A $13 Billion Bet on Robots?

The Motley Fool Hidden Gems Investing播客首先讨论了英伟达(NVIDIA)据报道有意以130亿美元收购Hugging Face的事件，这凸显了其在机器人和人工智能领域的一项重大押注。Foolish贡献者瑞秋·沃伦(Rachel Warren)和马特·弗兰克尔(Matt Frankel)与主持人乔恩·夸斯特(Jon Quast)一同探讨了这次潜在收购，随后深入探讨了听众关于投资论点的问题。 瑞秋·沃伦解释说，Hugging Face是人工智能世界的中心“城镇广场”，它是一个云存储库，数百万开发人员和研究人员在此协作，共享、测试和托管开源人工智能模型、数据集和软件库。开源意味着底层代码是公开可用的，可以进行修改，这与那些将自己的人工智能专有化的科技巨头形成对比。市值5万亿美元的英伟达将支付Hugging Face当前营收的86到129倍，瑞秋形容这笔金额对他们来说是“九牛一毛”。此次收购将使英伟达控制全球开发者社区的关键分发层，并掌握llama.cpp背后的团队，后者使人工智能模型能够在消费设备上运行。马特·弗兰克尔和瑞秋认为，英伟达的动机也可能是为了阻止OpenAI和Alphabet等人工智能模型公司变得过于强大，尤其是在它们开发定制人工智能芯片以减少对英伟达依赖的情况下。此举将为英伟达提供一个数据金矿，在开源领域获得优势，并为其硬件和软件生态系统赢得一个专属客户群。 对话随后转向了与Hugging Face合作的Micro Duck机器人。乔恩·夸斯特将其描述为“一只在你房子里走来走去的独眼鸭子”，介于菲比(Furby)和皮克斯(Pixar)台灯之间。Micro Duck是一个25厘米、800克的双足机器人，由Hugging Face收购的公司Pollen Robotics开发。它定价399美元，拥有十多个电机、一个摄像头、微型激光雷达和一个可动喙。它的目的是为开发人员和学生提供一个经济实惠的平台，用于测试现实世界中的人工智能软件，弥合数字代码与现实之间的鸿沟。用户可以在模拟环境中训练数字孪生体，并直接将自定义行为部署到物理机器人上。凭借其开源代码和模拟环境，Micro Duck将消费者转变为创造者，在发布24小时内获得了超过260万美元的预订（超过6500台），引起了巨大轰动。 马特·弗兰克尔将Micro Duck主要视为英伟达的“数据策略”。他表示，英伟达对拥有Micro Duck本身不感兴趣，而是对其从数千个在不同家庭中运行的设备生成的大量数据集感兴趣，这些数据对于运动和操控研究来说是无价的。瑞秋·沃伦补充说，英伟达进军开源机器人领域，可能会挑战特斯拉的擎天柱(Optimus)和Figure等公司的“封闭生态系统”，这表明解决物理人工智能问题对任何一家公司来说都过于复杂。一个开源框架能够使人工智能技术栈民主化，汇集来自不同环境的真实世界数据，并加速机器学习，从而惠及小型初创公司和学术研究实验室。 播客随后回答了听众的问题，首先是拉维(Ravi)关于投资论点的提问。马特·弗兰克尔将投资论点定义为股票价值可能上涨背后的理由，它需要具体性，并回答以下问题：1) 要证明论点正确需要发生什么，2) 预期何时发生，以及3) 什么会证明论点是错误的。拉维的五步提问流程，特别是关于“主要风险/看跌情景”和“反驳证据”的部分，受到了称赞。马特强调了理解空头论点比空头本身更重要的意义，并引用了查理·芒格(Charlie Munger)的一句话。投资者应该确定具体的、可衡量的卖出触发点，例如公司云收入减速或市场份额流失。瑞秋补充说，接受看跌思维有助于对抗确认偏误，她敦促投资者“像检查员一样寻找问题”。她建议撰写一份简短报告，反驳自己的立场，并查阅公司的10K风险因素和独立的看跌研究报告。 最后，主持人讨论了纳温(Nawin)的经历：他在一家小型股SPAC IPO公司被收购后成功出售股票并获利，尽管这只股票仅占其投资组合的1%。纳温质疑这是否只是运气，还是一个可靠的论点。马特认为这“两者皆有”，承认收购带来的回报有运气成分，但也证实了将投机性仓位保持在较小比例的策略是有效的，这样既能限制下行风险，又能让表现出色的投资获得显著回报。瑞秋同意，该公司强劲的基本面（无债务、正现金流、低估值）使其成为一个引人注目的价值投资和主要收购目标。她重申，保持小仓位是明智的风险管理。在构建更广泛的投资论点方面，瑞秋强调估值指标只是一个起点；一个引人注目的论点还需要定性因素，如长期顺风、不断扩大的竞争护城河和优质管理层。乔恩·夸斯特最后引用了摩根·豪塞尔(Morgan Housel)的话，强调“运气和风险是孪生兄弟”，长期坚持是至关重要的。

The Motley Fool Hidden Gems Investing podcast began by discussing NVIDIA's reported interest in acquiring Hugging Face for $13 billion, highlighting a significant bet in robotics and AI. Foolish contributors Rachel Warren and Matt Frankel joined host Jon Quast to explore this potential acquisition and later delve into listener questions about investment theses. Rachel Warren explained that Hugging Face serves as a central "town square" for the AI world, functioning as a cloud repository where millions of developers and researchers collaborate, sharing, testing, and hosting open-source AI models, datasets, and software libraries. Open-source means the underlying code is publicly available for modification, contrasting with tech giants who keep their AI proprietary. NVIDIA, a $5 trillion company, would be paying 86 to 129 times Hugging Face's current revenue, an amount Rachel described as "a drop in the bucket" for them. The acquisition would give NVIDIA control over a crucial distribution layer for the global developer community and the team behind llama.cpp, which enables AI models to run on consumer devices. Matt Frankel and Rachel suggested NVIDIA's motivation could also be to prevent AI model companies like OpenAI and Alphabet from becoming overly powerful, especially as they develop custom AI chips to reduce dependency on NVIDIA. This move would provide NVIDIA with a data goldmine, an edge in the open-source space, and a captive customer base for its hardware and software ecosystem. The conversation then shifted to the Micro Duck robot, a collaboration with Hugging Face. Described by Jon Quast as a "Cyclops duck that walks around your house" and a mix between a Furby and the Pixar lamp, the Micro Duck is a 25-centimeter, 800-gram bipedal robot developed by Pollen Robotics, a company Hugging Face acquired. Priced at $399, it features over a dozen motors, a camera, miniature LiDAR, and an articulated beak. Its purpose is to offer an affordable platform for developers and students to test real-world AI software, bridging the gap between digital code and reality. Users can train a digital twin in a simulation and deploy custom behaviors directly onto the physical robot. With its open-source code and simulation environments, the Micro Duck transforms consumers into creators, generating significant excitement with over $2.6 million in pre-orders (more than 6,500 units) within its first 24 hours. Matt Frankel viewed the Micro Duck primarily as a "data play" for NVIDIA. He stated that NVIDIA isn't interested in owning the Micro Duck itself but rather the vast dataset it generates from thousands of units operating in diverse households, which would be invaluable for locomotion and manipulation research. Rachel Warren added that NVIDIA's push into open-source robotics, potentially challenging the "locked ecosystems" of companies like Tesla's Optimus and Figure, acknowledges that solving physical AI is too complex for any single company. An open-source framework democratizes the AI stack, aggregates real-world data from diverse environments, and can accelerate machine learning, benefiting smaller startups and academic research labs. The podcast then addressed listener questions, starting with Ravi's query about investment theses. Matt Frankel defined an investment thesis as the reasoning behind why a stock's value might increase, requiring specificity and answers to: 1) what needs to happen to be right, 2) when it's expected, and 3) what would prove it wrong. Ravi's five-question process, particularly regarding "major risks/bear case scenarios" and "disconfirming evidence," was commended. Matt emphasized the importance of understanding the bear argument better than the bears themselves, citing a Charlie Munger quote. Investors should identify specific, measurable triggers for selling, such as a company's cloud revenue decelerating or losing market share. Rachel added that embracing bearish thinking helps counter confirmation bias, urging investors to "think like an inspector looking for damage." She suggested writing a short report arguing against one's own position and reviewing a company's 10K risk factors and independent bearish research. Finally, the hosts discussed Nawin's experience of successfully selling a small-cap SPAC IPO stock for a profit after it was acquired, despite keeping it a small 1% portfolio position. Nawin questioned if it was luck or a sound thesis. Matt believed it was "both," acknowledging the luck of an acquisition-driven return while validating the strategy of keeping speculative positions small to limit downside risk and still allow for significant returns if the investment performs exceptionally. Rachel agreed that the company's strong fundamentals (no debt, positive cash flow, low valuation) made it a compelling value investment and a prime acquisition target. She reiterated that maintaining a small position was wise risk management. In terms of building a broader investment thesis, Rachel stressed that valuation metrics are just a starting point; a compelling thesis also requires qualitative factors like secular tailwinds, a widening competitive moat, and quality management. Matt noted that the approach to thesis building varies depending on the stock type, with growth stocks emphasizing qualitative aspects and established compounders focusing more on metrics. Jon Quast concluded with a quote from Morgan Housel, highlighting that "luck and risk are siblings" and that staying in the game long term is paramount.