Ark Invest: On Tesla-SpaceX Merger, Optimus Rollout, Grok 4.5 & AI Models

以下是ARK Invest 的 Cathie Wood 和 Tasha Keeney 问答环节的总结，包括所有讨论的新闻和观点： **整体背景：** 本次问答环节由 ARK Invest 的 Tasha Keeney 和 Cathie Wood 主讲，讨论了特斯拉、SpaceX、xAI、擎天柱（Optimus）以及近期发布的 AI 模型。主持人也插入了他自己强烈的主观意见和预测。 **1. 特斯拉与 SpaceX 的合并可能性：** * **问题：** 特斯拉股东为何会批准与 SpaceX 传闻中的合并案？尤其是在 ARK 对特斯拉机器人出租车业务机会的估值高于潜在收购价格的情况下？ * **Tasha Keeney 的批准理由：** * 市场需要时间才能充分认识到特斯拉机器人出租车的机会，ARK 认为，这将在五年内贡献特斯拉大部分的企业价值。 * SpaceX “当然有可能”（主持人称“绝对肯定”）需要为特斯拉支付 **溢价** 才能获得股东批准，因为弃权票等同于反对票。 * **合并的理由：** * **减轻管理层压力：** 缓解埃隆·马斯克及高管团队在多个行业推动前沿探索的负担。 * **相互需求：** SpaceX 可以利用特斯拉高利润、经常性的机器人出租车现金流来资助其雄心勃勃的 AI 项目。 * 特斯拉需要 SpaceX，特别是 **xAI 模型 (Grok)**，它可为 **擎天柱 (Optimus)** 提供编排和规划层。 * 两家公司在 **TerraFab 项目** 上也相互需要，该项目旨在内部制造芯片，以解决当前 AI 计算芯片的限制。 * ARK 的初步模型表明，将特斯拉的现金流用于 SpaceX 的 AI 雄心，将为合并后的实体带来 **更高的投资者价值**。 * **主持人的澄清/观点：** * 他澄清说，特斯拉机器人出租车的利润资助 SpaceX 的 AI 项目，对特斯拉股东并无不利，相反，这是将资本用于“最佳用途”，SpaceX 可以在 AI 投资上产生“更大的回报”，从而带来“天文数字般的收入和利润”。 * 他指出，合并还可能赋予 **埃隆·马斯克超级投票权**，这是他目前在特斯拉所没有的。 * **Cathie Wood 的观点：** 她认为埃隆·马斯克是富有远见的管理者，并相信 SpaceX 将成为“世界上垂直整合度最高的公司”，参与几乎所有融合技术。 * **主持人的强烈意见：** 他认为埃隆·马斯克比普通股东做出更好的商业决策，并倾向于让马斯克在他的上市公司中拥有“更大的发言权”和更多的控制权。 **2. Grok/xAI 与基因组学 (主持人预测)：** * **主持人预测：** 他表示，**SpaceX AI 宣布一个专注于基因组学的项目** 只是“时间问题”，旨在了解人类生物学、长寿、疾病预防和治疗。 * **理由：** 他认为所有生命都是“运行的代码”（DNA），而 AI（例如用于编程的 GrokBuild）非常适合“梳理海量的遗传信息”，“优化”、“改变”、“改进”和“重写代码”。衰老被描述为生物代码中错误的积累，而 AI 可以“调试和修复这些错误”。 **3. 擎天柱人形机器人的推出：** * **Tasha Keeney 的估计：** * 特斯拉表示，擎天柱“在大约未来一年内”可能会对外销售。 * 更大规模的推广“很可能”在未来五年内实现。 * 关键转折点是当擎天柱的能力提升到“人类水平的熟练程度”时。 * ARK 的“大创意”研究表明，这可能在 **2028 年** 发生。 * 擎天柱将首先销售给 **企业**，因为企业对价格的承受能力更高，并且存在“车队销售”（批量销售）的潜力。 * 人形机器人的整体市场机会估计为 **26 万亿美元**，制造业和家用机器人各占一半，家用机器人的应用会晚一些。 * **主持人的评论：** * “手部灵活性”是最初迭代“硬件设计中最重要的部分”。 * 软件更新和车队学习将使第一代擎天柱随着时间的推移变得越来越强大。 * 他解释说，机器人工作时间是人类的四倍，即使速度只有人类的四分之一，也能达到同等水平。 * 商业应用将先于家庭使用，因为公司可以为增加利润而支付成本（每年 10 万美元），而家庭劳动通常不获得报酬，因此需要“显著”降低成本才能在家庭中可行。 **4. 近期 AI 模型发布 (Grok 4.5 和 GPT 5.6)：** * **约瑟夫的问题：** 近期前沿 AI 模型发布及其竞争地位的更新。 * **约瑟夫/主持人的分析：** * OpenAI 上周发布了 **GPT 5.6**，xAI 发布了 **Grok 4.5**。 * 这两款模型在性能上都代表了“重大飞跃”。 * 两者的“真正亮点”在于与前几代模型以及 Anthropic 的 Mythos 等竞争对手相比，其 **效率和成本节约**。 * **Grok 4.5** 以“极低的成本提供卓越的性能”。 * 旗舰版 **GPT 5.6 Sol** 提供相似的输出质量，但成本降低了 2-3 倍。 * 这两款模型的较小版本（Grok 4.5 和最小的 GPT 5.6）对于不需要前沿智能的任务，提供了“更显著的成本节约”。 * **关键见解：** 大多数用户不需要“地球上最聪明的头脑”。那些“工蜂模型”——那些“几乎和最好的模型一样好”（性能差距在 5% 以内）但成本极低的模型——将占据大部分 AI 收入。 * 实际表现仍有待验证。 * 预计更高效的产品将获得“显著关注”。 * **主持人的补充观点：** * “前沿模型”（今天的“爱因斯坦”）由于快速发展，将在短短几个月内被更高效、接近爱因斯坦级别的模型在能力上超越。 * 投资者应该关注这些“工蜂模型”，因为它们代表着巨大的机会。

Here's a summary of the Q&A with Cathie Wood and Tasha Keeney from ARK Invest, including all discussed news and points: **Overall Context:** The Q&A features Tasha Keeney and Cathie Wood from ARK Invest, discussing Tesla, SpaceX, xAI, Optimus, and recent AI model releases. The host also interjects with his own strong opinions and predictions. **1. Tesla & SpaceX Merger Potential:** * **The Question:** Why would Tesla shareholders approve a rumored merger with SpaceX, especially if ARK's valuation for Tesla's robo-taxi opportunity is higher than a potential acquisition price? * **Tasha Keeney's Argument for Approval:** * The market needs time to fully realize Tesla's robo-taxi opportunity, which ARK believes will contribute to the majority of Tesla's enterprise value within five years. * SpaceX would "certainly possible" (host says "absolutely certain") need to pay a **premium** for Tesla to get shareholder approval, as non-votes count as "no." * **Reasons the merger makes sense:** * **Reduces management strain:** Eases the burden on Elon Musk and executive teams pushing frontiers in multiple industries. * **Mutual need:** SpaceX could use Tesla's highly profitable, recurring robo-taxi cash flow to fund its ambitious AI projects. * Tesla needs SpaceX, particularly the **xAI models (Grok)**, which could power the orchestration and planning layer for **Optimus**. * Both companies need each other for the **TerraFab project**, an ambition to make chips in-house, addressing current AI compute chip constraints. * ARK's initial modeling suggests that factoring in Tesla's cash flow for SpaceX's AI ambitions would offer **more value to investors** in a combined entity. * **Host's Clarification/Perspective:** * He clarifies that Tesla's RoboTaxi profits funding SpaceX AI is not detrimental to Tesla shareholders, but rather puts the capital to its "best use," where SpaceX can generate a "much greater return" on AI investment, leading to "astronomical revenue and profits." * He notes that a merger would also likely grant **super-voting status to Elon Musk**, which he currently lacks at Tesla. * **Cathie Wood's Input:** She views Elon Musk as visionary management and believes SpaceX will become the "most vertically integrated company in the world," participating in almost all converging technologies. * **Host's Strong Opinion:** He believes Elon Musk makes better business decisions than average shareholders and prefers Musk to have an "outsized say" and more control in his publicly traded companies. **2. Grok/xAI and Genomics (Host's Prediction):** * **Host's Prediction:** He states it's a "matter of time" until **SpaceX AI announces a project focused on genomics**, aiming to understand human biology, longevity, disease prevention, and cures. * **Rationale:** He argues that all life is "running code" (DNA), and AI (like GrokBuild for coding) is ideally suited to "trawl through the monumental amounts of genetic information," "optimize," "change," "improve," and "rewrite the code." Aging is described as the accumulation of errors in biological code, and AI could "debug and repair those errors." **3. Optimus Humanoid Robot Rollout:** * **Tasha Keeney's Estimates:** * Tesla has stated Optimus could be sold externally "within the next year or so." * A wider rollout is "likely" within the next five years. * The key turning point is when Optimus's capability improves to "human level proficiency." * ARK's "Big Ideas" research suggests this could happen by **2028**. * Optimus will first be sold to **businesses** due to their higher price tolerance and potential for "fleet sales" (volume). * The overall market opportunity for humanoid robots is estimated at **$26 trillion**, evenly split between manufacturing and home robotics, with home use coming later. * **Host's Commentary:** * **Hand dexterity** is the "most important part of the hardware design" for initial iterations. * Software updates and fleet learning will make first-gen Optimus increasingly capable over time. * He explains that robots working four times as many hours as humans can be on par even at a quarter of the speed. * Commercial adoption will precede home use because companies can justify the cost ($100,000/year) to increase profits, whereas home labor isn't typically compensated, requiring a "meaningfully" lower cost for household viability. **4. Recent AI Model Releases (Grok 4.5 & GPT 5.6):** * **Joseph's Question:** Update on recent frontier AI model releases and their competitive standing. * **Joseph/Host's Analysis:** * OpenAI released **GPT 5.6** and xAI released **Grok 4.5** "last week." * Both models represent "significant leaps forward" in performance. * The "real headline" for both is their **efficiency and cost savings** compared to previous generations and competitors like Anthropic's Mythos. * **Grok 4.5** offers "phenomenal performance for a fraction of the cost." * The flagship **GPT 5.6 Sol** offers similar quality of output while being 2-3 times cheaper. * Smaller versions of both models (Grok 4.5 and the smallest GPT 5.6) offer "even more substantial cost savings" for tasks not requiring frontier intelligence. * **Key Insight:** Most users don't need "the biggest brain on the planet." The "worker bee models" – those "almost as good as the best" (within 5% of performance) but at a fraction of the cost – will capture most of the AI revenue. * Real-world performance still needs to be verified. * More efficient offerings are expected to gain "significant traction." * **Host's Further Point:** * The "frontier models" (the "Einsteins" of today) will be overtaken in capability by more efficient, nearly-Einstein models in just a few months due to rapid advancement. * Investors should pay attention to these "worker bee models" because they represent the massive opportunity.