Cerebras Raises IPO Target to $4.8 Billion as Demand Surges

需求激增，Cerebras将IPO目标估值上调至48亿美元 对AI智能体推理速度的强劲需求，推动Cerebras将其公开招股价格区间上调至每股160美元。 主持人深入剖析了多项重点市场与地缘政治热点，首先探讨了Audemars Piguet与Swatch备受争议的“Royal Pop”联名合作以及奢华腕表的转售趋势。随后，他们分析了电动汽车制造商BYD为拓展欧洲市场聘请Daniel Craig，以及在其晶圆级推理硬件强劲需求的推动下，Cerebras大幅上调IPO定价。最后，他们讨论了特朗普总统携美国顶级CEO与中国领导人习近平举行的高风险北京峰会，并以对Palmer Luckey禁止邮政垃圾邮件提议的轻松调侃作结。 爱彼（Audemars Piguet）与 Swatch 联名推出灵感源自皇家橡树的官方平价腕表“Royal Pop”，预计发售价约 400 美元且仅限线下门店发售，二级市场预估溢价达 5 至 12 倍。 (0:02) 比亚迪（BYD）聘请前 007 主演丹尼尔·克雷格为其高端电动车腾势（Denza）拍摄欧洲电影感广告，顶配车型在欧洲售价达 13.4 万欧元，力图撕掉低价标签。 (10:52) AI 芯片厂商 Cerebras 更新 IPO 招股书，在机构超额认购达 20 倍的热度下，将发行价格区间上调至每股 150 至 160 美元，预计募资总额增至 48 亿美元。 (14:20) 美国总统特朗普率豪华商业高管代表团访问北京，除探讨贸易与 AI 供应链外，因中东战事导致霍尔木兹海峡关闭引发的能源危机成为核心议题。 (20:41)

TBPN explores shifting commercial and geopolitical boundaries, highlighted by Audemars Piguet's controversial Swatch collaboration, Cerebras Systems' surging public market debut, and high-stakes U.S.-China diplomatic talks. Audemars Piguet Downmarkets With Swatch (0:02) Audemars Piguet partnered with Swatch to release the "Royal Pop," an accessible adaptation of the iconic Royal Oak that mirrors the strategy of the Omega MoonSwatch. The hosts argue that with Royal Oak secondary market prices reaching approximately $30,000—and the Code 11.59 struggling to capture equivalent interest in the low-to-mid $20,000 range—AP lacked an entry point into its brand ecosystem amidst a flood of convincing replicas. Aero Grevner noted that licensing the design serves as intellectual property damage control after AP lost trademark fights over its octagonal bezel and dial in Japan in 2024 and the U.S. in 2025. Resellers anticipate an in-store drop retailing near $400, with secondary markups forecasted between 5x and 12x. BYD Targets European Luxury Market (10:52) Chinese electric vehicle maker BYD recruited Daniel Craig to front a European advertising campaign for its high-end Denza brand. The Denza lineup retails between $40,000 and $60,000, but top-spec Denza Z configurations reach up to €134,000 (roughly $60,000 to $140,000). The hosts observe that the commercial incorporates synthesized internal combustion engine audio, underscoring BYD's strategic push to shed its legacy reputation for low-cost commuter cars and compete directly with incumbent European luxury automakers. Cerebras Upsizes IPO Amid Inference Demand (14:20) Ahead of its Nasdaq listing under ticker CBRS, chipmaker Cerebras expanded its offering size from 28 million to 30 million shares and raised its price target to $150–$160 per share, aiming to raise $4.8 billion. Supported by sustained transformer architecture dominance and the demands of AI agents, Reuters reported an order book oversubscribed by 20x, indicating roughly $100 billion in institutional demand from clients such as Amazon's AWS and OpenAI. Polymarket forecasted the company's valuation to clear $50 billion on day one, positioning early backer Benchmark—which retains an ownership stake above 20%—for a historic venture return on Cerebras' single-wafer hardware design. Trump Summit Tackles Middle East Energy Crisis (20:41) President Donald Trump traveled to Beijing accompanied by a delegation of prominent U.S. business leaders, including Tim Cook, Larry Fink, Stephen Schwarzman, and Jane Fraser. Although tech executives pushed to discuss GPU export rules, rare earth supplies, and AI governance—such as Chinese testing requirements through CAISI—talks were overshadowed by the U.S.-Israel-Iran conflict that closed the Strait of Hormuz. Because the passage handles roughly one-fifth of global crude flows, both administrations face strong economic incentives to ease energy market disruptions, even as Trump cautioned that an existing peace proposal remained on "major life support." Across consumer luxury, high-performance computing, and cross-border diplomacy, legacy operators are being forced to adapt to disruptive market pressures, though ongoing geopolitical conflicts around critical transit routes like the Strait of Hormuz threaten to complicate macroeconomic stabilization.

本期节目聚焦于前沿科技资本与地缘经济格局的剧烈震荡：AI芯片先锋Cerebras在推理需求爆发驱动下大幅上调IPO发行规模与定价区间；爱彼（Audemars Piguet）通过与斯沃琪（Swatch）联名推出平价表款，在知识产权诉讼失利与二级市场泡沫中寻求防御转型；在国际宏观层面，特朗普率领豪华商界代表团访华，核心斡旋霍尔木兹海峡危机对全球能源供给的致命威胁。 Cerebras晶圆级芯片与IPO估值跃升 (14:20) 据路透社报道，AI芯片制造商Cerebras更新了招股说明书，将IPO目标日期定于5月14日（周四）。由于机构投资者的爆发式需求，公司将发行规模从2800万股扩大至3000万股，发行价格区间由每股115–125美元大幅提高至150–160美元，预期募集资金总额从35亿美元跃升至48亿美元。报道显示，该IPO认购订单已达20倍超额认购，对应约1000亿美元的机构需求体量。早期投资机构Benchmark仍持有超过20%的股权，有望收获风投史上回报最高的基金之一。预测市场Polymarket数据显示，Cerebras（纳斯达克代码：CBRS）首个交易日收盘市值预计将跨过500亿美元门槛，远高于其早前报道的260亿美元目标估值。 技术层面上，芯片行业正经历Ben Thompson所称的“向推理端迁移”进程。Cerebras颠覆了将300毫米硅晶圆切割封装为微小芯片的传统半导体制造路径，直接将整片晶圆作为单一连续计算单元，单片集成了4万亿个晶体管、90万个AI计算核心以及PB级（petabits-per-second）片内带宽。招股文件披露其重要客户包括亚马逊AWS与OpenAI。在Codex Desktop结合GPT-5.3 Spark模型的实际应用中，晶圆级架构能够实现高智能度答案的近乎即时全页输出，彻底消除了传统AI推理数分钟的延迟瓶颈，主持团队预计该技术将被Cursor、Anthropic、Meta与Google等前沿实验室加速采用。 爱彼与斯沃琪联名：IP抗辩与平价下沉 (0:02) 瑞士制表名厂爱彼（AP）出人意料地与大众品牌斯沃琪（Swatch）合作，推出致敬其旗舰表款皇家橡树（Royal Oak）的“Royal Pop”联名腕表，引发收藏圈广泛争议。虽然有批评者戏称其为“官方山寨”，但分析人士将其比作此前斯沃琪与欧米茄（Omega）联名推出MoonSwatch的经典商业案例。目前在二级市场上，皇家橡树的入门门槛已被炒至30,000美元以上，而Code 11.59系列售价亦在20,000至25,000美元之间，远远脱离了中产收入增长水平，导致爱彼在品牌生态中极度缺乏可企及的入门级产品。 评论人士“Aero Grevner”指出，更关键的诱因在于法律风险防御：爱彼此前在2024年（日本）与2025年（美国）接连输掉了关于八角形表圈和表盘设计的关键商标诉讼。鉴于全球高仿表产业工艺成熟、几百美元即可买到高精度仿品，通过授权斯沃琪推出官方平价版本是一项战略性IP止损。表商与黄牛群体（gem changer）预测，该表款将于5月16日（周六）在线下实体店独占发售，官方零售价约为400美元且限购每人每天一块，二级市场预期转售溢价将达零售价的5倍至12倍。与此相对，市场上也存在标价高达37万美元的18K金Chrome Hearts定制款劳力士迪通拿（Rolex Daytona，但表盘疑似后改），反映出腕表市场两极分化的炒作现象。 特朗普访华与霍尔木兹海峡能源危机 (20:41) 美国总统特朗普赴北京与中国领导人展开高规格峰会，随行团队包括苹果公司Tim Cook、贝莱德Larry Fink、黑石Stephen Schwarzman、花旗Jane Fraser、思科Chuck Robbins、高盛David Solomon以及据报包含马斯克在内的顶级商界高管代表团。尽管科技界寄望于就GPU出口管制、人工智能供应链依赖与稀土矿产展开谈判，但峰会核心议题完全被涉及美以伊三国冲突、导致霍尔木兹海峡关闭的战局所笼罩。 霍尔木兹海峡承担着全球约五分之一的原油海运流量，其封锁造成国际能源市场陷入动荡，既切断了中国关键的平价原油补给，也严重损耗了美国的国内政治资本。特朗普宣称和平协议正处于“重大维持生命状态（major life support）”，并威胁美军可能重新启动“自由工程（Project Freedom）”海上护航及未公开的附加反制行动。除中东能源局势外，峰会经济议题还涉及中方承诺采购美国农产品、能源及航空航天产品，以及探索特殊投资通道；在AI治理方面，针对中国AI安全研究所（CAISI）的监管动向，节目也揭示了双方围绕以安全测试为名申请获取顶级AI模型权重展开的隐秘较量。 腾势出海、极客健康与垃圾邮件禁令 (10:52) 中国电动车龙头比亚迪（BYD）正通过旗下高端品牌腾势（Denza）全力打入欧洲高端汽车市场，并聘请前“詹姆斯·邦德”演员丹尼尔·克雷格（Daniel Craig）主演电影质感的宣传广告。腾势旗舰车型Denza Z预估售价在6万至14万美元之间，欧洲高配版本甚至高达13.4万欧元（常规系列在4万至6万美元之间），并专门搭载通过车内扬声器模拟传统跑车的合成发动机声浪，标志着比亚迪正突破过往“以丰田凯美瑞价格买法拉利性能”的平民标签，向超豪华电动车市场发起冲击。 在科技消费与政策层面，前Oculus创始人Palmer Luckey提出激进公共倡议，呼吁美国邮政（USPS）彻底立法取缔实体垃圾邮件。他比照FCC针对骚扰电话与2003年反垃圾邮件法（CAN-SPAM）的监管框架，批评准政府垄断机构为投递无用广告每年毁灭1亿棵树木，造成巨大资源浪费。此外，消费硬件极客群体中也出现了另类趋势，如将机械表与Whoop健康手环共用一条表带佩戴，或将Whoop监测到的原始体征数据直接输入大语言模型以自我排查睡眠呼吸暂停（sleep apnea）症状，折射出官方医疗审批滞后背景下由用户主导的DIY健康AI应用潮流。 本期节目集中反映了从底层算力物理架构到全球地缘宏观博弈的多维变迁：Cerebras以整片晶圆为赌注押注AI推理红利，爱彼在IP失利与二级市场断层中向快消模式低头，而中美在能源咽喉封锁下的仓促接触则暴露了全球经济对传统航道脆弱性的极度依赖。未决的核心悬念在于霍尔木兹海峡停火谈判能否真正落地以平抑通胀，以及Cerebras激进的单一晶圆制造模式能否在规模化量产交付中经受住各大前沿AI实验室的长期工程检验。

## Background 本期视频由知名科技谈话节目 TBPN 于 2026 年 5 月 22 日发布在 YouTube 平台。节目属于其 30 分钟精简栏目 Diet TBPN，由 John Coogan 和 Jordi Hays 主持，当期内容聚焦 Swatch 与爱彼联名、Cerebras 上调 IPO 募资与定价、以及特朗普携科技高管团访华等前沿科技与商业重磅热点。 ## TL;DR 本期节目首先深度解析了爱彼（AP）联手斯沃琪（Swatch）推出平价皇家橡树背后的商标维权败诉与商业破局逻辑，并探讨了由此引发的二级市场倒卖狂潮。随后，主播们聚焦丹尼尔·克雷格代言比亚迪高端品牌腾势进军欧洲的广告现象，以及 AI 芯片独角兽 Cerebras 上调 IPO 发行规模与指导价至 48 亿美元、首日市值预期冲破 500 亿美元的狂热行情。最后，节目探讨了特朗普携苹果、特斯拉等科技巨头高管访华斡旋地缘政治与供应链议题，并串讲了 Palmer Luckey 炮轰实体垃圾邮件、旧金山穿 AI 卫衣看房抬价等硅谷科技趣谈。 ## Atmosphere & style 整期节目节奏紧凑、风趣犀利，两位主持人保持着硅谷极客圈特有的松弛感与机智讽刺。对话充满密集的商业梗与调侃，从辨别手表的老派眼力一路聊到把黑匣子造飞机的芯片神话，主持人们在一搭一唱的默契拆台与调侃中，将原本硬核繁复的资本运作、芯片架构与地缘博弈化为一场兼具深度剖析与爆笑段子的科技脱口秀听觉盛宴。 ## Bulletin - [00:02](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=2s) 瑞士顶级制表工坊爱彼（Audemars Piguet）联手斯沃琪（Swatch）推出平价版皇家橡树设计腕表（代号 Royal Pop），定价约 400 美元且仅限线下发售，被视作爱彼因八角形表圈商标诉讼接连失利后主动出击的危机公关与破圈策略。 - [14:20](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=860s) 芯片独角兽 Cerebras 全面提升 IPO 发行规模与价格区间，募资额上调至 48 亿美元，机构认购倍数高达 20 倍（对应近 1000 亿美元资金争抢），首日收盘市值预期突破 500 亿美元。 - [10:52](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=652s) 比亚迪签约“007”丹尼尔·克雷格代言旗下高端品牌腾势（Denza）以全面发力欧洲市场，顶配车型定价高达 13.4 万欧元，重构中国新能源车的高溢价豪华心智。 - [20:39](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1239s) 特朗普携蒂姆·库克、拉里·芬克等多位顶级科技与金融领袖访华，核心聚焦由美以与伊朗冲突导致的霍尔木兹海峡石油危机斡旋，以及中长期中美 AI 供应链与 GPU 出口管制议题。 ## Key takeaways - [00:02](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=2s) 爱彼（AP）联手 Swatch 推出亲民款皇家橡树腕表，开辟低门槛入门级产品线，打破高端正品动辄数万美元带来的断层壁垒。 - [06:56](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=416s) 爱彼在 2024 年（日本）和 2025 年（美国）接连输掉商标维权诉讼，其八角形表圈与格纹表盘被裁定不具备排他性垄断特征，促使品牌通过官方平价联名直接反制市场仿制品。 - [09:46](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=586s) Swatch 与爱彼联名款采用全线下肉身排队、每人限购一块的发售策略，市场预期转手溢价将达零售价的 5 至 12 倍，引发黄牛转战二线城市门店的抢购热潮。 - [10:52](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=652s) 比亚迪聘请丹尼尔·克雷格作为高端子品牌腾势的代言人，摆脱过去的平价代步车标签，以对标超豪华车型的百万元级定价冲击欧洲市场。 - [14:20](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=860s) Cerebras 上调股票发行量至 3000 万股并将股价指导区间提至 150-160 美元，在 Transformer 仍为主流及 AI 推理加速需求爆发的背景下获得 20 倍超额认购。 - [19:28](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1168s) Cerebras 颠覆传统的全晶圆级芯片架构（WSE）虽然被调侃为简单粗暴的“黑匣子造飞机”，但在 AI Agent 大爆发时代成功实现了极高的推理吞吐与极低延迟。 - [19:28](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1168s) 早期风投机构 Benchmark 仍持有 Cerebras 超 20% 股权，若对标中国国内半导体企业的估值倍数，该笔投资极有可能创造风险投资史上的最高回报纪录之一。 - [20:39](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1239s) 特朗普率领由苹果、高盛、思科等顶级巨头组成的“科技高管考察团”访华，但首要议程被中东战局及承载全球两成海运石油的霍尔木兹海峡封锁危机所主导。 - [24:45](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1485s) 帕尔默·拉奇（Palmer Luckey）公开炮轰美国邮政署（USPS）的垄断特权，指出每年因无法退订的纸质垃圾邮件砍伐 1 亿棵树木，并呼吁出台实体信件版《反垃圾邮件法》。 - [26:52](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1612s) 密歇根大学捐赠基金早在 ChatGPT 爆火及微软入局前就向 OpenAI 押注了 2000 万美元，因该组织设立有限合伙营利实体而阴差阳错斩获巨额股权回报。 ## Detailed walkthrough ### [00:02](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=2s) 爱彼（AP）联手斯沃琪（Swatch）：当顶级奢表走向平民化 节目一开场，主持人约翰·库根（John Coogan）与霍尔迪·海斯（Jordi Hays）预告了当天长达两小时的豪华嘉宾阵容，包括 Match Group 首席执行官斯宾塞·拉斯考夫（Spencer Rascoff）、Long Lake Management 的亚历克斯·塔布曼（Alex Tubman），以及二手名表交易平台 Bezel 的联合创始人奎德（Quade）。而当天引爆话题的开场头条，则是瑞士顶级制表工坊爱彼（Audemars Piguet，简称 AP）与斯沃琪（Swatch）达成的震撼联名——双方合作推出了一款基于皇家橡树（Royal Oak）打造的亲民手表。主持人调侃这本质上就是官方亲自下场搞“山寨”，自己复刻自己的经典王牌，背后牵涉出极深层的商业逻辑。 随后节目放出了 Swatch 官方账号发布的先导宣传片。置身于 2026 年的技术环境下，两人甚至开始热烈辩论这段短片究竟是 AI 生成还是传统 CGI 制作。约翰感叹如今很难一眼看出二者区别，霍尔迪则笃定认为是“手工精雕细琢的纯正 CGI，就像传统的机械制表工艺一样，见不到半个 Transformer 架构的影子，完全是我祖父钟爱的老派作风”。目前官方只释放了悬念预告，全网充斥着各种以假乱真的民间渲染海报，真实表款究竟采用机械机芯还是石英机芯依旧成谜。约翰质疑机械机芯包含极其繁复的微型齿轮与组装工序，很难想象会压到几十美元的玩具价位，但霍尔迪立刻打趣道，约翰这是刚认清几百美元的手表其实同样拥有骇人听闻的暴利空间；毕竟在阿里巴巴等平台上，低至一两百美元就能买到具备类似机械构件的中国仿制爱彼或百达翡丽鹦鹉螺。约翰还提到腕表博主 Nico Leonard 近期在格雷厄姆·斯蒂芬（Graham Stephan）的《The Iced Coffee Hour》播客上玩辨表盲测，面对六枚混杂真假（甚至包括高仿百达翡丽）的腕表，凭借惊人眼力全部精准挑出。 - **嘉宾预告**：Match Group 首席执行官 Spencer Rascoff、Long Lake Management 的 Alex Tubman、Bezel 的 Quade。 - **重磅事件**：Audemars Piguet 联名 Swatch，推出代号或被称为“Royal Pop”的平价版皇家橡树设计腕表。 - **历史先例**：此前 Swatch 与同集团欧米茄（Omega）联名推出的 MoonSwatch（以超霸月球表 Moonwatch 为蓝本）大获成功，开辟了低门槛入坑奢表的新模式。 --- ### [03:28](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=208s) 二级市场泡沫、天价门槛与腕表圈的疯狂炒作 对于这次前所未有的联名，表圈舆论呈现两极分化，许多收藏家与代购歇斯底里地哀号“天塌了，赶紧 500 美元甩卖手中所有的正牌 AP，爱彼已死”。然而两位主持人分析指出，爱彼近年来一直在通过破圈联名与潮流明星合作走挑衅性路线，品牌本身生命力依旧极其强韧。更根本的行业现实在于，全球市场早就被各类高仿表充斥，既然防不胜防，品牌索性官方下场拥抱这股浪潮。与此同时，近年来高端腕表二级市场价格飞涨速度远远脱离了普通人的收入增长速度，爱彼官方定价甚至跟风二级市场水涨船高，直接导致进入爱彼生态的起步门槛高达 3 万美元，完全缺失让年轻玩家“由浅入深（walk, crawl, run）”的入门级梯队。 话题随后延伸到了其他令人咋舌的腕表联名与配货怪象。约翰展示了一款在二级市场引发轰动的改款表——18K 黄金打造的克罗心（Chrome Hearts）定制版劳力士迪通拿（Rolex Daytona），起拍开价竟高达 37 万美元，表盘甚至被指出可能是后期改装。网友 Dylan Dennis 借此狠损爱彼专柜令人窒息的配货潜规则：“我兴冲冲跑去排队买 Swatch 联名的皇家橡树，销售却告诉我必须先配货买一块 Code 11.59 联名款。”作为爱彼近年主推的新系列，Code 11.59 因轮廓缺乏皇家橡树那般鲜明的辨识度而饱受老藏家诟病，市场表现大幅逊色于劳力士的新推款式（如 Land-Dweller），沦为表圈长期调侃的“配货滞销王”。 ```mermaid flowchart TD accTitle: AP 联名 Swatch 商业动因分析 accDescr: 展示爱彼推出平价联名款背后的外部市场压力、品牌结构问题及商业对策。 subgraph External_Pressure["外部市场压力"] A["正品 AP 二级市场飙升<br/>门槛突破 30,000 美元"] B["全球仿冒品泛滥<br/>百元级高仿充斥网络"] end subgraph Internal_Dilemma["品牌内部痛点"] C["入门产品线断层<br/>年轻消费者无法入局"] D["Code 11.59 系列缺乏辨识度<br/>销量不及预期沦为配货"] end subgraph Strategy["Swatch 联名破局战略"] E["推出 Royal Pop 亲民款<br/>官方吸纳中低端流量"] F["制造排队效应与话题炒作<br/>打响年轻群体品牌认知"] end A --> C B --> E C --> E D -.-> F ``` - **价格断层**：正统皇家橡树正品二手行情动辄数万美元起步，而 Code 11.59 二级市场价格多在 2 万多美元区间。 - **极致奢华炒作**：第三方改款克罗心 18K 黄金劳力士迪通拿挂牌起步价叫到 370,000 美元。 --- ### [06:56](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=416s) 双表狂魔、智能硬件与 AP 专利战败的“危机公关” 除了名表本身，玩家们的佩戴癖好与科技穿戴设备的结合也成为趣味笑谈。霍尔迪分享了一张 Reddit 社区的热帖神图：一位表友居然用一条 Whoop 健康手环当表带，腕部内侧贴着 Whoop 传感器，外侧硬生生绑着一枚欧米茄机械表。约翰对此吐槽这种“双表合一”的戴法极其诡异，但承认 Whoop 等健康监测设备正在席卷硅谷与中产圈。约翰提到身边朋友通过大语言模型（LLM）分析自己的 Whoop 导出数据，竟然自行诊断出了睡眠呼吸暂停（sleep apnea），由于医疗监管审查严苛，官方功能迭代缓慢，反而催生了大量极客的“民间 DIY 科学”。霍尔迪则大开玩笑，称这种佩戴如果嫌累赘，不如直接做成脚踝监测器，戏称为“波士顿版 Fitbit”（暗指电子脚镣）。 讨论的高潮转向了爱彼此次动作背后隐藏的法律风暴。知识产权分析人士 Aero Grevner 指出，这次合作堪称“教科书级的危机公关”。爱彼在 2024 年于日本、2025 年于美国接连输掉了重磅商标维权官司，法院裁定爱彼标志性的“八角形表圈”和“格纹表盘”不足以构成排他性的合法垄断特征，爱彼实质上丧失了保护这套标志性外观的护城河。既然无法在法律上阻止外部模仿，爱彼干脆将核心资产授权给斯沃琪，推出官方且便宜的“Royal Pop”，用海量正规平价产品淹没仿制品。与此同时，黄牛与转售市场已闻风而动，知名炒家在社交网络宣称这是“所有无业游民今年最轻松斩获四位数美金的机会”——该联名表款定于 5 月 16 日星期六极速突击发售，发售官方定价在 400 美元左右，完全不设线上购买渠道，仅限线下实体店肉身排队，一场长达十几个小时的抢购大战即将上演。 - **健康数据 DIY**：用户绕过官方迟缓的医疗认证更新，通过将原生 Whoop 手环数据导入 LLM，成功自主诊断出睡眠呼吸暂停。 - **商标判决失利**：爱彼先后在 2024 年（日本法院）和 2025 年（美国法院）输掉商标官司，八角形表圈与经典表盘设计被判定不具备独占商标显著性。 - **发售规则**： - **发售时间**：5 月 16 日星期六（仅提前一周预热预告）。 - **官方定价**：约 400 美元区间。 - **销售渠道**：全线下实体店发售，无线上渠道，催生极大的二级市场溢价空间。 ### [09:46](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=586s) 联名腕表的倒卖淘金热与二级市场预期 围绕 Swatch 与爱彼（Audemars Piguet）联名表款的发售，品牌方制定了极其严苛的限购政策：每家门店、每人、每天仅限购买一块，而不是两块。这意味着想要规模化套现的黄牛党必须采取“按人头算”的策略，动员尽可能多的人肉排队大军。市场上关于这只表的二级溢价预期十分惊人，有人预测转手价格可能会冲到零售价的 5 倍甚至 12 倍。只要有耐心在店门口排队抢到，转手挂到网上就能迅速大赚一笔。 圈内人士还针对排队地点给出了实用建议：务必避开纽约 SoHo、时代广场、伦敦以及新加坡等竞争惨烈的热门核心旗舰店，转战密歇根州特洛伊（Troy）、宾夕法尼亚州普鲁士王市（King of Prussia）、加州卡诺加公园（Canoga Park）乃至夏威夷檀香山等相对次级的城市。“哪怕专程飞一趟檀香山去抢都值。”尽管网络上充斥着唱衰的声音，但主播们一致认为这款表必将大火，因为普通大众极为青睐类似卡西欧 G-Shock 风格的耐用与趣味设计，这为初涉钟表收藏的年轻人提供了一个绝佳且门槛适中的切入点。 - **限购规则**：每人每店每天限购 1 枚，杜绝批量购入。 - **二手溢价预期**：预计二级市场交易价可达零售价的 5 倍至 12 倍。 - **推荐排队地点**：避开一线都市核心商圈，建议前往 Troy、King of Prussia、Canoga Park 或 Honolulu 等次级城市门店。 --- ### [10:52](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=652s) 比亚迪携手“邦德”丹尼尔·克雷格：腾势豪车进军欧洲 比亚迪（BYD）为了进军欧洲高端市场，正式签下“007 詹姆斯·邦德”的扮演者丹尼尔·克雷格（Daniel Craig）作为旗下高端品牌腾势（Denza）的全新代言人。这则广告在社交网络上引发了广泛赞叹，评论员直呼“中国车企借用詹姆斯·邦德发力欧洲市场，世界真的变了”。广告画面极具质感，克雷格在车内深情念出独白：“严冬过后必有暖春，昔日的自我与过往的身份终将转变，生活不正是在要求我们走出阴影、拥抱新生吗？” ```mermaid flowchart LR accTitle: 比亚迪腾势借势丹尼尔·克雷格进军欧洲 accDescr: 展示比亚迪利用克雷格的邦德银幕光环树立腾势豪华品牌形象并进军欧洲市场的流程 A["邦德银幕光环<br/>(优雅/权威/硬汉)"] --> B["代言比亚迪腾势<br/>(Denza 豪华 EV)"] B --> C["打破平价代步标签<br/>(进军欧洲高端市场)"] C --> D["对标超豪华品牌<br/>(单车定价最高达 14 万美元)"] ``` 在播放广告片时，演播室里爆发出不少趣味插曲。主播们先是对车后座一只佩戴特殊安全带的宠物狗忍俊不禁，接着敏锐地听到视频中传出类似内燃机的轰鸣声，甚至专门倒带十秒仔细确认：“这难道不是内燃机的轰鸣吗？难道是通过车内扬声器模拟出来的声浪？现在的比亚迪让人分不清到底纯电还是混动。”谈到克雷格本人，主持人们感叹他彻底将“邦德光环”变现到了极致——无论是早期的黑帮犯罪片《夹心蛋糕》（Layer Cake），还是在《利刃出鞘》（Knives Out）中饰演的名侦探布兰科（Benoit Blanc），他身上始终散发着无可替代的威严与酷劲，成为汽车品牌梦寐以求的代言面孔。 - **欧洲售价**：腾势在欧洲的顶配标价高达 13.4 万欧元/美元；其中 Denza Z 预计定价在 6 万至 14 万美元之间，其他走量主力车型集中在 4 万至 6 万美元区间。 - **品牌转型**：以往比亚迪在海外常被视为“以丰田凯美瑞的价格买法拉利的性能”，而腾势正在真正建立起高溢价的豪华车心智。 --- ### [14:20](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=860s) Cerebras 上调 IPO 定价与规模：AI 推理算力迎来百亿狂欢 主播调侃道，丹尼尔·克雷格下一次代言最应该去接 Cerebras 的超级碗广告，在片中抱怨 AI 推理速度太慢，最后换上 Cerebras 芯片瞬间解决问题。根据路透社的最新披露，芯片独角兽 Cerebras 的上市进程正如火如荼：公司不仅将发行规模从原先的 2800 万股上调至 3000 万股，价格指导区间更是从每股 115–125 美元大幅提升至 150–160 美元，募资总额从 35 亿美元跃升至 48 亿美元。市场认购极其狂热，机构认购倍数高达 20 倍，换言之，面对不足 50 亿美元的融资额，市场涌入了近 1000 亿美元的疯狂争抢。科技博主 Ben Thompson 更是直言：“如果你想挑一个最佳时机让芯片公司上市，2026 年 5 月几乎无可匹敌。” 主播们提到，早先市场上围绕 Cerebras 存在诸多看空噪音（FUD），比如芯片从流片到量产耗时极长、初代产品缺乏灵活性，以及过度依赖单一客户导致客户集中度过高。此外，早期还有人担忧其全晶圆架构过度针对 Transformer 做了专用优化，一旦底层算法抛弃注意力机制就会沦为废铁。然而现实证明，Transformer 架构地位依旧稳固，而在 AI Agent 大爆发的时代，推理速度和推理成本已成为生死攸关的胜负手。在实测体验中，用户在 Codex Desktop 中选择搭载 Cerebras 算力的 GPT-5.3 Spark 模型，以往需要等待数分钟甚至数小时的复杂任务，现在瞬间就能生成整页高质量回答。除 OpenAI 深度绑定外，亚马逊 AWS 等算力巨头也已进场部署。预测市场 Polymarket 显示，Cerebras 在纳斯达克挂牌（股票代码 CBRS）首日收盘市值预计将突破 500 亿美元大关，较原先 260 亿美元的估值翻了一番。 - **IPO 关键条款调整**： - 发行股数：由 2800 万股增加至 3000 万股。 - 股价指导区间：由 115–125 美元上调至 150–160 美元。 - 募资规模：从 35 亿美元攀升至 48 亿美元。 - 认购倍数：超额认购达 20 倍（对应近 1000 亿美元的认购需求）。 - 上市地点与代码：纳斯达克挂牌，代码为 `CBRS`；市场预期首日市值或超 500 亿美元。 - **颠覆性全晶圆级芯片架构（WSE）**：与传统芯片制造“先切片成数百个小晶片再封装互连”的方案不同，Cerebras 直接利用整张 300 毫米硅晶圆制造单颗芯片，集成 4 万亿个晶体管、90 万个 AI 专用计算核心，并拥有高达 PB 级每秒（Petabits per second）的超高片上片间互连带宽与海量片上高速缓存。 ### [19:28](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1168s) Cerebras 估值狂想与“黑匣子造飞机”的芯片神话 谈及 AI 芯片黑马 Cerebras，主播们对风险投资机构 Benchmark 的独到眼光惊叹不已。据报道，Benchmark 至今仍持有 Cerebras 超过 20% 的股份。如果将中国上海对摩尔线程（Moore Threads）和寒武纪（Cambricon）给出的估值作为参照物，“哪怕 Cerebras 未来的交易估值只有它们的一半，也会在不到两年内飙升至 5000 亿美元以上”。这种回报率足以直接击穿传统的风险投资回报模型，若 Benchmark 选择坚守持股，极有可能创造风投史上收益最恐怖的一期基金。 节目中分享了一条调侃 Cerebras 技术路线的爆笑推文：大家完全可以想象 Cerebras 的诞生起源——某位对技术一窍不通的婴儿潮一代管理者信口开河：“行吧，但你们为什么不能直接往这颗芯片上怼个 50GB 的 L3 缓存呢？”而被当场架在火上烤的工程师只能无奈妥协：“呃……我想大概也行吧。”这种简单粗暴的技术直觉引发了网友的共鸣，一位历史系网友在评论区回忆自己也曾给搞计算机视觉的高材生室友出过类似馊主意：“我们为什么不把这个功能模块直接焊死在另一个上面呢？你有没有想过干脆用飞机的‘黑匣子’材料造一整架飞机？”可 Cerebras 偏偏就把这种看似荒诞的“黑匣子造飞机”做成了现实，而且这架飞机还能以“10 马赫”的恐怖速度狂飙。 - **Benchmark 持股**：依然持有 Cerebras 20% 以上的股权，有望斩获风投史上第一神级基金。 - **中国芯片参照系**：对标摩尔线程与寒武纪在二级市场的估值倍数，推演 Cerebras 的天花板直指 5000 亿美元大关。 --- ### [20:39](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1239s) 特朗普访华“科技高管考察团”与霍尔木兹海峡阴影 话题转到了国际地缘政治：唐纳德·特朗普本周将与中国领导人在北京会晤。令人瞩目的是，这趟行程被外界戏称为“豪华实地考察团（field trip）”，随行人员名单几乎汇聚了美国顶尖科技与金融巨头：苹果 CEO 蒂姆·库克（Tim Cook）、贝莱德的拉里·芬克（Larry Fink）、花旗集团的简·弗雷泽（Jane Fraser）、思科的查克·罗宾斯（Chuck Robbins）、高盛的大卫·所罗门（David Solomon），甚至据传埃隆·马斯克（Elon Musk）也在队伍之中。 然而，商界领袖能否在会场插得上话还是个未知数，因为整个峰会都被中东战事及伊朗危机所笼罩。据《华尔街日报》报道，美以对伊朗的冲突引发霍尔木兹海峡封锁，这处咽喉水道承载着全球约五分之一的石油运输，冲突不仅重创全球经济，也在持续消耗特朗普的国内政治资本。虽然特朗普当天上午声称和平协议目前处于“依靠生命维持系统（on major life support）”的垂危状态，但他此行意在借力严重依赖伊朗廉价石油的中国共同促成停火。 ```mermaid flowchart TD accTitle: 中美高层峰会议程结构 accDescr: 展示地缘政治危机如何压制科技与经贸谈判议程 subgraph Geopolitics["首要地缘政治议程"] Iran["伊朗危机与霍尔木兹海峡封锁<br/>（原油断供 / 全球供应链受阻）"] Peace["斡旋停火协议<br/>（协议目前处于'生命维持'状态）"] Iran --> Peace end subgraph Economy["经贸与科技诉求"] Trade["传统经贸采购<br/>（农产品、能源、航空航天）"] Tech["核心科技关切<br/>（GPU 出口管制、AI 供应链、稀土）"] end Peace -.->|"压制并决定谈判窗口"| Economy ``` 科技界则期盼战火平息后，谈判能迅速推进到 GPU 出口管制、AI 供应链以及稀土供应等命脉议题上。围绕中美 AI 监管的走向，主播们戏谑地模仿起可能出现的离谱场景：中国会不会直接开出条件要求“把 Mythos 也开放给我们，给完全访问权限方便我们快速蒸馏；或者更省事点，为什么不干脆把模型权重和跑模型所需的 GPU 全送过来？我们好替你们‘检查合规’”。不过正如大家所指出的，库克和马斯克这帮大腕绝不可能跨过大洋去充当提倡过度安全限制的“安全派（safety-pilled）”。 - **石油命脉**：霍尔木兹海峡承载全球约 20% 的石油海运，地缘冲突刺激油价大幅攀升。 - **自由行动（Project Freedom）**：特朗普威胁将重启美方护航行动，并语带威胁地宣称将包含“其他手段”。 --- ### [24:45](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1485s) Palmer Luckey 炮轰垃圾邮件：美国邮政的实体“投毒”与 CAN-SPAM 法案 科技圈名流帕尔默·拉奇（Palmer Luckey）在社交媒体上公开发难，强烈呼吁美国邮政署（USPS）全面禁止纸质垃圾邮件（junk mail）。帕尔默指出，联邦通信委员会（FCC）早已严令禁止骚扰电话，电子邮件也有 2003 年《反垃圾邮件法》（CAN-SPAM Act）严格监管，然而实体邮箱里绝大多数的信件却全是毫无用处的传单废物，为此每年要砍伐 1 亿棵树木。 主播在热烈赞同的同时反思道，自己以往把邮箱没有垃圾邮件全归功于 Google 的垃圾箱过滤，其实《CAN-SPAM 法案》确实起到了相当大的震慑作用，反倒是如今未指名的骚扰电话和短信呈泛滥之势。帕尔默直言这是一种不可理喻的荒谬垄断：美国赋予了一个准政府机构对信件投递的绝对专营权，结果该机构却利用这种特权，违背居民意愿将大量废纸塞入千家万户。如果换作联邦快递（FedEx）或 UPS 这么干，公众早就将其撕碎；如果按全美每个成年人每天浪费 30 秒翻检、剔除这些垃圾邮件来计算，造成的国民生产力损失每年就高达十几亿美元。对此，主播打趣给出了硅谷式解法：买个像 Earth Class Mail 那样的实体信箱扫描服务，转成电子件后直接接入 Claude 等大模型，让 AI 替你把关决定哪封信配被你看到。 --- ### [26:52](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1612s) 穿 OpenAI 卫衣抬房价与密歇根大学的百倍神投 随后节目分享了一则关于旧金山房地产的经典段子：“为旧金山房东及买家提供新服务：我将身着 OpenAI 或 Anthropic 的官方周边卫衣出席你们的房屋开放日（Open House），现场假装看房抬高抢购预期，事成仅收 5% 佣金。”这种被戏称为“吊灯房地产（Chandelier Real Estate）”的骚操作引发热议：即便真实需求寥寥，普通买家一看到现场站着顶尖 AI 实验室的“新贵码农”，往往也会吓得咬牙开出极具攻击性的超高溢价。 话题进一步延伸到了早期的 AI 资本博弈：在 ChatGPT 席卷世界之前、在微软掏出 10 亿美元巨资下注之前、在所谓几千亿 IPO 计划之前，密歇根大学（University of Michigan）的捐赠基金竟然早在早期就为 OpenAI 押注了 2000 万美元。现场主持人打趣着指导后台导播 Tyler 切换屏幕素材，并调侃这笔投资的戏剧性：密歇根大学最初可能只是以慈善资助心态接触 OpenAI 这个非营利组织，中途突然发现他们成立了有限合伙营利实体，于是顺水推舟决定“那不如换成股权投资试试吧”，阴差阳错押中了世纪大奖。 节目的尾声切入了一则令人啼笑皆非的科技家庭代沟趣闻：一位网友吐槽自己的母亲科技素养低到令人发指——老妈原本只是想在 Google 上搜搜“什么是功能饮料”，在手机上一通误触瞎点后，竟然稀里糊涂地直接在 Google 地图上把自家住宅注册认证成了一家对外营业的实体商铺。 ### [28:48](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1728s) AI 扩散速度的调侃与节目尾声 在节目的最后几秒，主持人们以一贯轻松幽默的语调收尾，顺手调侃了一个关于包裹开箱与具体收件地址的梗，提到要把确切地址留在名为“盒子里面是什么”（"what is in the box"）的名义之下。紧接着，话题瞬间切回科技圈最火热的议题，主持人打趣地抛出一句反讽：“还有人觉得 AI 会迅速渗透普及开来”（"And people think AI will diffuse quickly"），搭档随即斩钉截铁地附和道：“这可有待商榷”（"Debatable"）。两人在对 AI 落地扩散速度的打趣中，展现了对技术狂热与现实落地之间差距的一丝审慎与戏谑。 伴随着轻松的对话节奏，节目迅速进入收尾呼吁。主持人们提醒听众和观众在 Apple Podcasts 以及 Spotify 平台上为节目留下五星好评，并号召大家前往官方网站订阅电子简报（Newsletter），在简短干脆的一声告别中结束了当天的节目。 - **互动与支持渠道**：听众可在 Apple Podcasts 与 Spotify 为节目评分（5星好评）。 - **官方网站资讯**：访问 TBPN.com 即可注册订阅官方最新通讯 Newsletter。 ## Notable quotes > [00:02](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=2s) 手工精雕细琢的纯正 CGI，就像传统的机械制表工艺一样，见不到半个 Transformer 架构的影子，完全是我祖父钟爱的老派作风 — 霍尔迪·海斯（Jordi Hays） > [03:28](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=208s) 天塌了，赶紧 500 美元甩卖手中所有的正牌 AP，爱彼已死 > [03:28](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=208s) 我兴冲冲跑去排队买 Swatch 联名的皇家橡树，销售却告诉我必须先配货买一块 Code 11.59 联名款。 — Dylan Dennis > [06:56](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=416s) 所有无业游民今年最轻松斩获四位数美金的机会 > [10:52](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=652s) 严冬过后必有暖春，昔日的自我与过往的身份终将转变，生活不正是在要求我们走出阴影、拥抱新生吗？ — 丹尼尔·克雷格（Daniel Craig） > [14:20](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=860s) 如果你想挑一个最佳时机让芯片公司上市，2026 年 5 月几乎无可匹敌。 — Ben Thompson > [19:28](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1168s) 哪怕 Cerebras 未来的交易估值只有它们的一半，也会在不到两年内飙升至 5000 亿美元以上 > [19:28](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1168s) 行吧，但你们为什么不能直接往这颗芯片上怼个 50GB 的 L3 缓存呢？ > [20:39](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1239s) 依靠生命维持系统（on major life support） — 唐纳德·特朗普（Donald Trump） > [26:52](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1612s) 为旧金山房东及买家提供新服务：我将身着 OpenAI 或 Anthropic 的官方周边卫衣出席你们的房屋开放日（Open House），现场假装看房抬高抢购预期，事成仅收 5% 佣金。 > [28:48](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1728s) And people think AI will diffuse quickly — 约翰·库根（John Coogan） > [28:48](https://www.youtube.com/watch?v=l4Yg8lNxWrc&t=1728s) Debatable — 霍尔迪·海斯（Jordi Hays） ## Mentioned 爱彼（Audemars Piguet / AP）, 斯沃琪（Swatch）, 约翰·库根（John Coogan）, 霍尔迪·海斯（Jordi Hays）, 斯宾塞·拉斯考夫（Spencer Rascoff）, 亚历克斯·塔布曼（Alex Tubman）, Match Group, Long Lake Management, Bezel, 奎德（Quade）, 皇家橡树（Royal Oak）, MoonSwatch, 欧米茄（Omega）, Nico Leonard, 格雷厄姆·斯蒂芬（Graham Stephan）, The Iced Coffee Hour, 百达翡丽（Patek Philippe）, 克罗心（Chrome Hearts）, 劳力士（Rolex）, 劳力士迪通拿（Rolex Daytona）, Code 11.59, Land-Dweller, Dylan Dennis, Whoop, Aero Grevner, Royal Pop, 比亚迪（BYD）, 腾势（Denza）, 丹尼尔·克雷格（Daniel Craig）, 詹姆斯·邦德（James Bond）, Denza Z, Cerebras, Ben Thompson, Codex Desktop, OpenAI, GPT-5.3 Spark, 亚马逊 AWS, Polymarket, Benchmark, 摩尔线程（Moore Threads）, 寒武纪（Cambricon）, 唐纳德·特朗普（Donald Trump）, 蒂姆·库克（Tim Cook）, 拉里·芬克（Larry Fink）, 简·弗雷泽（Jane Fraser）, 查克·罗宾斯（Chuck Robbins）, 大卫·所罗门（David Solomon）, 埃隆·马斯克（Elon Musk）, 苹果（Apple）, 贝莱德（BlackRock）, 花旗集团（Citigroup）, 思科（Cisco）, 高盛（Goldman Sachs）, 帕尔默·拉奇（Palmer Luckey）, 美国邮政署（USPS）, 联邦通信委员会（FCC）, CAN-SPAM 法案, 联邦快递（FedEx）, UPS, Earth Class Mail, Claude, Anthropic, 密歇根大学（University of Michigan）, 微软（Microsoft）, Google, Google 地图, Tyler ## Video description 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. 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 [Source](https://www.youtube.com/watch?v=l4Yg8lNxWrc)

[0:02] 霍尔迪·海斯（Jordi Hays）: 手工精雕细琢的纯正 CGI，就像传统的机械制表工艺一样，见不到半个 Transformer 架构的影子，完全是我祖父钟爱的老派作风 [3:28] 天塌了，赶紧 500 美元甩卖手中所有的正牌 AP，爱彼已死 [3:28] Dylan Dennis: 我兴冲冲跑去排队买 Swatch 联名的皇家橡树，销售却告诉我必须先配货买一块 Code 11.59 联名款。 [6:56] 所有无业游民今年最轻松斩获四位数美金的机会 [10:52] 丹尼尔·克雷格（Daniel Craig）: 严冬过后必有暖春，昔日的自我与过往的身份终将转变，生活不正是在要求我们走出阴影、拥抱新生吗？ [14:20] Ben Thompson: 如果你想挑一个最佳时机让芯片公司上市，2026 年 5 月几乎无可匹敌。 [19:28] 哪怕 Cerebras 未来的交易估值只有它们的一半，也会在不到两年内飙升至 5000 亿美元以上 [19:28] 行吧，但你们为什么不能直接往这颗芯片上怼个 50GB 的 L3 缓存呢？ [20:39] 唐纳德·特朗普（Donald Trump）: 依靠生命维持系统（on major life support） [26:52] 为旧金山房东及买家提供新服务：我将身着 OpenAI 或 Anthropic 的官方周边卫衣出席你们的房屋开放日（Open House），现场假装看房抬高抢购预期，事成仅收 5% 佣金。 [28:48] 约翰·库根（John Coogan）: And people think AI will diffuse quickly [28:48] 霍尔迪·海斯（Jordi Hays）: Debatable

[0:02] not a transformer in sight, the way my grandfather used to like it. [14:20] Ben Thompson: The Inference Shift [14:20] being a chip company in May 2026 is hard to beat. [16:22] Attention Is All You Need [16:22] Non-technical manager: Why can't you just put 50 gigs of L3 cache on this chip? [16:22] Engineer: I guess you could. [20:41] Donald Trump: As of this morning, the peace deal was, according to Trump, on major life support... Is that good? It's better than it being dead. [20:41] Donald Trump: Project Freedom [20:41] Send us the weights... We'll inspect it and make sure it's okay for the Chinese population. [24:45] what is in an energy drink [24:45] what is in the box [24:45] And people think AI will diffuse quickly. Debatable.

[0:02] We got a banger show for you today, [0:03] folks. We got a lot of guests coming on. [0:05] We got 2 hours of back-to-back [0:07] interviews with everyone from Spencer [0:09] Rascoff, the CEO of Match Group, to Alex [0:12] Tubman of Long Lake Management. Quade's [0:14] coming on from Bezel. He's going to take [0:16] us deeper on our first story, which is, [0:19] of course, Audemars Piguet AP is [0:22] partnering with Swatch to launch a watch [0:25] that's [0:26] uh you could call it a knockoff of the [0:28] Royal Oak. It's certainly [0:30] >> It's not a knockoff cuz it's official. [0:31] It's from the actual They knocked [0:33] themselves off. And there's a bunch of [0:34] interesting business implications of why [0:36] they did this, what it means, what will [0:38] happen. Uh let's pull up the video of [0:40] the launch first. And you got to tell [0:43] me, do you think this is AI or CGI? [0:46] Because it's a launch video, and in [0:48] 2026, it's hard to tell the difference. [0:50] So, we're going to play the [0:52] This is from the official Swatch [0:54] account. This is something that looked [0:55] like uh like a fake clickbait video, but [0:58] it's real. Here it is. What do you [1:00] think? [1:01] AI or CGI? [1:05] >> CGI. Precision handmade CGI. Just like [1:09] >> Yes. Yes. [1:10] >> watches. [1:10] >> Not a transformer in sight, the way my [1:13] grandfather used to like it. [1:15] Uh there will be interesting pushbacks. [1:17] They're like, "Oh, this this movie is [1:19] awesome. It only used CGI. [1:21] No AI involved whatsoever." Wait, it's [1:23] So, is it a manual watch? Is it a [1:26] >> I think it's unclear. Right. So, what [1:28] what there's been a massive amount of of [1:31] uh everyone has basically created [1:33] somewhat realistic-looking posters for [1:36] these. And the actual watch has not been [1:38] revealed at all. [1:38] >> And that's the only thing that's been [1:39] Just based off of the fact that Swatch [1:42] did a collaboration with a luxury watch [1:45] brand, the MoonSwatch, which was based [1:48] on the Moonwatch, right? [1:50] >> Oh, yeah. I forgot about that. [1:51] >> Yeah. And so, the MoonSwatch was [1:53] successful and was a uh a more [1:55] accessible entry point to the Moonwatch, [1:58] which is from Omega, which is sort of in [2:00] that like Rolex tier. And I don't know [2:04] if the MoonSwatch what how they saved [2:07] costs because there's a certain amount [2:09] of cost that just goes into making a [2:12] a movement a mechanical movement. It's a [2:14] lot of small pieces. Sure, you can put [2:16] it on a manufacturing line and press [2:18] them and stuff. [2:19] >> John learning that couple hundred bucks [2:21] watches have egregious profit margin. [2:24] >> I know they have egregious profit [2:25] margins, but still like like the [2:28] the work to put together all the gears [2:30] and manufacturing it like I would be [2:32] surprised if if if it's a mechanical [2:34] watch is it really going to be like $5? [2:37] No, it's not a happy meal toy. [2:38] >> So there's there's the fake Chinese [2:42] version of all of these popular watches [2:44] Nautilus APs etc. You can get them like [2:47] you could go on Alibaba get them for in [2:50] the low hundreds of dollars, right? And [2:52] so it is totally possible to put [2:54] together a watch that has a lot of the [2:56] same componentry. [2:58] >> Yeah, I saw Nico Leonard on who's a [3:00] great YouTube watch reviewer fun very [3:02] fun creator on the Ice Coffee Hour with [3:04] Graham Stephan and they gave him is it a [3:07] fun little [3:08] game that they they're getting really [3:09] good at playing games with their guests [3:11] on this podcast. So they give him six [3:13] watches and they tell him that three of [3:15] them are knockoffs and three of them are [3:19] are authentic and he has to guess and he [3:21] nails it. He gets all of them correct. [3:23] >> Yeah. [3:23] >> Even though some of them were very [3:25] convincing fakes especially of Patek. [3:28] >> So a lot of people are saying this is [3:30] bad. [3:31] >> It's over. They're saying sell [3:33] everything freak out sell everything [3:35] sell all your APs for $500. Rest in [3:37] peace. [3:37] >> AP has [3:39] you know consistently done things over [3:41] the last few years to that were [3:43] provocative, right? Some of the various [3:45] partnerships they've had with talent, [3:48] right? Celebrities etc. have been [3:50] somewhat provocative. Uh but overall I [3:54] the brand seems healthy, right? It's not [3:56] going to be for everyone. I think in [3:57] some ways this decision could be seen [4:00] and again I'm no watch expert, but [4:01] >> Yep. [4:02] >> uh fakes are flooding the market [4:04] globally [4:05] >> Mhm. [4:05] >> and why not just lean into that [4:08] basically? And uh the other thing is [4:10] they don't have any entry level like the [4:12] gap between, you know, what I'm sure [4:15] this this um this Swatch AP will end up [4:18] retailing for far more or sorry not [4:21] retail, but secondary will be far more [4:22] than than whatever it goes out at. But [4:25] the gap even then between that and a [4:27] real Royal Oak will be immense, right? [4:30] >> And the gap has been growing because the [4:31] aftermarket prices have been increasing [4:33] >> And they [4:34] >> faster than incomes. [4:35] >> They basically were like, oh, this is [4:37] what our watches are worth aftermarket. [4:39] We should price there, right? So [4:41] >> And so you quickly wound up with like it [4:42] to get in the game you're in you're at [4:44] 30k. And that's just a lot and there's [4:46] not as much of a like a walk crawl run [4:49] to get into the ecosystem that some [4:50] brands have. AP certainly has not had [4:52] that. Now this is [4:54] >> with the code. You go with the code. [4:56] >> Aren't those like 60 now? [4:57] >> Oh, they are? [4:58] >> I think they're really expensive. No, [4:59] no, I think they're really pricey. I [5:01] don't know. [5:01] Um [5:02] but uh yeah, maybe they'll bring down [5:03] the price on that. Live Monitor says lit [5:05] spend half a million dollars or $500 and [5:08] you get something that looks pretty [5:10] similar. Of course very different [5:11] materials, very [5:12] >> You can get a [5:13] uh Code 11.59 for low to mid 20s. [5:16] >> 20s, okay. But there's another watch [5:19] collab that I want your reaction to. You [5:21] got to see this. This one's going to fly [5:24] off the shelves, potentially selling [5:25] more units than the Swatch AP collab. [5:28] It's the [5:30] Rolex collab. We can pull this up. [5:32] >> Something you've never seen before and [5:34] you won't see again. A Chrome Hearts [5:37] Rolex Daytona 18 karat gold. Starting [5:41] offer 370,000, [5:45] okay? If you want it, there's only one [5:47] way to get it. I've never seen this [5:49] watch before, neither have you. That's [5:51] why I'm showing it to you. [5:53] This is the first time I've ever seen [5:56] this watch before. Official Rolex over [5:57] here telling me that the dial might be [6:00] after market. [6:01] I don't know, you figure that out. [6:02] >> know. I don't I don't follow it. It's [6:04] funny [6:06] somebody on X over the weekend was [6:10] assuming that I was into Chrome Hearts [6:11] because I joked a lot about Chrome [6:13] Hearts, but no, I'm not enough of an [6:15] expert [6:16] to [6:17] >> No, that's Dylan. [6:19] Dennis is playing off of your joke [6:21] saying I tried to buy the Swatch AP [6:23] Royal Pop collab, but they told me I had [6:25] to buy this collab first and it's the [6:27] Code 11:59. If you're not familiar with [6:29] the Code 11:59, it's [6:30] newest watch from AP, but it has a less [6:34] distinct silhouette than the Royal Oak [6:36] and has been not loved by the biggest [6:39] fans of AP broadly and so it has been [6:42] under selling probably relative to the [6:45] the new Rolex. [6:46] >> Which which one's the new Rolex? The [6:48] Land Dweller, which has been I think [6:51] selling very well and [6:53] it this one has been the Code 11:59 has [6:55] not been doing as well. [6:56] >> Let's head head over to Reddit and check [6:58] out what they're doing there. [6:59] >> What are they doing there? [7:00] >> Somebody has figured this out ages ago. [7:03] I want your reaction, John. Omega on one [7:05] side, but the Omega is strapped to the [7:08] wrist with a Whoop band. [7:09] >> Yeah, Whoop on the inside, Omega on the [7:12] outside. It just doesn't feel right. It [7:14] feels a little unnatural to me, but I [7:16] did I have seen a lot of people wearing [7:17] the Whoop bands lately and I think that [7:19] there's some remarkable data. I was [7:20] talking to someone who connected their [7:23] their Whoop data and and found out that [7:26] they had sleep apnea by analyzing it [7:29] with an LLM and [7:31] which is something you would expect [7:32] Whoop to be doing on their side, but for [7:34] regulatory reasons, it might be slower [7:36] for Whoop to roll out that feature of [7:38] like health monitoring and so So a lot [7:40] of like DIY science that comes from it. [7:42] So I don't know. It's weird no matter [7:44] what because if you have a Whoop on one [7:45] end and then on one hand and then a [7:48] watch on the other, that's an odd [7:49] choice. I feel like isn't there an [7:52] opportunity to put the Whoop band [7:53] somewhere more discreet? Like even like [7:55] a chest monitor would be less weird. [7:58] >> you can. [7:58] >> The Aura ring is not very intrusive, but [8:01] they should really make the [8:02] What do they call it? The [8:04] the Boston Fitbit? It's the ankle [8:06] monitor. [8:09] I can say I can say whatever I want. [8:11] But there was some debate over AP's [8:13] motivations. Aero Grevner says, "As an [8:15] IP nerd, I love this. AP's trademark [8:18] loss means they lost the moat around [8:20] their octagonal bezel and their dial. So [8:23] what do they do? They lounge they [8:25] license its crown jewels that to Swatch [8:28] for a flood of legit affordable royal [8:31] pop pieces, a masterclass in damage [8:34] control." This is on this news that AP [8:36] lost a trademark fight in Japan in 2024 [8:40] and in the US in 2025. The courts ruled [8:43] that the bezel and the dial aren't [8:45] distinctive enough to legally own, but [8:48] there's some pushback in the community. [8:49] No. [8:49] >> Here's the thing, they they have managed [8:52] I think maintain a trademark around the [8:54] octagon. [8:55] >> Yeah, well there's a trade dress, so you [8:57] can never do a full [8:58] a full knockoff of a direct product, but [9:01] they couldn't lock down the idea of an [9:04] an octagon. That was like simply too [9:06] much. I don't know how much of this was [9:07] was damage control around the [9:09] intellectual property, but it's [9:11] certainly certainly an interesting [9:12] thing. And there's also some people [9:14] thinking that this is a way to make [9:15] money. You see this gem changer said, [9:17] "Every unemployed guy with a group chat [9:19] of equally unemployed friends, this post [9:21] is for you. The [9:24] royal pop drops Saturday, May 16th." [9:26] They're coming out with this quickly, [9:27] just one week teaser and then it's out. [9:29] >> Smart. [9:30] >> He says, "This is the easiest [9:31] four-figure week you'll have all year if [9:33] you're willing to do something that [9:34] resembles work resembles work for 14 [9:37] hours if you're willing. [9:39] Uh let me lay out exactly what to do. [9:41] Retail is in the $400 range. It's in [9:44] store only. There's no online sales. [9:46] They're limiting it to one person one [9:48] piece per person per store per day, not [9:51] two. So, your hustle is per warm body, [9:54] not per pair. And then uh there he [9:57] shares some uh some expect some [9:59] expectation about where this might [10:01] trade, maybe uh 12x retail, maybe 5x [10:04] retail. And so, if you go and you wait [10:06] in line and you buy this and then you [10:09] sell it online, you might be able to [10:11] make a pretty penny pretty quickly, but [10:12] you'll need a bunch of friends. And he [10:13] gives a bunch of advice on where you [10:15] should go. Avoid Soho, Times Square, [10:17] London, Singapore. You got to go to [10:20] secondary cities, Troy, Michigan, King [10:22] of Prussia, uh Canoga Park, Honolulu. [10:26] Going to Honolulu just for this. Um [10:29] >> I think this is I think this is going to [10:30] be a hit. I think the haters are wrong. [10:32] People love G-Shocks. Sort of a [10:34] combination. [10:35] >> I think so, too. It'll do well. It just [10:36] seems like a fun a fun watch. And I and [10:38] I do think it's a nice entry point for [10:40] someone who's getting into watches. Like [10:41] start with this and then get something [10:43] else. It's like a striver, like, you [10:44] know, uh [10:46] uh it's a point along a curve, which I [10:48] think will be popular. And it's also [10:49] just like a lot of fun and and it'll [10:51] look good. Anyway, [10:52] >> Head over to China. [10:53] >> What are they doing over in China? [10:54] They're [10:55] >> BYD [10:56] >> BYD [10:56] >> got Daniel Craig as the new face. [10:59] >> Build your dreams. They're building [11:00] their dreams for the Denza luxury EV. [11:04] James Wood says it's an amazing ad. [11:08] And Adam Thomas says China uses James [11:10] Bond for a Euro push. The world is [11:12] changing. Let's play this advertisement [11:14] from BYD. [11:15] >> A world with the capacity to find [11:17] change. [11:18] But just as spring follows a harsh [11:20] winter and [11:21] summer looks back on a routine spring, [11:25] old selves, past identities, they shift. [11:28] They have to. [11:30] It's so exciting. Do you think? [11:34] And it doesn't life ask us to step out [11:36] of the shadows and embrace the new? [11:38] >> Are there car seats for dogs? [11:42] >> I don't think so. I think it gets pretty [11:44] dangerous. Although that dog looks like [11:45] it's wearing some sort of harness that [11:47] could be [11:48] tied in. [11:51] See, post James Bond he has a lot more [11:54] comedic timing. He's done SNL a few [11:56] times. He He's done a couple comedies. [11:59] There's a lot more to it, but [12:01] he will still just never not be James [12:03] Bond because he had such a successful [12:06] run [12:07] of James Bond performances. This is He's [12:09] really driving this car. [12:11] >> You can hear but but pause pause pause [12:14] pause. [12:14] Rewind 10 seconds. [12:16] >> Okay. What are we listening for? [12:18] >> Listen cuz there's some uh [12:27] Doesn't that Doesn't that sound like a [12:30] internal combustion engine for a second [12:31] there? [12:32] >> Yeah. So they're larping. It's played [12:33] through the speakers while you're [12:35] driving. Is that an option? [12:36] >> Maybe. [12:36] >> I feel like it has to be an option if [12:38] they're advertising it like that. [12:39] >> Yeah. [12:41] >> But who knows? You never know with these [12:43] with these BYDs if there's actually an [12:46] internal combustion engine in there [12:48] somewhere that can activate at a certain [12:50] point. Like everything feels like a [12:51] hybrid these days. But I mean it says [12:52] luxury EVs. What a great partnership and [12:54] what a great run from Daniel Craig to be [12:56] able to just like cash in on the the [12:59] aura of being James Bond forever even [13:02] after the franchise ends and he has to [13:04] move on to other things. It really is [13:06] remarkable [13:07] if you're Daniel Craig. And it also [13:09] helped that Do you ever see Layer Cake? [13:11] No one's seen Layer Cake here? Oh, such [13:14] a good movie. And he plays He plays like [13:16] sort of a someone involved in like the [13:18] drug trade in Europe. But it's a very [13:21] James Bond-esque character. And so even [13:24] throughout his portfolio of movies when [13:26] he's And then And then he plays Benoit [13:28] Blanc from Knives Out. And even that [13:32] character, even though it has a [13:33] different even though he has a different [13:34] accent, it still feels like he carries [13:36] the authority of a James Bond-like [13:39] figure. And so, he's always had this [13:41] same demeanor and aura around him that's [13:44] been built through his entire cinematic [13:45] portfolio that he can continue to cash [13:47] in on. And we when you're thinking about [13:49] advertising a particular car like an [13:51] Aston Martin, like this luxury BYD, your [13:54] mind goes to him before anyone else, [13:56] really. [13:57] >> European price tag is 134. [13:59] >> Ooh, okay. They're actually getting [14:00] pricey. Usually, like when you see these [14:02] BYD numbers, it's always like it's the [14:05] performance of a Ferrari for the price [14:06] of a Camry. [14:07] >> Well, so there's the Denza Z, which is [14:10] six The The The estimated price is 60 to [14:12] 140. And then but a lot of the other [14:15] Denza's are in the 40 to 60K range. [14:18] >> Okay. Okay. Not too bad. Not too bad. [14:20] Well, uh you know where I'd like to see [14:22] Daniel Craig do his next endorsement? [14:25] Cerebras. I would love to see a Super [14:26] Bowl ad where [14:28] >> he's lamenting the [14:30] slow speed of AI inference. And he [14:34] >> solves his problem in this Super Bowl ad [14:37] by partnering with Cerebras, firing up [14:40] some Cerebras chips. Uh the IPO is [14:43] looking like it's going very, very well. [14:44] So, Cerebras updated their filing, [14:46] according to Reuters. The IPO date will [14:49] be Thursday, May 14th. I heard maybe [14:52] Wednesday, but any day now. And they're [14:55] offering 30 million shares. That's up [14:57] from 28 million. So, they're they're [14:58] offering more shares than they were [15:00] expecting to. And they also increased [15:02] their price range from 115 to 125 up to [15:06] 150, 160. And so, they're going to [15:09] raise, instead of 3.5 billion, they'll [15:11] be raising 4.8 billion. And allegedly, [15:14] the the round is massively [15:16] oversubscribed to the tune of 20X demand [15:19] for that for billion. So, [15:23] uh something like [15:24] wait, 100 billion of demand for that 5 [15:27] billion, which is remarkable at that [15:29] price. Now, does that mean it's going to [15:31] 10x on day one? No, but it's certainly a [15:33] good sign going into this this IPO. And [15:36] so, that's why I called it like a [15:37] potential $50 IPO just to sort of have [15:40] some some parallelism with the $500 [15:42] watch and the 50 trillion of IP of GDP [15:45] meeting in China that we'll talk about [15:46] later. [15:47] So, there was a bunch of fun around [15:49] >> says if you were looking for an ID the [15:51] ideal time to IPO being a chip company [15:54] in May 2026 is hard to beat. [15:56] >> It really is. It really is. He had a [15:58] great piece called the inference shift [15:59] today on Stratechery. Go check it out. [16:01] There has been a bunch of fun about [16:02] Cerebras. I mean, for a long time they [16:04] were just sort of like building in [16:06] stealth. They're talking about the idea. [16:08] It takes a really long time to [16:10] design these chips, tape them out, and [16:11] then actually produce them. And then the [16:13] first version is less flexible, less [16:16] designed collaboratively with the [16:17] companies that are using them. So, there [16:19] was like one customer that was buying [16:21] them, and there was a lot of customer [16:22] concentration. Now, the chips have [16:25] actually been deployed. And there was [16:26] this big narrative about like okay, [16:28] well, they're maybe overly optimizing [16:30] for the transformer architecture. What [16:32] happens if the AI researchers come out [16:34] with like attention is all you actually [16:37] don't need that much. Attention is nice [16:39] and useful, but we have a new thing [16:40] that's better. And that didn't happen. [16:42] And so, attention and transformer-based [16:44] architectures are still dominant, and [16:46] inference costs are extremely important [16:48] in the age of AI agents. And and speed [16:51] is is so so important. And so, demand [16:53] is, you know, 10x-ing every few months [16:55] at this point. And there's a very very [16:57] clear business story. Tyler, do you have [16:58] anything else on Cerebras? [16:59] >> Uh yeah, I mean, just like if you use [17:01] the Cerebras chips, like you can use it [17:02] GPT-5.3 Spark in Codex. It's like insane [17:06] how fast [17:06] >> It's crazy. It's crazy. [17:08] Yeah, if you want to give it a try and [17:09] actually demo it, which I think is [17:11] important with these like [17:13] with these these these like [17:15] semiconductor companies, if they're like [17:16] abstract and you're like I don't know if [17:17] it's like a real company or something. [17:19] Like, you can actually just go down, [17:20] download Codex Desktop, pick from the [17:22] down from the drop down 5.3 Spark, and [17:25] then you can you don't even have to get [17:27] it to do code. You can ask it history of [17:29] the Roman Empire, and it will just [17:31] instantly tell you a full page of [17:33] exactly the response with 5.3 level [17:35] intelligence, which is pretty good. [17:37] And it's a pretty remarkable experience, [17:40] and you can imagine this coming to every [17:43] LLM interface, every AI experience, [17:45] which has normally been like for any [17:47] meaningful work, fired off, come back 5 [17:49] minutes later, sometimes 2 hours later, [17:52] will cut all of that in in, you know, in [17:55] half or by 10, and that's where this is [17:58] going. So, you can see, you know, [18:00] significant demand even though there's [18:01] this like customer concentration thing. [18:03] I don't know why there wouldn't be a lot [18:05] of different customers lining up. Every [18:08] every lab that has exploding demand, [18:11] Cursor, Anthropic, [18:14] Meta, and Google, like [18:16] unless they have like a direct answer to [18:18] this, I I would see them being a buyer [18:19] in the near term. So, the Cerebras [18:22] upsized its IPO and topped the new [18:24] range. Reuters says the IPO drew orders [18:27] for more than 20 x the shares available. [18:29] Cerebras makes AI inference chips and [18:31] lists Amazon and OpenAI among customers. [18:34] So, [18:35] there was there was a question about [18:36] like was it all OpenAI, and I guess [18:38] Amazon has jumped on. [18:40] Very flexible with regard to the chips [18:42] that they rack over at AWS. Uh Poll [18:44] market is projecting Cerebras to close [18:46] above 50 billion market cap by the end [18:49] of day one. That would be about 100% [18:51] above the target valuation of 26 billion [18:53] that was previously reported. So, [18:55] preparing for an IPO on the Nasdaq next [18:57] week under the ticker CBRS. [19:00] And traditional semiconductor [19:01] manufacturing works like this. A silicon [19:03] wafer is fabricated. The wafer is cut [19:05] into hundreds of smaller chips. The [19:06] chips are packaged individually and [19:08] connected together in systems. Cerebras [19:10] took a completely different approach. [19:12] They use the entire 300 mm wafer, 4 [19:15] trillion transistors, 900,000 AI [19:18] AI oriented compute cores, and the big [19:21] thing is the petabits per second of [19:23] internal bandwidth. So, better memory [19:26] bandwidth for [19:27] KB caches and everything that you need [19:28] to do in AI, I think, something like [19:30] that. Benchmark is going to be [19:33] absolutely cleaning up. They still own [19:34] over 20% of Cerebrus, apparently. [19:37] If it trades at even half of how Shang [19:40] High priced more threads in Camber Con, [19:42] it will be over 500 billion in less than [19:46] 2 years. Okay, that's a big step. [19:49] It will break the venture model if they [19:51] hold. Has a shot to deliver the number [19:53] one fund in VC history. Benchmark on an [19:56] absolute tear. Quant again says, [19:58] "Cerebrus is pretty funny because you [20:00] can just imagine the origin story being [20:02] some boomer non-technical manager going, [20:05] 'Okay, but why can't you just put 50 [20:08] gigs of of L3 cache on this chip?' And [20:11] the engineer being put on the spot and [20:13] going, 'I guess you could.'" [20:15] And uh [20:17] said, [20:18] "Someone else in the comments here [20:19] chiming in, I I random history major had [20:22] a much less important version of this [20:23] conversation with my college roommate [20:25] fancy engineer during computer vision [20:26] stuff one time and still feel really [20:28] good about it. I think it helped. Just [20:30] we weld the thing that does that onto [20:32] the other thing. Have you considered [20:34] building the entire plane out of the [20:35] black box? But it works and the plane [20:37] goes Mach 10." That's exactly what [20:39] happened. Well, let's move over to [20:41] China. Donald Trump is meeting with uh [20:44] Xi Jinping this week and you were you [20:47] were sharing some info on how many [20:50] journalists are going over there. [20:51] >> They're calling it a field trip. [20:52] >> field trip. [20:53] >> They're calling it a field trip because [20:55] Tim Cook is going, Larry Fink, Stephen [20:57] >> Oh, okay. I thought it was journalists, [20:59] but I guess it's tech people. [21:00] >> Jane from City, Chuck Robbins is headed [21:03] over there from from Cisco, David [21:05] Solomon. [21:06] >> Interesting. [21:07] >> And a whole bunch of others. I guess [21:10] Elon is supposedly on the trip as well. [21:12] >> Mhm. [21:13] >> And uh [21:13] >> they can get a word in edgewise because [21:16] the the the vast majority of the [21:18] discussions will obviously center around [21:19] the war in Iran. This is from the Wall [21:21] Street Journal as the heads of the [21:23] world's two superpowers meet in Beijing [21:25] this week. President Trump and Chinese [21:27] leader Xi Jinping will have another [21:29] nation looming over their summit, Iran. [21:31] The long-anticipated meeting has been [21:33] delayed once due to the US and Israel's [21:36] war against Iran, which has led to the [21:38] closure of the Strait of Hormuz. Trump [21:40] is eager to move on from the Middle East [21:41] war that is sapping his domestic power [21:43] and straining the global economy. The [21:46] >> As of this morning, the peace deal was, [21:48] according to Trump, on major life [21:50] support. [21:50] >> On major life support. Is that good? [21:53] It's better than it being dead. So, [21:55] hopefully we get a peace deal because if [21:57] you're on life support, sometimes you [21:58] can have a miraculous comeback. I don't [22:00] know. That's what I'm pulling for. He [22:01] will land in Beijing prepared to push [22:03] China, which relies on Iran for low-cost [22:06] oil in their transactional relationship, [22:07] to help broker an agreement that ends [22:09] the conflict. Uh Xi Jinping also wants [22:11] the fighting to stop as Middle East [22:13] turmoil restricts China's oil supply and [22:15] shrinks countries' ability to buy [22:18] Chinese goods. Finding a resolution [22:21] could raise Xi Jinping's stature as a [22:23] global statesman who swooped in at the [22:25] precipice of a possible military [22:27] escalation. Trump on Friday threatened [22:29] to resume Project Freedom, the US-led [22:31] operation helped ships navigate the [22:33] strait safely, adding this time that the [22:36] operation would include quote other [22:37] things. Very ominous. So, I think uh [22:41] there's a few different things. So, uh [22:42] Trump uh the the deal's on life support, [22:45] but most recently he rejected Iran's [22:47] latest response to a US peace proposal. [22:50] They're going back and forth. Oil prices [22:52] have climbed amid fears of a prolonged [22:55] disruption through a choke point that [22:57] carries roughly 1/5 of global oil flows. [23:00] And the summit is focused heavily on [23:03] Iran, but also uh trade deals, [23:05] specifically Chinese purchases of [23:07] American agriculture, energy, aerospace [23:09] products, and some other related [23:11] investment mechanisms. But, the tech [23:13] industry is obviously hoping to like [23:15] wind down the conflict peacefully and [23:17] quickly and then move on to discussions [23:19] of export restrictions, GPUs, the AI [23:21] supply chain, rare earths, all the [23:23] different things that go into [23:25] what the tech industry needs to [23:27] flourish. [23:28] But, I was thinking we wouldn't get that [23:30] much movement or that many sound bites [23:32] from this trip based on how large Iran [23:36] is looming, but with all of those tech [23:37] CEOs there, you would imagine that [23:39] there's some conversation that happens [23:41] there. [23:41] >> might be clip farming? [23:42] >> Potentially. Potentially or farming. [23:44] Potentially frame mugging each other. [23:46] You never know. Tyler, what do you [23:47] think? [23:47] >> Uh yeah, I think it it would be [23:48] interesting if it's also, you know, at a [23:50] higher level than just the supply chain. [23:51] Cuz like last week there's all the news [23:53] about CAAISI, right, doing the new [23:57] AI regulation, right? Like what's going [23:58] to happen? Like what what are we going [24:00] to you do much test before the models [24:01] come out? It seems like they're kind of [24:02] moving away from that, like less kind of [24:04] safety focused. [24:06] >> China should be like, "Send us mythos, [24:08] too. Give us unfettered access. Ideally [24:10] like enough to distill it really [24:11] quickly, and then we will also say [24:14] whether or not it can be released in [24:15] China." [24:16] >> Well, why not why stop there? Why not [24:18] just send the weights? [24:19] >> Yeah, just send the weights. That's a [24:20] that's actually way more efficient. [24:21] >> it. [24:21] >> Yeah, way more efficient. We'll inspect [24:23] it. We'll make sure that it's okay for [24:24] the Chinese population. [24:25] >> Yeah, but you know, like the we we've [24:27] talked about this a little bit on the [24:28] show. [24:28] >> Actually send send all the GPUs I need [24:29] to write that, too. [24:31] Sorry. [24:32] >> We we've talked about this a little bit, [24:33] but like there are people in China who [24:34] are actually worried about like the the, [24:35] you know, very like safety pill. [24:37] But, it seems like people going over [24:38] there are like Tim Cook, [24:39] >> Yeah, I mean there's a Chinese [24:40] restaurant called Elon. [24:41] >> These people are not going to be arguing [24:42] in favor of like safety. No. [24:45] >> Palmer brings up an important point. [24:47] It's time for the United States Postal [24:49] Service to ban junk mail. [24:51] Unsolicited spam calls are already [24:53] prohibited by the FCC. Emails are [24:55] heavily regulated by the CAN-SPAM Act of [24:59] 2003. Junk mail is the majority of mail. [25:01] 100 million trees per year. Enough. It [25:04] really is way too much. [25:07] >> This is very interesting. I have I I I [25:09] thought I put all of the the blame or I [25:12] I guess the credit I gave all the credit [25:13] to Google with the fact that like I [25:15] don't get spam emails. I get emails if I [25:19] buy something online and I forget to [25:21] uncheck the box and that's kind of on me [25:23] or if I'm subscribed to a newsletter and [25:25] it gets boring and I'm like ah this is [25:27] junk. I'll need to deal with that. But [25:29] like very rarely do I just get a true [25:32] spam email. Just like truly slob junk [25:35] like just complete nonsense. It's pretty [25:38] pretty rare. Uh and I think some of [25:40] that's the filtering but also the [25:41] CAN-SPAM Act seems to be somewhat [25:44] effective. I wind up getting a lot more [25:45] spam phone calls and a lot more spam [25:48] text messages these days than spam [25:49] emails in terms of like cold outreach [25:52] that's completely undirected. And so [25:54] yeah, maybe they need to expand the [25:56] CAN-SPAM Act. Shouldn't it be the [25:58] CAN'T-SPAM Act? I don't know why they [26:00] call it CAN but uh it has been [26:02] successful at least in email. But I [26:04] agree with this. This is good. Uh Palmer [26:06] says it's insane that America has given [26:08] a monopoly on letter delivery to a [26:11] quasi-governmental agency that then uses [26:14] it to flood our homes with useless [26:16] garbage against our will. America would [26:18] never allow FedEx, UPS, DHL, or anyone [26:21] else to force this on us even ignoring [26:24] the wasted tax payer money, insane moral [26:26] hazards, and ecological impact. The lost [26:29] time and productivity is inexcusable. I [26:31] agree. If the average American spends [26:33] only 30 seconds sorting their mostly [26:36] spam mail each day looking for the real [26:38] stuff, that's over a billion dollars. [26:40] Well, Earth Class Mail, Palmer, I think [26:42] you need a PO box that scans the emails [26:45] or scans the physical mail and delivers [26:47] it to an email inbox. Throw Open Claw in [26:49] front of it and have it decide what's [26:52] what's what makes it through. New [26:53] service for SF retailers and home [26:55] buyers. I will show up to your open [26:57] houses wearing open AI or Anthropic [27:00] merch. I charge 5% commission. Just to [27:03] just to fully pump up the price. Uh it's [27:05] called chandelier real estate. [27:06] >> will that actually help though? Because [27:09] I think some people [27:11] would just be like I'm not even going to [27:12] bid. [27:12] >> But I think some some people at the open [27:14] house tour might see oh there's an open [27:16] AI or Anthropic person. I should make my [27:18] offer particularly strong if I want this [27:21] even if there even if there isn't that [27:23] much [27:24] uh that much demand. I love the founder [27:26] railway. It's great. Um what else is [27:28] going on here? Uh [27:30] new before chat GPT release before [27:32] Microsoft's $1 billion bet and long [27:35] before plans for an IPO there was the [27:37] University of Michigan putting 20 [27:39] million into AI. [27:41] >> Pull up Tyler. Pull up Tyler. [27:43] >> What what happened here? Like how did [27:45] this actually happen? [27:47] There we go. [27:50] There you go. Nailed it. [27:51] >> Tyler, your homework learn how to get [27:53] that more dialed because I hit that [27:55] pretty often for you and every time you [27:57] go the wrong something to work on. [27:58] That's just some constructive feedback. [28:00] I believe that you could be better at [28:02] this at this task. Thanks. [28:07] >> This this feels like the University of [28:08] Michigan was considering donating and [28:10] then at some point it just became they [28:13] just became aware of like oh well like [28:14] you could also participate in the [28:15] for-profit and they're like oh that [28:16] sounds better maybe. Like let's let's [28:19] put this in. My mom is so bad with [28:21] technology. She literally tried to [28:24] search up info about energy drinks and [28:27] accidentally set up set our house up as [28:30] a business. [28:38] What is an energy drinks? [28:40] Can you imagine? Like starting with a [28:43] Google search and ending up like [28:45] creating a business on Google Maps for [28:48] your exact address under the name what [28:51] is in the box. [28:53] And and people think AI will diffuse [28:55] quickly. Debatable. [28:57] Debatable. Leave us five stars on Apple [28:59] podcast and Spotify. Sign up for our [29:01] newsletter to bpn.com. Goodbye.