元的延期补偿费（ticking fee）。鉴于州反垄断审判定于2027年5月2日开庭，若诉讼拖延至判决出炉，Paramount累计支付的罚金可能超过12亿美元。Bonta坚持要求进行资产剥离等结构性补救措施，而Ellison意在通过放弃好莱坞标志性总部这一“极端手段”，迫使监管机构在法庭程序外提前达成和解。 Musk薪酬协议控制权条款或开启万亿美元套现捷径 (11:28) 《华尔街日报》披露了Tesla股东批准的2025年Elon Musk薪酬方案中隐藏的控制权变更机制。在常规路径下，Musk若想解锁全部12个批次、共计4.23亿股的股票期权（价值约8240亿至1万亿美元），必须实现多项极端苛刻的长期运营指标，其中包括Tesla市值达到8.5万亿美元、累计交付2000万辆汽车、拥有1000万个FSD全自动驾驶活跃订阅、部署100万台Optimus人形机器人以及投放100万辆商用Robotaxi。 然而，该协议包含一项控制权变更条款：若Tesla被另一实体（例如Musk旗下的SpaceX）收购，所有运营考核指标将被自动免除或视为达成，股权解锁将完全根据交易时的市值进行评估。这意味着，若SpaceX以8.5万亿美元的隐含估值收购Tesla，即便Optimus或交付量等运营目标未达成，Musk也能依据交易前市值或收购价的较高者直接解锁全部4.23亿股期权。这种架构为Musk在业务进展滞后时整合商业帝国并获取顶格薪酬提供了制度捷径。 Meta推进AI定位与科技商业动态速览 (13:35) Mark Zuckerberg在发表6000字AI开源宣言后积极对外发声，称Meta即将发布能力显著提升的下一代模型。讨论指出，业内传闻Meta超级智能实验室（MSL）目前在AI格局中稳居第三并领先于xAI。与Anthropic首席执行官Dario Amodei长期坚持的反开源和对华鹰派立场不同，Zuckerberg的沟通策略更为多变；而Google DeepMind的Demis Hassabis则以诺贝尔奖得主的科研形象，成为向政策监管者与公众沟通AI前景的标杆公关代表。 在其他商业动态方面，意大利科技公司Bending Spoons在收购Airtable后估值升至330亿至340亿美元，其市盈率高达265倍，自IPO以来上涨50%；Zillow前首席运营官Jun Choo在LinkedIn上发表了极度简短的离职声明引发外界猜测；演员Leonardo DiCaprio则利用其社交平台动员公众反对智利的一项输电线路工程，以保护仅存约1000只的极度濒危物种棘胸蛙。 本期内容突显了科技巨头在重资产算力证券化、并购监管对抗及高管薪酬结构设计上的深层财技。然而，Nvidia算力资产打包能否抵御长周期技术淘汰，Paramount的政治施压能否换取反垄断让步，以及SpaceX并购Tesla的设想是否存在严苛的监管与法律障碍，仍是留待市场检验的核心疑问。

## Background 本期内容由科技商业谈话节目 TBPN 于 2026 年 8 月 13 日发布在 YouTube 平台。该栏目为 Diet TBPN 精华版，聚焦当下硅谷与华尔街最受关注的重磅商业交锋，涵盖英伟达高达 5000 亿美元的算力融资计划、派拉蒙对抗加州反垄断监管的迁址威胁，以及马斯克借助 SpaceX 收购特斯拉绕道解锁万亿薪酬的资本内幕。 ## TL;DR 本期节目深入解析了英伟达携手华尔街顶级私募财团推出的 5000 亿美元 AI 基础设施融资模式，探讨了 GPU 资产证券化如何将算力投融资从 VC 风险投资转向商业地产逻辑。随后，主播们聚焦派拉蒙在合并受阻后向加利福尼亚州发出的撤离通牒，剖析了每日数百万美元延期罚金下的法律与商业博弈。节目还拆解了扎克伯格近期密集公关背后的 Meta 模型竞争格局，并深度揭秘了特斯拉 2025 年薪酬方案中深藏的“控制权变更”条款——一旦 SpaceX 实施吸收合并，马斯克即可豁免严苛的量产与运营指标直取万亿美元级股票期权。最后，节目轻松盘点了 Bending Spoons 的疯狂并购潮、Zillow 高管极简离职声明以及好莱坞影星小李子为智利濒危青蛙发声等热点趣闻。 ## Atmosphere & style 整期节目节奏紧凑、信息密度极高，两位主持人搭档默契，在辛辣讽刺与硬核商业金融拆解之间自如切换。从调侃华尔街大佬齐聚 CNBC 的“名利场”躁动，到戏谑扎克伯格一边钓鱼一边宣讲技术哲学的用力过猛，对谈洋溢着硅谷内行特有的敏锐与幽默。听众既能领略到数千亿美元资本博弈的惊心动魄，又能在片尾关于职场八卦与环保名流的欢快吐槽中感受到轻松惬意的脱口秀氛围。 ## Bulletin - [00:02](https://www.youtube.com/watch?v=I8tsEszPcBY&t=2s) 英伟达联手高盛、黑石等华尔街巨头推出 5000 亿美元 AI 融资计划，黄仁勋通过提供最高 25% 折旧保险与标准化参考设计，将 GPU 资产包装为 ABS 等低成本地产级债券，Apollo 预计未来整个 AI 基建将撬动超 8 万亿美元资本投入。 - [07:56](https://www.youtube.com/watch?v=I8tsEszPcBY&t=476s) 派拉蒙威胁在 10 月 1 日前全盘撤出加州以抗衡反垄断阻挠；若与华纳兄弟探索公司的并购案诉讼拖至 2027 年开庭，派拉蒙每日 700 万美元的延期费将累积高达约 12 亿美元。 - [11:28](https://www.youtube.com/watch?v=I8tsEszPcBY&t=688s) 马斯克 2025 年特斯拉薪酬对赌协议暗藏“控制权变更”条款，若 SpaceX 实施收购，半数严苛业绩考核将直接豁免，为马斯克解锁高达 8240 亿美元至 1 万亿美元的股票期权提供制度捷径。 - [13:35](https://www.youtube.com/watch?v=I8tsEszPcBY&t=815s) 扎克伯格发表 6000 字长文力推下一代前沿大模型，展现 Meta 在开源与模型研发上的强劲执行力；同时欧洲软件聚合巨头 Bending Spoons 凭借 265 倍市盈率收购 Airtable 后市值攀升至约 340 亿美元。 ## Key takeaways - [00:02](https://www.youtube.com/watch?v=I8tsEszPcBY&t=2s) 英伟达协同高盛、黑石、贝莱德等顶级华尔街机构推出 5000 亿美元 AI 基础设施融资方案，黄仁勋在 CNBC 传奇圆桌拆解这笔世纪资本布局。 - [01:35](https://www.youtube.com/watch?v=I8tsEszPcBY&t=95s) 黄仁勋指出算力供应链上下游面临物理限制但 Token 利润率极高，反驳市场对 AI 初创企业未盈利的担忧；5000 亿美元融资对应约 10 吉瓦算力，仅够维持头部前沿实验室一到两年的扩张消耗。 - [04:30](https://www.youtube.com/watch?v=I8tsEszPcBY&t=270s) 英伟达通过提供 25% 折旧保险并制定标准化机房参考设计，帮助投行将 GPU 算力债务打包为 ABS/CLO 证券，成功将客户融资成本从风投资本拉低至商业地产借贷水平。 - [07:56](https://www.youtube.com/watch?v=I8tsEszPcBY&t=476s) 派拉蒙 CEO David Ellison 启动“核选项”，向加州总检察长发出最后通牒，若拒绝就华纳兄弟并购案和解，派拉蒙将于 10 月 1 日前将洛杉矶总部及制片业务迁往得州或佐治亚州。 - [09:23](https://www.youtube.com/watch?v=I8tsEszPcBY&t=563s) 因反垄断诉讼悬而未决，派拉蒙自 10 月 1 日起每日需支付 700 万美元倒计时费，若案件拖至 2027 年开庭，累计罚金将高达约 12 亿美元。 - [11:28](https://www.youtube.com/watch?v=I8tsEszPcBY&t=688s) 《华尔街日报》披露马斯克若通过 SpaceX 发起对特斯拉的并购，将触发薪酬协议中的“控制权变更”豁免条款，规避 100 万辆 Robotaxi 与 Optimus 量产等严苛考核。 - [13:35](https://www.youtube.com/watch?v=I8tsEszPcBY&t=815s) 扎克伯格发表 6000 字 Pro-AI 哲学宣言并在钓鱼时向媒体吹风，意在为 Meta 即将推出的下一代大幅跃升的前沿大模型铺路，展现出超越部分闭源对手的研发势头。 - [21:02](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1262s) 特斯拉常规薪酬方案要求特斯拉市值冲刺至 8.5 万亿美元，并达成交付 2000 万辆汽车、1000 万 FSD 订阅、100 万台擎天柱机器人和 100 万辆商业运营 Robotaxi 的极端指标。 - [21:02](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1262s) 虽然擎天柱量产遥远，但特斯拉 FSD 在 HW4 芯片加持下驾驶平稳度已超越普通人类司机，Robotaxi 采取街边落客方案已逼近商业落地的技术临界点。 - [24:11](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1451s) 若触发控制权变更条款，所有运营指标作废并纯按交易估值结算，马斯克可拿满顶格 4.23 亿股奖励，潜在现值高达约 8240 亿美元。 - [27:39](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1659s) 软件聚合商 Bending Spoons 以 265 倍高市盈率接连收购老牌现金流 SaaS，宣布收购 Airtable 后市值暴涨 110 亿美元突破 340 亿美元。 - [27:39](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1659s) 房产平台 Zillow 首席运营官 Jun Choo 在 LinkedIn 发布极简冷峻的单句卸任声明，引发行业对其幕后内情的强烈猜测。 - [28:51](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1731s) 影星莱昂纳多·迪卡普里奥调动 6000 万 Instagram 粉丝向智利政府喊话，要求叫停威胁野生仅存千只刺胸蛙生存的拟建输电项目。 ## Detailed walkthrough ### [00:02](https://www.youtube.com/watch?v=I8tsEszPcBY&t=2s) 英伟达联手华尔街巨头推出 5000 亿美元 AI 融资计划 节目开场，主持人提到了登上《金融时报》（Financial Times）头条的轰动性新闻：华尔街各大机构与英伟达（Nvidia）联手推出了高达 5000 亿美元的 AI 融资方案，堪称具有里程碑意义的借贷计划。黄仁勋（Jensen Huang）集结了金融界的“全明星阵容”，并在 CNBC 上完成了一场被主持人称为“史上最精彩之一”的传奇圆桌访谈。除了贝莱德（BlackRock）的 Larry Fink 线上参会外，高盛（Goldman Sachs）CEO David Solomon、黑石（Blackstone）的 Jon Gray、阿波罗全球管理（Apollo Global Management）的 Jim Zelter 以及 Brookfield 的 Bruce Flatt 全部亲临演播室，由黄仁勋亲自向全场拆解这场世纪资本布局。 主持人打趣称，演播室外的好莱坞名流们恐怕都快把玻璃敲碎了，想在这群每天在市场上“孤注一掷”的顶尖资本家、金融“坏小子”（bad boys）身上寻找新目标。根据 Apollo 总裁的预测，未来预计将有超过 8 万亿美元的资本投入 AI 基础设施建设，这一惊人规模促使私有资本与公共资本全面加速进场，华尔街丝毫没有减速踩刹车的迹象。 ```mermaid flowchart TD accTitle: Nvidia $500B AI Infrastructure Financing Structure accDescr: Diagram illustrating how private equity and Wall Street firms partner with Nvidia to fund AI compute infrastructure for labs and startups. subgraph WallStreet["华尔街金融财团"] GS["高盛 Goldman Sachs"] BX["黑石 Blackstone"] AP["阿波罗 Apollo"] BK["贝莱德 BlackRock / Brookfield"] end subgraph NvidiaRole["英伟达战略支撑"] NV["Nvidia 硬件与生态架构"] INS["折旧保险与参考设计标准化"] end subgraph Assets["资产与基础设施"] DC["10 吉瓦级 AI 工厂 / 数据中心"] ABS["ABS / CLO / CDO 证券化债权产品"] end subgraph EndUsers["终端算力需求方"] Labs["AI 实验室与前沿初创企业<br/>(OpenAI, Anthropic 等)"] NeoCloud["SpaceX / 新型云服务商"] end WallStreet -->|"5000 亿美元债务与股权融资"| DC NvidiaRole -->|"提供芯片架构与信用担保"| WallStreet DC -->|"算力租赁与 Token 生成"| Labs DC -->|"资产打包分层分级"| ABS NeoCloud -.->|"数据中心承建与算力外包"| DC ``` ### [01:35](https://www.youtube.com/watch?v=I8tsEszPcBY&t=95s) 算力供应链全线告急与“不可思议的 Token 利润率” 在 CNBC 的访谈片段中，黄仁勋被问及对 AI 算力需求是否过热或无法满足的担忧。黄仁勋明确表示，供应链在相当长一段时间内仍将全线紧绷——从芯片、内存、先进封装、系统组件、光通信器件、连接器，到土地储备、电力供应以及现场建筑工人，上下游整个链条都面临物理限制。但他坚信这一进程无可阻挡，因为 AI 正在全球范围内产生真实的生产力，并且“AI Token 的利润率高得令人难以置信”（AI tokens are profitable, incredibly profitable）。 面对“未盈利初创公司是否难以吸引公共资本”的质疑，黄仁勋强硬反驳，宣称在短短几个月内市场就会看清这些 AI 实验室是极其赚钱的，它们是“科技史上成长速度最快的企业”。在过去六个月中，全球向 AI 初创企业注资约 5000 亿美元，创下了现代商业史上的资本投资峰值。针对这笔 5000 亿美元的融资体量，节目主播分析指出，按照目前每吉瓦（GW）500 亿至 600 亿美元的数据中心建设成本计算，5000 亿美元大致对应 10 吉瓦的算力规模。考虑到目前各大头部 AI 实验室合计拥有的算力也不过 3 吉瓦左右，且 Meta 一家就已提出 10 吉瓦的扩张蓝图，按现有的 3 倍算力扩展定律（3x scaling），这笔天量资金实际上也仅够支撑接下来一两年的算力消耗。 - **华尔街圆桌阵容**：Jensen Huang（Nvidia）、David Solomon（Goldman Sachs）、Jon Gray（Blackstone）、Jim Zelter（Apollo）、Bruce Flatt（Brookfield）、Larry Fink（BlackRock，远程）。 - **基建投资预期**：Apollo 预计整个 AI 基础设施最终将撬动超过 8 万亿美元的资本开支。 - **算力换算成本**：数据中心建造成本约为每吉瓦 500 亿至 600 亿美元，5000 亿美元融资对应约 10 吉瓦算力容量。 - **初创融资纪录**：过去半年全球 AI 创业生态吸纳了约 5000 亿美元资金。 ### [04:30](https://www.youtube.com/watch?v=I8tsEszPcBY&t=270s) 从风险投资走向房地产逻辑：GPU 资产证券化大戏 两位主持人进一步探讨了算力生态的隐忧与深层资本结构。当前推动公开股票市场前行的两大 AI 领头羊实际上仍是私有公司，其真实财务细节只能靠零星泄露来拼凑。与此同​​时，SpaceX 也在通过接入英伟达硬件向新型云（Neocloud）转型，在供应链中扮演重要承建者角色，为 xAI Colossus 等大型集群提供服务。 黄仁勋在社交平台 X 上发表了一篇获得破万点赞的深度长文，强调“英伟达计算不仅仅是芯片，而是一整套 AI 工厂平台，涵盖加速计算、网络架构、系统软件、AI 框架与全球开发者生态”。分析人士指出，黄仁勋正在完成资本维度的“闭环”： 传统银行一直抗拒将 GPU 用作贷款抵押品，核心原因在于新一代芯片的代际淘汰会导致资产发生不可预测的残值贬值。黄仁勋凭借掌握英伟达产品技术路线图的优势，直接向银行提供最高达 25% 的“折旧保险”（depreciation insurance），帮金融机构消除坏账顾虑。同时，英伟达为银行制定标准化、可互换（fungible）的数据中心参考设计指南，使数据中心能像标准化写字楼一样被风险评估与归类。 一旦数据中心资产实现标准化与同质化，华尔街投行即可将底层算力债务打包为资产支持证券（ABS）、贷款抵押凭证（CLO）甚至类似 2008 年的抵押债务凭证（CDO）。通过分级切块（tranching），这些高风险债务可以变身为能够出售给养老金和保险基金的投资级债券，将个别项目的特质信用风险转化为全行业的信用风险。黄仁勋此举的实质，是将英伟达客户的融资成本从高昂的“风投资本股权”（venture equity）拉低到平价的“商业地产融资”（real estate financing）水平，正式将数据中心投融资从 VC 游戏带入了顶级机构主导的深水区。 ### [07:56](https://www.youtube.com/watch?v=I8tsEszPcBY&t=476s) 派拉蒙威胁搬离加州：与州总检察长反垄断僵局白热化 节目的第二个重磅话题转向好莱坞与传媒行业：派拉蒙（Paramount）向加州政府发出最后通牒，威胁若加州拒绝就华纳兄弟探索公司（Warner Bros. Discovery）并购案的诉讼展开和解谈判，派拉蒙将在 10 月 1 日前将业务全盘撤出加利福尼亚州。 根据《Variety》的独家报道，派拉蒙 CEO David Ellison 极力寻求快速解决由加州等 12 个州总检察长联手提起的反垄断诉讼。Ellison 已向公司高层通报，由于加州总检察长 Rob Bonta 成为阻击这桩巨额并购的带头人，Paramount-Skydance 董事会已正式批准了搬迁方案。如果 Bonta 拒绝在 8 月 1 日前开启实质性谈判，派拉蒙将启动撤离程序，其位于洛杉矶的历史性总部最早将于 10 月迁出加州。 这一威胁让这场司法反垄断博弈的赌注急剧升级。Bonta 虽未公开具体的和解清单，但他已明确表态，任何可接受的救济措施都必须是“结构性剥离”（如剥离部分核心资产），而绝不会接受口头的“行为承诺”（例如仅仅口头承诺未来每年保留 12 部或 24 部电影的制作量）。双方在企业架构与合并合法性上的冲突已陷入剑拔弩张的僵局。 ### [09:23](https://www.youtube.com/watch?v=I8tsEszPcBY&t=563s) Paramount 巨额倒计时费与撤离加州的“核选项”威胁 Paramount 正在面临极其昂贵的时间压力。从 10 月 1 日开始，由于与 Warner Bros. Discovery 的并购交易悬而未决，Paramount 每天必须向对方股东支付高达 700 万美元的“倒计时费”（ticking fee），直到交易正式完成。然而，加利福尼亚州发起的反垄断审讯预计要等到 2027 年 5 月 2 日才会正式开庭。这意味着如果官司一直拖到审判结束，Paramount 累计支付的延迟罚金可能高达 12 亿美元左右。为了打破这一死局，接管派拉蒙的 David Ellison 决定反向给加州政府施压：如果州政府不肯配合为合并扫清障碍，Paramount 就准备将工作岗位全部撤离加州。 两位主持人指出，如果反垄断审查拖延数年，不仅将彻底冻结 Ellison 家族所有的业务整合计划，更会造成巨大的财务失血。因此，威胁离开加州几乎是 Ellison 手中唯一的“核选项”（nuclear option）。根据《Variety》杂志的报道，派拉蒙最具标志性的洛杉矶总部将成为第一批外迁的业务部门，Ellison 甚至已经制定了一项为期 5 年的长期规划，准备将绝大部分制片厂工作岗位彻底迁出加州，候选落脚点包括佐治亚州、德克萨斯州和田纳西州。主持人调侃道，如果真的把带有经典水塔标志的派拉蒙好莱坞历史片场连根拔起，这在象征意义上无异于给加州投下了一枚震撼弹，也必将激怒整个洛杉矶影视工业圈。 - **每日延期费**：自 10 月 1 日起，派拉蒙每天需支付 700 万美元。 - **潜在损失**：若反垄断审讯拖延至 2027 年 5 月结案，累计费用可达约 12 亿美元。 - **拟迁出地**：佐治亚州（Georgia）、德克萨斯州（Texas）以及田纳西州（Tennessee）。 --- ### [11:28](https://www.youtube.com/watch?v=I8tsEszPcBY&t=688s) SpaceX 收购 Tesla：马斯克斩获万亿薪酬包的隐秘跳板 随后，话题转向了资本市场传闻已久的重磅重组猜想——SpaceX 收购 Tesla。《华尔街日报》在头版披露了一个极其离奇但具备可操作性的商业剧本：埃隆·马斯克（Elon Musk）或将借助两家公司的合并，以惊人的“走捷径”方式多赚取 1 万亿美元的个人报酬。虽然这套方案错综复杂且在短期内难以一步到位，但其背后的公司治理架构与条款设计引发了市场的强烈关注。 这套逻辑的核心源于 Tesla 股东此前批准的 2025 年马斯克高额薪酬方案。在那份对赌协议中，马斯克立下了极度激进的军令状：要求 Tesla 在规定期限内交付上百万辆 Robotaxi 自主出租车、大规模量产出货 Optimus 人形机器人，并将公司市值推升至约 10 万亿美元的天文数字，才能全额解锁新一批海量股票期权。然而，《华尔街日报》挖掘出该薪酬协议中藏有一项鲜为人知的“隐秘条款”：一旦 Tesla 发生控制权变更（Change of Control，例如被外部实体或关联实体 SpaceX 吸收收购），附加在该股票奖励上的半数严苛业绩考核门槛将被直接豁免抹除。 ```mermaid flowchart TD accTitle: 马斯克薪酬解锁路径对比 accDescr: 展示特斯拉常规业绩考核解锁与控制权变更豁免机制的对比流程 subgraph NormalPath["常规解锁路径（难度极大）"] A["2025 年薪酬方案"] --> B["达成全部激进业务指标\n(100万辆 Robotaxi + Optimus 量产)"] B --> C["市值冲刺约 10 万亿美元"] C --> D["全额解锁 1 万亿美元期权奖励"] end subgraph MergerShortcut["收购跳板路径（潜在捷径）"] E["SpaceX 发起对 Tesla 的并购"] --> F["触发『控制权变更』(Change of Control) 条款"] F --> G["直接豁免 50% 核心业绩硬性指标"] G --> H["大幅降低万亿级薪酬兑现门槛"] end ``` --- ### [13:35](https://www.youtube.com/watch?v=I8tsEszPcBY&t=815s) 扎克伯格的“钓鱼宣言”与 Meta 的 AI 雄心 节目聊天室里观众起哄，希望看到 Jordi 再次猛烈吐槽扎克伯格（“Feisty Jordy is based AF”）。事情的起因是马克·扎克伯格刚发表了一篇长达 6,000 字的亲 AI 宣言长文，甚至还在周末戴着 Ray-Ban Meta 智能眼镜边钓鱼边给知名科技记者 Alex Heath 打电话宣讲理念，随后在社交平台上发布意味深长的预告，声称 Meta 即将迎来“重大突破”。扎克伯格对外解释称，之所以在这个节点公开自己的技术哲学，是因为 Meta 即将推出实力显著跃升的下一代前沿大模型，他希望在这些更强大的模型问世之前，让全世界先理解他的价值观。 面对扎克伯格的这番表态，两位主持人展开了辛辣的讨论。Jordi 调侃扎克伯格所谓的“宣称价值观”与他的“实际行动”完全脱节：Meta 在商业上从来都是“看到热门产品就买下、抄袭，或者疯狂追逐新风口”，其公开形象总给人一种用力过猛、迫切求认可的“求关注实验室”（pick-me lab）的感觉。但不可否认的是，Meta 在底层模型上的工程执行力极其可怕，不仅在开源领域站稳脚跟，甚至在研发速度上有望超越许多闭门竞争对手。如果 Meta 下一代模型能够在表现上超越 xAI，甚至在 Gemini 4 发布后仍将其压制在第四位，那将是科技行业极为惊人的大逆转。 聊到科技领袖的公众信誉时，主持人对比了扎克伯格与 DeepMind 创始人 Demis Hassabis。他们指出，扎克伯格身负过去十余年社交网络算法争议和隐私诉讼的历史包袱，无论怎样以亲民形象出现，都很难在 AI 议题上充当消除公众疑虑的说客；而刚刚荣获诺贝尔奖的 Sir Demis Hassabis 说话纯粹聚焦于科学与乐观愿景，没有负面历史包袱，才是科技界向外界普及 AI 价值时最完美的代言人。最后，主持人笑称科技圈的争议永远聊不完，相比扎克伯格的宏大哲学，公众或许还是更想回到《华尔街日报》上关于马斯克如何再赚一万亿美元的现实商业戏码上。 - **扎克伯格长文规格**：发表了一篇长达 6,000 字的 Pro-AI 哲学宣言。 - **公关名场面**：扎克伯格戴着 Meta 智能眼镜一边钓鱼一边与科技记者 Alex Heath 通话，透露新模型动态。 - **行业格局对比**：Meta 的模型迭代展现出极强竞争力，正向前沿头部梯队发起强力冲击。 ### [19:10](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1150s) 坚持立场的敬佩与马斯克的“万亿薪酬”回归 在聊完开源与闭源的摇摆后，主持人不禁感叹 Anthropic 首席执行官 Dario Amodei 的知行合一：无论外界喜欢还是讨厌他，Dario 从第一天起就对技术扩散持强硬怀疑态度，且始终坚定反对无节制的开源。这种从不为了“赚取印象分”而见风使舵的一贯性，在科技圈反倒显得格外稀缺与令人钦佩。顺着这种“执着”的话题，对话自然转向了另一位将意志贯彻到底的代表人物——埃隆·马斯克（Elon Musk）。马斯克曾经体验过问鼎万亿美元身家的滋味，正如俗话所说“赚到第一个万亿永远是最难的”，如今他正试图在特斯拉重演奇迹，甚至将目标翻倍。 在特斯拉股东批准的薪酬协议中，马斯克最高可获得多达 4.23 亿股特斯拉股票，按 12 个批次逐步解锁。但按照正常履约条件，每一批解锁都要求特斯拉同时达成极其严苛的市值门槛与运营里程碑。最终的终极目标甚至要求特斯拉达到 8.5 万亿美元的恐怖市值。这一计划规模之宏大，不仅令业界咋舌，也几乎把马斯克个人的利益与特斯拉多项最前沿的技术产品彻底捆绑在了一起。 - **薪酬方案规模**：分 12 批次解锁，上限为 4.23 亿股特斯拉股票。 - **终极市值目标**：特斯拉市值需攀升至 8.5 万亿美元。 ### [21:02](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1262s) 苛刻的运营里程碑：从 FSD、擎天柱到百万人形车队 常规薪酬方案所设立的业务里程碑不仅跨度极大，难度也极高：特斯拉需要交付 2000 万辆汽车、拥有 1000 万活跃的 FSD（完全自动驾驶）订阅用户、交付 100 万台 Optimus（擎天柱）人形机器人，并将 100 万辆 Robotaxi 投入商业运营。主持人们分析指出，Optimus 仍处于极早期阶段，交付百万台看似天方夜谭；但 100 万辆 Robotaxi 和 FSD 订阅目标却并非遥不可及。目前特斯拉仅在单半年的交付量就已达到约 83.8 万辆，凭借强大的制造能力，一到两年内生产出百万量级的无人出租车完全可行。 关于 FSD 的成熟度，主持人们聊起了日常体验中的生动细节。有乘客反映在打 Uber 时，若是特斯拉车型，他们甚至希望能强制司机开启 FSD。因为许多人类新手司机无法熟练掌控特斯拉的动能回收制动，车辆行驶时顿挫严重、极易让人晕车不适；相反，升级到 Hardware 4（HW4）芯片后的 FSD 表现得异常平稳柔和，驾驶体验在很多场景下已经超越了普通人类司机。虽然当前车辆进出私人车道或特定小院仍偶尔需要接管，但 Robotaxi 完全可以采取靠街边落客的策略，“让乘客走完最后的 100 英尺”，技术门槛已无限逼近商业化落地的奇点。 - **运营里程碑目标**：累计交付 2000 万辆汽车、1000 万 FSD 活跃订阅、100 万台 Optimus 机器人、100 万辆商业运营 Robotaxi。 - **交付量现状**：特斯拉交付量虽在 2024 年和 2025 年经历波动，但上半年交付量已达 83.8 万辆，呈现回升态势。 - **无人驾驶对比**：目前 Waymo 全球运营车队规模约为 10,000 辆，特斯拉若达成百万量级将是其规模的近百倍。 ### [24:11](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1451s) “控制权变更”暗门：SpaceX 收购特斯拉的万亿捷径 然而，这份薪酬协议中深藏着一条重磅豁免条款：一旦特斯拉发生“控制权变更”（Change of Control，即被另一家实体收购），所有复杂的运营里程碑将瞬间全部作废。届时，马斯克能解锁多少批次股票将纯粹由收购时的估值决定——取交易前夕的市值与收购报价隐含价值之间的较高者。如果收购估值达到 8.5 万亿美元，即便特斯拉一台 Optimus 机器人都没有交付、一辆商业化 Robotaxi 都未上路，马斯克依然能直接拿满全部 4.23 亿股奖励，按当前估算其最高潜在价值高达约 8240 亿美元。 ```mermaid flowchart TD accTitle: 特斯拉薪酬协议的双轨解锁路径 accDescr: 展示常规严苛里程碑与控制权变更豁免条款两条路径的差异 subgraph RegularPath["常规路径：双重严苛门槛"] A["市值目标逐步递增<br/>（最高 8.5 万亿美元）"] B["四项极端运营指标<br/>（20M 交付 / 10M FSD / 1M 机器人 / 1M Robotaxi）"] A & B --> C["分 12 批解锁<br/>4.23 亿股期权"] end subgraph AcquisitionPath["捷径：控制权变更（如 SpaceX 收购）"] D["触发控制权变更协议"] --> E["所有运营里程碑自动作废"] E --> F["仅依据交易估值结算<br/>估值若达 $8.5T"] F --> G["直接全额解锁<br/>12 批次全部股票"] end ``` 这种架构使得市场盛传的“SpaceX 收购特斯拉”设想变得耐人寻味。马斯克在 SpaceX 拥有更高的持股比例，虽然要撬动数万亿美元的超大规模并购犹如天方夜谭、需要投行团队在多块显示屏前反复推演极其复杂的财务模型，但这一条款无疑为马斯克避开长达十年的艰苦业务爬坡提供了合法通道。两位主持人调侃这甚至可以做成特斯拉的新款保险杠贴纸：“我买这辆车，就是为了帮埃隆拿到他的一万亿美元薪酬”。更重要的是，合并将帮助马斯克进一步巩固对特斯拉的实际控制权，把他庞大的商业帝国真正统摄于同一屋檐之下。 - **免责条款机制**：若发生被收购或控制权转移，免除全部交付量、Robotaxi 及机器人数量的硬性要求，纯看估值结算。 - **顶格奖励现值**：虽然俗称“1 万亿美元薪酬包”，按《华尔街日报》预估，顶格 4.23 亿股奖励在目前的估值折算下约为 8240 亿美元。 - **博弈激励曲线**：马斯克的持股重心导致收购估值存在 U 型激励——估值极低或极高均对其有利，中间估值区间反而最为复杂。 ### [27:39](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1659s) Bending Spoons 的疯狂并购与 Zillow 高管离职风波 节目尾声，主持人们盘点了科技圈的其他热点动态。欧洲软件聚合平台 Bending Spoons 近期势头极为凶猛，凭借高达 265 倍的市盈率（P/E ratio），接连大举收购那些增长放缓但具备稳定现金流的“无聊老牌 SaaS”产品。在宣布收购协作平台 Airtable 后，Bending Spoons 市值单日暴涨 110 亿美元，整体估值跃升至 330 亿至 340 亿美元区间，较其 IPO 价格飙升逾 50%。科技界纷纷戏称“千万别让你的勺子被他们掰弯”（Don't get your spoon bent），争相讨论谁会是下一个被其吞并的目标。 节目的最后以一则颇具戏剧色彩的职场动态收尾：房产巨头 Zillow 的前首席运营官 Jun Choo 近日在 LinkedIn 上发文宣告离职，全文仅有极其冷峻的一句话：“我已经卸任 Zillow 首席运营官一职。这就是我要说的全部内容。”（"I have stepped down from my role as COO of Zillow. That's all I have to say about that."）这种宛如“摔麦走人”般的告别声明给外界留下了无尽的悬念与猜测，主持人们也笑着表示，大家很快就会知道这句意味深长的话背后究竟隐藏着怎样的职场内幕。 - **Bending Spoons 并购动态**：市盈率高达 265 倍，宣布收购 Airtable 后市值剧增 110 亿美元，总市值攀至约 340 亿美元，自 IPO 以来涨幅超 50%。 - **Zillow 高管变动**：原首席运营官 Jun Choo 在 LinkedIn 发布极简声明卸任，引发行业广泛关注与离职内幕猜测。 ### [28:51](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1731s) 莱昂纳多为濒危青蛙发声与节目尾声 在聊完前面的爆料八卦后，主持人打趣提到当事人欲言又止的模样就像是在为日后的自传式内幕爆料（tell-all）做铺垫，显然碍于某些限制现在什么都不能说。随后，两位主持人转向了当天的最后一则轻松趣闻：好莱坞影星莱昂纳多·迪卡普里奥（Leonardo DiCaprio）公开呼吁智利政府采取行动，保护一种正面临拟建输电项目威胁的极度濒危青蛙。主持人调侃道，这听起来完全是一件可以两全其美的事，“我们既能保护青蛙，又能把输电线路工程给建了”。 另一位主持人随即补充了这只青蛙的背景信息——这是野生数量仅存约 1,000 只的刺胸蛙（spiny chest frog）。两人开始幽默地出谋划策：既然莱昂纳多这么关心，干脆让这些小青蛙直接搬去莱昂纳多的豪宅住，或者接到美国送进动物园。他们随后站在智利官员的角度设身处地地吐槽：“当智利政府官员也太难了”，日常处理基建规划时突然天降名流，“心想这小李子到底是谁啊？直接从天而降指手画脚”，甚至还动员了他庞大的社交媒体力量——把自己的 6,000 万 Instagram 粉丝大军引流到智利官方账号下疯狂施压，处境实在太惨烈了。 节目在欢声笑语中画上句号。主持人们提醒观众现在正值夏天，别光顾着盯屏幕，“快替我们去外面踩踩草坪（go touch some grass）”，并预告明天同一时间太平洋时间直播再见，与观众道别结束了当天的节目。 - **争议物种**：刺胸蛙（Spiny chest frog），目前野外仅存约 1,000 只，栖息地受到当地电力输送工程项目的威胁。 - **明星效应**：莱昂纳多·迪卡普里奥利用其拥有约 6,000 万粉丝的 Instagram 社交声量向智利当局喊话，引发网络舆论对该基建项目的强烈关注。 ## Notable quotes > [01:35](https://www.youtube.com/watch?v=I8tsEszPcBY&t=95s) AI tokens are profitable, incredibly profitable — Jensen Huang > [04:30](https://www.youtube.com/watch?v=I8tsEszPcBY&t=270s) 英伟达计算不仅仅是芯片，而是一整套 AI 工厂平台，涵盖加速计算、网络架构、系统软件、AI 框架与全球开发者生态 — Jensen Huang > [13:35](https://www.youtube.com/watch?v=I8tsEszPcBY&t=815s) Feisty Jordy is based AF > [19:10](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1150s) 赚到第一个万亿永远是最难的 > [21:02](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1262s) 让乘客走完最后的 100 英尺 > [24:11](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1451s) 我买这辆车，就是为了帮埃隆拿到他的一万亿美元薪酬 > [27:39](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1659s) Don't get your spoon bent > [27:39](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1659s) I have stepped down from my role as COO of Zillow. That's all I have to say about that. — Jun Choo > [28:51](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1731s) 我们既能保护青蛙，又能把输电线路工程给建了 > [28:51](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1731s) 心想这小李子到底是谁啊？直接从天而降指手画脚 > [28:51](https://www.youtube.com/watch?v=I8tsEszPcBY&t=1731s) go touch some grass ## Mentioned Nvidia, Jensen Huang, CNBC, Financial Times, Larry Fink, BlackRock, David Solomon, Goldman Sachs, Jon Gray, Blackstone, Jim Zelter, Apollo Global Management, Bruce Flatt, Brookfield, OpenAI, Anthropic, Dario Amodei, Meta, Mark Zuckerberg, Ray-Ban Meta, Alex Heath, xAI, Colossus, Gemini 4, DeepMind, Demis Hassabis, SpaceX, Tesla, Elon Musk, Optimus, Robotaxi, FSD, Waymo, Paramount, David Ellison, Warner Bros. Discovery, Rob Bonta, Skydance, Variety, Wall Street Journal, Bending Spoons, Airtable, Zillow, Jun Choo, Leonardo DiCaprio, Instagram, John Coogan, Jordi Hays ## Video description Sign up for TBPN’s daily newsletter at TBPN.com Diet TBPN delivers the best of today’s TBPN episode in 30 minutes. TBPN is a live tech talk show hosted by John Coogan and Jordi Hays, streaming weekdays 11–2 PT on X and YouTube, with each episode posted to podcast platforms right after. Described by The New York Times as “Silicon Valley’s newest obsession,” the show has recently featured Mark Zuckerberg, Sam Altman, Mark Cuban, and Satya Nadella. TBPN is made possible by: Ramp - https://ramp.com Public - https://public.com Cisco - https://www.cisco.com Console - https://www.console.com CrowdStrike - https://www.crowdstrike.com Figma - https://www.figma.com MongoDB - https://www.mongodb.com NYSE - https://www.nyse.com Railway - https://railway.com Shopify - https://www.shopify.com/ Follow TBPN: https://TBPN.com https://x.com/tbpn https://open.spotify.com/show/2L6WMqY3GUPCGBD0dX6p00?si=674252d53acf4231 https://podcasts.apple.com/us/podcast/technology-brothers/id1772360235 https://www.youtube.com/@TBPNLive Chapters: 00:00 Nvidia’s $500B AI Compute Financing Machine 07:46 Paramount Threatens to Leave California Over Warner Bros. Deal 11:28 SpaceX–Tesla Merger Could Shortcut Musk’s $1T Payday 13:35 Zuck Fires Back + Is Meta Actually Near the AI Frontier? 19:49 How Musk’s $1T Tesla Pay Shortcut Would Actually Work 27:25 Bending Spoons’ Tear, Zillow Drama + DiCaprio’s Frog Fight [Source](https://www.youtube.com/watch?v=I8tsEszPcBY)

[1:35] Jensen Huang: AI tokens are profitable, incredibly profitable [4:30] Jensen Huang: 英伟达计算不仅仅是芯片，而是一整套 AI 工厂平台，涵盖加速计算、网络架构、系统软件、AI 框架与全球开发者生态 [13:35] Feisty Jordy is based AF [19:10] 赚到第一个万亿永远是最难的 [21:02] 让乘客走完最后的 100 英尺 [24:11] 我买这辆车，就是为了帮埃隆拿到他的一万亿美元薪酬 [27:39] Don't get your spoon bent [27:39] Jun Choo: I have stepped down from my role as COO of Zillow. That's all I have to say about that. [28:51] 我们既能保护青蛙，又能把输电线路工程给建了 [28:51] 心想这小李子到底是谁啊？直接从天而降指手画脚 [28:51] go touch some grass

[13:35] once you go to Ibiza, you must go back to Ibiza [27:25] Jun Choo: I have stepped down from my role as COO of Zillow. That's all I have to say about that

[0:02] Is that the Nvidia compute deal alarm? [0:05] Huge deal on the cover of the Financial [0:08] Times today. Uh Wall Street big names [0:11] join Nvidia to build $500 billion AI [0:14] financing package. A landmark lending [0:17] plan. Uh Jensen really lined up the [0:19] murderers row of uh financeers for his [0:22] new AI financing package. He brought [0:24] >> legendary setup over on CNBC too. One of [0:27] the greatest roundts of all time. Yeah, [0:29] incredible. Only Larry Frink was remote. [0:31] They got David Solomon, CEO of Goldman [0:33] Sachs in studio, John Gray from [0:35] Blackstone, Jim Zeltzer from Apollo, [0:38] Bruce Flat from Brookfield, and of [0:40] course Jensen breaking it all down for [0:42] everyone. Uh, I'm sure the Hollywood [0:44] starlets were pounding on the glass [0:46] outside looking for some new arm candy, [0:48] don't you think? [0:49] >> 100%. [0:50] >> 100%. because it's very rare to see that [0:53] many individual [0:55] >> capital allocators specifically capital [0:56] allocators put it on put it all on the [0:58] line every day in the markets. [1:00] >> Some would call them bad boys. [1:01] >> Some might, some might, some have. Uh so [1:05] I'm sure they were lined up, but it's a [1:06] great segment. Um I wanted to play this [1:07] one clip of them discussing just the [1:09] profitability, why they see this as an [1:11] investable asset. There were a couple [1:13] quotes. The the sheer size of the AI [1:15] infrastructure buildout is [1:16] unprecedented. Uh the president of [1:18] Apollo said more than $8 trillion of [1:21] capital is expected to be invested, a [1:23] staggering sum. We see an enormous [1:25] opportunity for private capital to [1:27] finance a portion of this along with [1:29] public public capital. So Wall Street's [1:31] not taking their foot off the gas. Uh [1:33] anyway, let's play this clip from CNBC [1:35] >> more of your margin of safety of energy. [1:37] >> So are these concerns about whether we [1:39] can meet this demand over overdone at [1:41] this point? Do do you think Jensen that [1:44] from where you see things, the demand [1:46] level and how we're building up around [1:47] it that it's going to be okay? It'll all [1:50] work out. [1:51] >> We're going to be constrained for some [1:52] time and pretty much across the board [1:54] from chips to memories to packaging to [1:57] systems, um, photonics, connectors, [2:00] land, power, construction workers, uh, [2:04] the whole thing, the the entire supply [2:06] chain up and down behind behind me [2:09] upstream all the way downstream. Uh and [2:11] this is happening at a time when AI has [2:13] become useful [2:15] >> because it's starting to do productive [2:18] work and it's happening all over all [2:20] over the world and AI tokens are [2:24] profitable incredibly profitable. When [2:27] you have something profitable, everybody [2:28] wants to make more of it. [2:30] >> Yes. [2:30] >> Great demand, great profitability. The [2:33] conditions are exactly right for the [2:35] work that we're doing right now. [2:36] >> Jensen, why these companies? Uh and and [2:38] did you go to any partners who said no? [2:40] >> No one said no. But th this is the six [2:43] premier [2:44] >> world's premier institutional [2:47] >> financeers for infrastructure. This is [2:50] the best of the best. [2:51] >> What John said that right now you'll be [2:55] less likely to have public capital that [2:56] comes into this because a lot of these [2:59] are companies that aren't making money [3:01] yet. Um is he right on that or are there [3:03] going to be big banks and others that [3:05] kind of step up? I believe within months [3:07] you're going to realize that these [3:08] companies are extremely profitable. [3:10] These are the fastest growing technology [3:12] companies in history. [3:13] >> Your customers, [3:14] >> that's right. These are fastest growing [3:16] technology companies in history and the [3:18] tokens they're generating are incredibly [3:21] profitable. You know, if if uh the [3:23] wafers that we buy from TSMC are [3:25] incredibly profitable. There's [3:26] incredible demand for it. I'm going to [3:27] want to buy a lot more. [3:29] >> Who are we talking about? Your [3:30] customers, which customers will have [3:32] access to these? [3:32] >> AI labs. AI labs. AI labs are the ones [3:35] that's the ones that you think are [3:36] profitable. But this will this [3:37] >> AI labs AI startups um you know as you [3:41] know this last 6 months the world put in [3:44] about $500 billion in AI startups $500 [3:47] billion the largest capital investing [3:49] investing period probably in recent [3:52] history and uh these companies need [3:54] compute and so we now have the vehicle [3:57] to do so. It's great stuff. 500 billion, [4:00] huge number. [4:01] >> Um, but uh it feels a lot smaller when [4:04] he lays out the actual compute [4:06] calculation. 50 billion per gigawatt to [4:09] >> 50 to 60 [4:10] >> 50 to 60. So you're looking at 10 [4:12] billion or 10 gigawatts of powered [4:14] compute. Um, which is, you know, the [4:17] labs are at like three combined across a [4:19] few of them. Uh, and we've been on this [4:22] 3x scaling. So this is really just like [4:24] next year's compute [4:26] basically. I think Meta individually has [4:28] a 10 gawatt plan or something like that. [4:30] So, [4:30] >> he's having to talk about profitability [4:33] and the profitability of incremental [4:34] tokens. And uh I think one of the [4:37] reasons that a lot of people are just [4:38] very uncomfortable with this is that the [4:40] two leading companies in the space are [4:42] private and their various numbers leak [4:46] out from time to time, but you [4:47] definitely don't get a complete picture. [4:50] And it's very unusual to have the two [4:52] companies driving the public markets are [4:54] actually private. [4:55] >> Yes. Well, that there is the SpaceX [4:58] factor here too, which uh recently [5:00] announced that they're going all in on [5:02] Nvidia, but again, they are turning into [5:06] a Neocloud in many ways and licensing [5:08] out compute. So, there's this world [5:09] where you might see Wall Street. [5:11] >> Yeah, but I look but I but I but I'm [5:12] looking at SpaceX more on this on the [5:15] supply side, right? Not driving. [5:17] >> No, exactly. So, so it's like a six [5:19] layer cake and you're seeing the the [5:21] finance guys there, then Nvidia making [5:23] the chips, marshalling all this capital. [5:25] a lot of it's going to go to a Colossus [5:27] 5 data center and then be rented by a [5:29] lab. Like that will be one of the [5:31] potential outcomes. Uh but then they're [5:33] they're just in the supply chain. Um [5:35] because I don't think I don't think the [5:37] uh like the semi- analysis forecast was [5:39] for like Grock token demand [5:41] specifically. It was just overall [5:44] inference and compute demand because [5:45] they're capable of building data centers [5:47] very quickly. Anyway, Jensen also took [5:49] to X to post a a long essay 12,000 [5:52] likes. He's like fully on Axe now. Uh [5:55] he's I haven't I haven't seen him. [5:57] >> It's almost like he was [5:58] >> lying lurking. [6:00] >> You think he was the whole time? [6:02] >> Well, I don't know if he was studying [6:03] enough because people are getting mad at [6:04] him about for you for for using the uh [6:06] the forbidden phrase. He says Nvidia [6:08] compute is not just a chip. It is a [6:10] complete AI factory platform including [6:13] accelerate computing networking system [6:15] software AI frameworks and global [6:18] developer ecosystem. That doesn't sound [6:20] that AI to me. I don't know. Somebody [6:22] will run this through Pangram and we'll [6:23] we'll get to the bottom of it. Jensen is [6:25] completing the circle says Pash. Bankers [6:28] don't like GPUs as collateral because [6:29] the depreciation is unpredictable. It's [6:31] unpredictable because a new GPU can [6:33] obsolete an old one. But Jensen knows [6:35] his own road map. So he's offering [6:38] depreciation insurance to the banks. The [6:40] depreciation insurance up to 25% helps [6:42] the banks get marginal deals over time. [6:44] Speculation here from Picos. Uh Nvidia [6:47] will also advise the banks on reference [6:49] designs for data centers that will make [6:50] them fungeable. So you will know this is [6:52] a one gigawatt data center but it's in [6:55] this particular class this particular [6:57] configuration it's blackwell data center [6:59] it's powered this way and so if you can [7:02] put it in this bucket you can underwrite [7:03] it a particular way because it's it it's [7:05] it becomes more funible um having them [7:07] be fungeible means the debt can be [7:09] repackaged into asset back securities [7:11] collateralized loan obligations and [7:13] collateralized debt obligations abslos [7:16] and cos from 2008 haha I'm sure there [7:18] will be a lot of uh a lot of folks upset [7:20] about all of the comparisons to the [7:23] mortgage uh the mortgage back security [7:26] uh buildout in 2008. Uh this allows [7:28] tranching to get investment grade [7:30] ratings on the debt so it can be sold [7:32] and resold to pension funds and [7:34] insurance firms. It also allows the [7:35] banks to trade idiosyncratic project [7:37] specific credit risk for sectorwide [7:40] credit risk. So Jensen is trying to get [7:42] his customers the same cost of financing [7:44] as real estate rather than venture [7:46] equity. this is going to move the data [7:48] center game out of the VCs and into the [7:50] big leagues. And so, uh, I'm sure people [7:53] will be speculating all sorts of [7:54] different things on what happens next. [7:57] There are some other stories. Paramount [7:58] is threatening to leave California by [8:00] October 1st if the state refuses to [8:03] negotiate a settlement in the legal [8:05] fight over its Warner Brothers Discovery [8:06] merger. We've been covering the story [8:08] back and forth for a while, but [8:09] according to Variety, Paramount CEO [8:11] David Ellison wants a quick quick [8:13] resolution to the antitrust lawsuit [8:16] brought by 12 state attorneys general [8:18] seeking to block the deal. He has now [8:20] told senior executives that Paramount is [8:22] prepared to begin the process of moving [8:25] its operations out of California if [8:28] Attorney General Rob Bont doesn't enter [8:30] settlement talks. Ellison reportedly [8:32] told his leadership team last week that [8:35] the Paramount Sky Dance board has [8:37] already approved the relocation plans. [8:40] Uh if negotiations with Bont haven't [8:41] begun by August 1st, the company would [8:44] start preparing its exit with [8:45] Paramount's Los Angeles headquarters [8:47] potentially moving out of the state as [8:48] early as as October. The threat [8:51] dramatically raises the stakes in [8:53] Paramount's fight with Bont who has [8:55] emerged as a leading opponent of the [8:57] Warner Brothers Discovery acquisition. [8:58] Bont hasn't publicly detailed what [9:01] Paramount would need to offer to resolve [9:03] the case, but he has said that any [9:05] acceptable remedies would likely need to [9:07] be structural, such as asset divesters [9:09] rather than behavioral commitments like [9:11] maintaining certain levels of [9:12] production. So, it's not going to be [9:13] enough for Ellison say, "Hey, we're [9:15] still going to do 12 movies a year, 24 [9:17] movies a year." Uh, it needs to be [9:19] something specific about the struct like [9:21] the actual structure of the company. Uh, [9:23] Paramount is also racing against an [9:25] expensive clock. Beginning October 1st, [9:28] the company will owe Warner Brothers [9:30] Discovery shareholders a $7 million $7 [9:33] million per day ticking fee until the [9:36] transaction closes. The state's [9:38] antitrust trial isn't scheduled to begin [9:40] uh until May 2nd of 2027, so almost nine [9:45] months from now. Uh meaning Paramount [9:47] could rack up roughly $1.2 2 billion [9:50] dollar in payments by the time the case [9:52] is expected to conclude. Ellison is [9:54] effectively putting pressure on [9:55] California from the other direction. If [9:57] the state won't help find a path to [9:59] closing the merger, Paramount is [10:01] prepared to take start taking jobs [10:02] elsewhere. [10:03] >> Yeah. So, so he's he's trying to force [10:05] the issue. This gets uh extremely [10:09] painful for the Ellison's if this [10:12] antitrust thing just drags on. You could [10:14] imagine it dragging on for a couple [10:17] years. that puts a pause on all of their [10:21] integration plans. It just makes [10:23] everything a lot harder. [10:24] >> I don't think he has this. This seems [10:27] like his one option, right? It's kind of [10:28] the nuclear option. It It's going to [10:31] piss off a lot of the industry here in [10:34] LA. [10:35] >> It is the most symbolic move because [10:37] according to Variety, the company's LA [10:39] headquarters would be the first [10:41] operation to leave. And Ellison already [10:44] has a 5-year plan to move most studio [10:45] jobs out of California. No destination [10:47] has been selected, but Georgia, Texas, [10:49] and Tennessee reportedly under [10:50] consideration. So, it's like if you're [10:52] moving the LA headquarters, the iconic [10:54] Paramount headquarters, you've [10:56] everyone's seen the water tower. Um, uh, [10:59] that is a a a huge shot across the bow [11:02] as opposed to something that might [11:03] actually be more economically impactful [11:06] like just, oh, for this production, we [11:08] were going to make it in in LA, we're [11:10] going to do it in uh in Atlanta. that [11:12] might actually move more dollars around, [11:15] but maintaining that headquarters is so [11:17] symbolic, right? So Ellison and much of [11:19] Paramount senior leadership currently [11:21] work from the historic Paramount Picture [11:22] studio lot in Hollywood, but maybe it [11:25] will be moving to Georgia, Texas, or [11:27] Tennessee. I wonder where they'll go. [11:30] Um, anyway, speaking of publicly traded [11:32] companies, Elon Musk may have found a [11:34] shortcut to unlocking a trillion dollar [11:36] payday. He needed a win. [11:39] Uh, [11:40] >> you asked for it. He's delivering. [11:42] >> Yeah. The idea is having SpaceX buy [11:45] Tesla. This has been rumored for a long [11:47] time. Lots of speculation on when these [11:49] two companies will merge, uh, if they [11:52] will merge. Um, but, uh, the Wall Street [11:55] Journal on the front page outlines a [11:58] very odd scenario where Elon could wind [12:00] up making an incremental $1 trillion. [12:03] Uh, it's really complicated and it's not [12:06] it's not as simple as just, oh, he'll [12:08] just buy the companies and then he just [12:09] gets another trillion dollars. uh it's [12:11] more it's more complicated but there is [12:13] a potential outcome but there's some [12:15] mitigating factors but it's worth it's [12:18] worth understanding the mechanics of the [12:19] deal or the mechanics of potential deal [12:21] uh as it might play out over the next [12:23] few years because this is something that [12:24] couldn't happen right now but uh in the [12:27] future it is possible. So, the Wall [12:29] Street Journal reports that there's an [12:31] obscure provision in Musk's 2025 Tesla [12:36] pay agreement that's already been uh [12:38] approved by the shareholders. And it was [12:40] that crazy one we talked about where it [12:42] was like a million robo taxis by this [12:44] time and like the Optimus needs to be [12:46] shipping and it was a lot of really bold [12:48] claims, but if he hits all those and and [12:50] the stock goes to like 10 trillion, then [12:52] he gets a huge unlock of new stock. And [12:55] it was all like, okay, well, like it's a [12:57] lot of money, but that's a lot of [12:58] progress for this company. Yeah. Uh [13:00] because uh they there's a lot of [13:02] projects at Tesla that just are sort of [13:05] slow and steady, not really advancing [13:07] that quickly. And so he was sort of [13:10] throwing down the gauntlet saying, [13:11] "Okay, give me another couple years, and [13:13] I'm going to deliver in a really big [13:14] way, and if I do, I want to be [13:15] compensated for that." And the [13:16] shareholders approved. But the obscure [13:18] provision uh there's an obscure [13:20] provision in the already approved uh [13:23] Tesla pay agreement that could eliminate [13:24] half of those performance requirements [13:27] attached to the stock award if Tesla is [13:30] acquired. So change of control affects [13:32] those those pay packages. What are you [13:34] laughing about? [13:35] >> Mark in the chat says talk about [13:37] Zuckerberg again. I [laughter] like [13:39] Feisty Jordy. Feisty Jordy is based AF. [13:43] [laughter] [13:44] >> Yeah. Guess I woke up on the wrong side [13:47] of the bed yesterday. I had some strong [13:49] opinions. [13:50] >> We'll see. [13:51] >> Somebody had somebody had to say it. [13:52] >> Hey, hey, he fired back. He called out [13:54] Alex Heath. He said something big is [13:56] coming. He was vague posting directly to [13:58] Alex Heath. [13:59] >> And again, I just felt like that was so [14:00] mimedic with all the people that are [14:03] actually at the frontier. [laughter] [14:04] And uh it just the whole thing [14:08] [laughter] I'm like, "Yeah, I'll I'll [14:10] believe it when I see it." [14:11] >> Yeah. Yeah. It's kind of [14:13] >> what was the actual See, see Mark, now [14:15] you got me stuck. [14:16] >> Now you snapped him. [laughter] [14:18] >> Now I snapped. Over the weekend, Zuck [14:21] called me via his meta glasses while [14:23] fishing to discuss the 6,000word ProAI [14:27] manifesto he published this morning. Uh [14:29] yeah, so he's he's trying to balance [14:31] >> I love that aesthetically. That's that's [14:32] fantastic. [14:33] >> Good old American [14:35] >> American American. Um, he told me the [14:39] immediate reason he decided to publish [14:41] his philosophy now is that he believes [14:44] Meta is very close to having [14:47] substantially stronger models. [14:49] >> Probably true. [14:50] >> And he wants people to understand his [14:52] values before those models arrive. [14:55] >> Uh, we know your values. [14:59] >> We know your values which uh are [15:03] >> connecting the world. It's fine. [15:05] Instagram's fine. No, I mean the the [15:07] whatever whatever the stated values [15:10] versus the the lived values. The lived [15:12] values are [15:14] >> if there is a hot product [15:16] >> small businesses grow. What about that [15:19] value? Come on. Uh Instagram has birthed [15:22] so many companies. We know so many [15:23] people that their companies would not [15:25] exist if not for meta platforms. [15:27] >> How about great hours of enjoyable [15:29] content? [15:29] >> Yes. [15:30] >> Everyday Americans to watch every day. [15:31] Would would we be even remotely familiar [15:34] with Professor Sendi and the creation of [15:37] the wombo without Mark Zuckerberg? [15:39] >> You would have never seen that he turned [15:41] it down. [15:41] >> You you would have never seen that he [15:43] turned it down [15:44] >> or the dabbita. [15:46] >> Yeah. You wouldn't you wouldn't have you [15:47] would not be aware of that. [15:48] >> All these iconic moments. [15:50] >> You also wouldn't know that once you go [15:52] to IBA, you must go back to IBA. And you [15:56] would not know that there's a [15:57] 21-year-old bodybuilder who looks like [15:58] he's 35 who's on his way. [laughter] [16:02] >> See, it's enjoyable. Admit it. [16:04] >> Um I I don't have a problem with Mark. [16:08] >> Yeah. [16:08] >> And I don't have a problem with Meta [16:11] products. I do get enjoyment out of [16:12] them. [16:13] >> I [16:15] have built companies that are dependent [16:17] on Meta Platforms. I've invested in many [16:20] companies that that exist in their [16:22] current form entirely because of Meta [16:24] Platforms. [16:24] >> There you go. But I I know what Mark's [16:27] values are. [16:28] >> What are those? [16:29] >> Mark cares about [16:31] >> delighting customers, [16:33] profitable advertising. These are good [16:35] values. [16:37] >> I don't I wish he cared more about [16:38] advertising. [16:39] >> That's true. Yeah. [16:40] >> I wish he cared more about advertising. [16:42] >> That would be good. [16:42] >> But but you know, you just look at the [16:44] historical behavior. It's like buy or [16:47] copy or chase the hot thing. [16:50] >> Yeah. But that's just that's just a [16:51] game. like, you know, you're talking [16:53] about a10 billion dollar startup that's, [16:56] you know, coming after you with some new [16:58] form factor. I don't know. You copy it. [17:00] >> I I think you're discounting how good [17:01] MSL is. Everyone is saying that MSL is [17:03] like clearly in third place, right? That [17:05] means that they're ahead of XAI. [17:06] >> I mean, if they flip, it's crazy. If [17:09] they if they if they actually get the [17:11] next version of Spark out before Gemini [17:14] 4 and it's better and then Gemini 4 [17:16] launches and stays in fourth place, like [17:18] that is a crazy crazy reversal. They're [17:21] executing well. Yeah. Coming from [17:23] incredibly far behind. [17:25] >> They are approaching the frontier. [17:28] >> I just think that if he wants to be the [17:30] pick me lab, [17:31] >> you don't like the comms. [17:33] >> I don't. [17:34] >> Yeah. I just think it's not right. [17:38] >> I just don't think it's very authentic. [17:40] >> Okay. I I think it's the opposite. I [17:42] think I think it is authentic. I think [17:44] it is he he is thinking about this [17:46] stuff. He might not be the it is it is [17:50] just a little bit of a tough voice [17:52] because there's so much attention from [17:54] the social reckoning and the social [17:56] network and stuff and then like in terms [17:59] of the overall AI industry uh attaching [18:03] him it's like attaching Demis to the AI [18:06] industry was definitely better for the [18:08] voices of the AI industry because Demis [18:10] would stay on message and just talk [18:12] purely about uh science. He won the [18:14] Nobel Prize. And so if if I was like, [18:17] "Okay, I'm I'm dealing with an AI [18:19] skeptic who is paranoid about water use [18:22] and surveillance, like who can I put [18:24] them in the room with?" I'm like, "Yeah, [18:26] go sit down with Demis, Sir Demis. He's [18:29] going to put on a good show and like and [18:31] like walk you through this and give you [18:32] a really optim optimistic vision and [18:35] he's not going to have the baggage of [18:37] anything else, any lawsuits or anything [18:39] else, right?" Uh, and so yeah, it's a [18:42] little bit of a it's a little bit of a [18:43] tough tough go. Anyway, back to [18:45] >> anyways, I just I just think like again, [18:48] >> let's get back to less controversial [18:50] tech leaders. Let's talk about Elon Musk [18:52] and how he's going to make his next [18:53] trillion because this is what's in the [18:55] Wall Street Journal today. [laughter] [18:57] >> I wish I I'll just I'll just [19:00] >> I'll just end it here. I wish that Zuck [19:03] came out with like a five-point plan to [19:05] get his own trillion dollar pay package, [19:07] which was just like sell like trillions [19:10] of dollars worth of ads. [19:13] >> And I would I would just be praising [19:14] that all day. [19:15] >> That would be sick. [19:16] >> I would be praising that all day versus [19:18] being like, "Oh, we're doing open source [19:20] and now we're not doing open source and [19:22] now we're open uh we're open sourcing [19:24] again because I I'll get some brownie [19:26] points." Yeah. [19:27] >> Uh but also we're gonna [19:30] >> It doesn't feel like [19:30] >> still have closed models. [19:32] >> Yeah. I mean, you go to the Dario thing [19:33] and it's like the guy's been hard on [19:34] China. He's been anti-open source since [19:36] day one. [19:37] >> Extremely [19:37] >> extremely consistent. You love him, you [19:39] hate him, but like there's consistency [19:41] there and that is uh just reliable and [19:44] and in many ways admirable. I I [19:45] understand what you're saying. Anyway, [19:47] back to Elon Musk. He's been extremely [19:49] consistent. He's getting that trillion [19:50] dollars from [laughter] Tesla one way or [19:53] another. And there are tasted being a [19:54] trillionaire before [laughter] he wants [19:56] to run it back. [19:57] >> He's got to run it up. He's got to [19:58] double it. They say your first trillion [20:00] is the hardest and this is how he gets [20:02] his second trillion. So this is how he's [20:04] going to do it. So in Musk's 2025 Tesla [20:07] pay agreement, uh if he hits certain [20:10] milestones, he can eliminate a lot of [20:12] those performance uh performance [20:13] requirements that were previously [20:15] attached to a stock award uh if the [20:18] company's acquired. And so change of [20:21] control at this scale normally you would [20:24] think that's impossible. You cannot take [20:26] Tesla private. He tried it. It was not [20:27] impo it was too big of a company. And [20:30] then also once you're funding was [20:32] secured the funding was secured but uh [20:34] but there was a trillion dollar once [20:35] you're a trillion dollar company like [20:36] you can't just get acquired except there [20:38] we're in this very unique case where [20:40] SpaceX is also a trillion dollar [20:41] company. So Tesla shareholders approved [20:44] the compensation plan in November. last [20:47] November. Under its normal terms, Musk [20:50] can earn as many as 423 million Tesla [20:54] shares across 12 tranches, but each [20:56] tunch requires Tesla to hit both a [20:59] market cap and an operational milestone. [21:02] So, the goals are deliberately enormous. [21:04] Uh Tesla would eventually need to reach [21:06] an $ 8.5 trillion market cap while [21:09] accomplishing targets including [21:11] delivering 20 million vehicles. [21:13] Remember, I think the number of vehicle [21:14] deliveries is actually like declining [21:16] this year. Uh it it was a very very bold [21:19] plan. They need 10 million active FSD [21:21] subscriptions. That actually seems easy. [21:23] FSD is really good. They need to deliver [21:25] 1 million Optimus robots. That seems [21:27] crazy because Optimus is still so early [21:29] as a project. [21:30] >> They should try to interview someone [21:31] with a Tesla that doesn't get the FSD. [21:36] >> They turn it down. And I want to [21:37] understand [21:39] >> why. [21:40] >> I understand it. if you just bought an [21:41] old older Tesla, you like the ease of [21:43] charging and you haven't upgraded to the [21:45] newer hardware package, uh because it's [21:48] like like the the FSD is available on [21:51] the older hardware 3 uh technology, but [21:54] uh I think it's best on the hardware 4, [21:55] which is like 2024 onward. Not everyone [21:57] can upgrade. If it's a financial [21:58] decision, I understand it. Um but the [22:01] last one was uh they need to get a [22:03] million robo taxis into commercial [22:05] operation. That also seems pretty [22:07] doable. I saw a robo taxi driving around [22:09] LA recently in gold and uh like just [22:13] from using FSD it seems like it's ready [22:16] like there's probably some legal stuff [22:18] but uh in general I think they could [22:21] roll out the the robo taxis like very [22:23] quickly they can make a million cars [22:24] pretty quickly and they have the [22:26] technology so it's just about putting [22:27] those on the road [22:28] >> yeah I haven't spent a ton of time in [22:29] Teslas but the times recently where [22:32] human drivers had to take over was [22:35] getting into a driveway into effectively [22:38] like private property. It's a robo taxi [22:40] you can just pull over the side of the [22:42] street and say like you're walking the [22:43] last 100 feet. But [22:44] >> Yep. totally. [laughter] [22:45] Um and so yeah, I I mean that that one [22:48] doesn't seem that that difficult. Uh [22:50] although obviously it is a lot. I mean I [22:52] think that's like a hundred times as [22:54] many Whimos. I think there's like 10,000 [22:55] Whimos out there. So uh it would be it [22:58] would be a big a big move, but over a [23:00] couple years is that possible? [23:01] >> Yeah. vehicle deliveries uh were falling [23:05] in 2024 and 2025, but seem to be [23:09] rebounding, [23:09] >> but they're not they're not far off from [23:11] like a million vehicles, right? [23:13] >> Uh 838 in H1 [23:15] >> 838,000. Yeah. And so and so over a year [23:19] or two, they could probably manufacture [23:21] a million robo taxis. And I think the [23:23] technology is pretty much there. When [23:25] you actually uh look online and you see [23:27] the reviews of people talking about [23:28] Tesla Ubers, they're like, I wish there [23:31] was a Tesla product or an Uber product [23:34] where I could demand that if it's a [23:36] Tesla, they have to stay in full [23:38] self-driving mode because many Uber [23:40] riders regard the full self-driving [23:43] experience as smoother and less likely [23:46] to cause indigestion and sickness in the [23:48] back seat versus watching a a driver who [23:52] has a Tesla who isn't that experienced. [23:54] and doesn't understand how to use the [23:56] regenerative braking properly and so [23:58] that it's much more jerky when a human's [24:00] driving it because FSD is actually super [24:02] human relative to a newbie uh Uber [24:05] driver with a Tesla. So, uh I think the [24:07] techn is like very very close. They got [24:09] to manufacture it. Obviously, legal [24:10] stuff, but they'll get there. But [24:11] there's a major exception buried in the [24:13] agreement. If Tesla undergoes a change [24:15] of control, essentially if Tesla is [24:18] acquired, the operational requirements [24:20] disappear. you no longer have to hit a [24:24] million robo taxis or a million [24:25] optimuses uh to unlock those new tanches [24:29] of stock for Elon. If there's a change [24:31] of control, it's purely based on the [24:33] market cap. And so instead, Tesla will [24:35] determine how many of Musk's 12 tranches [24:39] have been earned solely by looking at [24:42] the company's value at the time of the [24:43] transaction. So the milestones don't [24:46] matter anymore. Only the market cap [24:48] matters. And so that could become [24:50] extremely important if the widely [24:52] speculated buyer turns out to be another [24:54] Musk controlled company, SpaceX. So if [24:57] SpaceX comes in and gives a really high [24:58] price for Tesla, Musk can unlock more of [25:00] those tanches and get more equity in [25:03] Tesla, which then rolls into the [25:04] combined entity, of course. So under [25:06] this agreement, uh Tesla's value in an [25:08] acquisition would be calculated using [25:11] whichever is higher, its market [25:12] capitalization immediately before the [25:14] deal or the value implied by the price [25:16] being paid to Tesla shareholders. If [25:19] that figure reached 8.5 trillion, all 12 [25:22] tanches tanches could qualify, putting [25:25] Musk in line for the full 423 million [25:28] share award without Tesla ever having to [25:30] accomplish many of the pay packages [25:32] operational milestones. So there's an [25:34] obvious catch which is how are you going [25:37] to acquire Tesla for 8.5 trillion uh [25:40] when when uh SpaceX is not a10 trillion [25:44] company or $50 trillion company. Uh it's [25:47] a hard pitch to do a merger at that [25:49] scale. Um but this is more of like a [25:51] what might happen in a number of years. [25:53] Well, yeah, and there's also the [25:54] trade-off between uh he has more [25:56] ownership of SpaceX, which would mean [25:59] that he would benefit from acquiring [26:01] Tesla at a lower valuation, but then [26:04] >> depending on these milestones, you know, [26:07] he's there's probably uh his bankers [26:10] probably have some very elaborate [26:12] spreadsheets, [laughter] [26:12] you know, six different monitors looking [26:15] at all the different [26:16] >> So, basically, there's sort of a U shape [26:18] to the incentive. If Tesla's a really [26:20] low price, Elon probably benefits from [26:22] acquiring it. And if Tesla's a really [26:24] high price, he he benefits from he [26:27] actually increases overall ownership [26:29] from buying it. Uh but there's sort of [26:30] like a messy middle where it gets a [26:32] little bit rougher. So uh 8.5 trillion [26:34] would be more than six times Tesla's [26:36] recent market cap. Uh and Tesla [26:38] shareholders would still have to approve [26:40] a SpaceX acquisition. The journal [26:41] estimates uh that Musk's maximum award [26:45] is currently worth about 824 billion [26:48] despite the package's familiar $1 [26:50] trillion label. Uh still the provision [26:53] creates an unusual path around some of [26:55] the hardest arguments uh hardest [26:56] requirements in Musk's compensation [26:58] package. Instead of spending the next [27:00] decade hitting a dozen separate [27:01] operating goals, a sufficiently [27:03] expensive acquisition of Tesla could [27:06] effectively declare those goals [27:07] accomplished. [27:08] >> New bumper stick. New Tesla bumper [27:10] sticker idea. [27:11] >> I bought this to help Elon achieve his [27:15] one trillion dollar pay package. [27:17] >> Yes. Full send on the pro Elon Tesla. [27:21] Get a lot of thumbs up. [27:23] >> A lot of positive honks. [27:24] >> For sure. For sure. Uh there could be [27:27] another major benefit for Musk. A [27:29] combination with SpaceX could increase [27:30] his effective control over Tesla. [27:32] something Musk has repeatedly sought [27:34] while consolidating even more of his [27:36] business empire under a single roof. [27:39] Well, uh it's a fascinating story. Any [27:41] other uh stories you want to cover? [27:44] Bending Spoons apparently on a tear. [27:46] 265x [27:48] p uh price to earnings ratio buying bor [27:51] old boring SAS with slow growth. People [27:53] are excited about that. I guess Bending [27:54] Spoons added 11 billion in market cap [27:57] after announcing the Air Table [27:58] acquisition. Can't make it up. Is that [28:00] true? Wow. 33 34 billion company. Uh way [28:06] up since IPO. Wow. Really really uh [28:09] really impressive. Up 50% since they [28:11] IPOed. Um they're on a tear. Who will be [28:14] the next Ben Spoon? That is the question [28:16] everyone's asking. Do not get [28:19] >> Don't get your spoon bent. [28:20] >> Don't get your spoon bent. Do we need to [28:22] >> unless that's the best possible outcome. [28:23] >> Yeah, maybe. [28:24] >> Um we'll leave this show with a post [28:28] from uh June Chu. Mhm. [28:31] >> who was the COO at Zillow until very [28:34] recently. He posted on LinkedIn, "I have [28:37] stepped down from my role as COO of [28:39] Zillow." That's all I have to say about [28:41] that. [28:41] >> Mic drop moment. There we go. [28:43] >> And yeah, I'm curious. Uh I'm sure we'll [28:46] find out in due time [28:49] what he really wants to say about that. [28:51] [laughter] [28:51] >> Yeah, it feels like teeing up a tell [28:53] all. Uh [28:54] >> feels like he maybe can't say anything. [28:57] >> Sure. For sure. Um, [28:59] [laughter] [28:59] >> anyway, there's a bunch of other good [29:01] stories. [29:02] >> Last last but not least, Leonardo [29:04] DiCaprio urged Chilean authorities to [29:07] protect a critically endangered frog [29:09] from a proposed power transmission [29:11] project. This feels like something [29:13] that's solvable. I think we can protect [29:15] the frogs and do the power transmission [29:18] project. [29:19] >> Yeah. [29:19] >> Do you agree, John? I know nothing about [29:21] the situation, but I have I have a [29:25] little faith. [29:25] >> Spiny chest frog. There's a thousand of [29:28] them that remain in the wild. [29:30] >> I think Leo should let them move in. [29:33] >> Bring them to America. Put them in a zoo [29:35] or something. I think they want to live [29:36] there. I think they I think you got to [29:38] do a lot of work. It's got to be rough [29:40] being the being the in the Chilean [29:42] government and just being like, "Who's [29:44] Leo?" Just comes over the top. [29:46] >> 60 He sent 60 million Instagram [29:50] followers our way and they're here on [29:52] us. [29:53] >> It's got to be brutal. [29:54] >> It's summer. [29:55] >> Yeah, it's summer. go touch some grass [29:57] for us [29:58] >> and we'll see you tomorrow Pacific. [30:00] >> Goodbye. [30:01] >> Love you.