引言：一种规避监管的新型控制策略

Narrator: In the race to dominate generative AI, Big Tech firms haven’t just been building, they’ve been buying. But there’s something strange about most of the deals that they have struck. Companies like Meta, Microsoft, Amazon, Google, and Nvidia are embedding themselves deep within the AI ecosystem through strategic investments, exclusive partnerships, and talent acquisitions- with deals that stop just short of formal takeovers, but the economic impact of these deals it turns out – is indistinguishable from full control.

讲述者: 在主导生成式AI（Generative AI：能够创造新内容的人工智能）的竞赛中，科技巨头们不仅在自主研发，更在积极收购。然而，他们达成的大多数交易都有些不同寻常。Meta、微软、亚马逊、谷歌和英伟达等公司正通过战略投资、排他性合作和人才招揽，将自身深深嵌入AI生态系统。这些交易在形式上恰好止步于正式收购，但其经济影响却与完全控制无异。

Narrator: This new approach—which some analysts are describing as “non-acquisition acquisitions”—is designed to work around antitrust law. The idea is to avoid drawing regulatory scrutiny with deals that allow tech giants to secure privileged access to foundational AI models, critical data pipelines, and elite research talent.

讲述者: 这种被一些分析师称为“非收购式收购”（Non-acquisition acquisitions：指在法律上不构成正式收购，但能实现类似战略控制的交易）的新方法，其目的在于规避反垄断法（Antitrust Law：旨在促进市场竞争、防止垄断行为的法律）。其核心思想是通过这些交易，在不引发监管审查的情况下，确保科技巨头能够获得对基础AI模型、关键数据管道和顶尖研究人才的优先使用权。

Narrator: Despite all the talk of slashing regulation, and despite the support that Trump has drawn from silicon valley - so far, the new administration is showing a surprising degree of continuity with the Biden-era approach to antitrust enforcement, particularly when it comes to Big Tech and AI-related deals. According to the law firm Goodwin Procter - US enforcers are still willing to challenge deals they view as anticompetitive under traditional horizontal theories of harm — particularly mergers between direct competitors with leading market positions. So far there has been no signs of reduced FTC enforcement.

讲述者: 尽管外界充斥着削减监管的言论，且特朗普获得了硅谷的支持，但到目前为止，新一届政府在反垄断执法上，尤其是在针对科技巨头和AI相关交易方面，出人意料地延续了拜登时代的做法。根据Goodwin Procter律师事务所的分析，美国执法机构仍然愿意依据传统的横向损害理论，对他们认为具有反竞争性的交易提出挑战，特别是那些涉及市场领先地位的直接竞争者之间的合并。迄今为止，尚未有迹象表明联邦贸易委员会（FTC, Federal Trade Commission: 美国负责执行反垄断法和保护消费者的独立政府机构）的执法力度有所减弱。

历史的回响：洛克菲勒的策略与当今科技巨头的相似之处

Narrator: The tactics used by today’s tech giants to quietly consolidate power in the AI sector are not without precedent. More than a century ago, John D. Rockefeller pioneered similar strategies during the rise of Standard Oil, where he often acquired companies without formal legal documentation using silent partnerships, verbal agreements, and trusted intermediaries to quietly absorb competitors.

讲述者: 当今科技巨头在AI领域悄然巩固权力的策略并非没有先例。一个多世纪前，约翰·D·洛克菲勒在标准石油公司崛起期间，就开创了类似的策略。他常常通过秘密合伙、口头协议和可信的中间人，在没有正式法律文件的情况下收购公司，从而悄无声息地吞并竞争对手。

Narrator: According to his biographer - Standard Oil executives worried that this secrecy could cause problems for the firm - as if a seller died unexpectedly - his heirs might mistakenly claim ownership of a refinery that Standard Oil had bought and paid for.

讲述者: 根据洛克菲勒的传记作者所述，标准石油的高管们曾担心这种保密做法会给公司带来麻烦——比如，如果卖家意外去世，其继承人可能会错误地声称对标准石油已经付款购买的炼油厂拥有所有权。

Narrator: To further obscure control of the businesses he had purchased, Rockefeller created the Standard Oil Trust in 1882, a legal innovation that allowed him to centralize authority over dozens of companies without technically owning them outright.

讲述者: 为了进一步掩盖对自己所购企业的控制权，洛克菲勒于1882年创建了标准石油托拉斯（Standard Oil Trust），这是一项法律创新，使他能够在技术上不直接拥有数十家公司的情况下，集中行使对它们的控制权。

Narrator: These tactics eventually led to the 1911 Supreme Court decision that broke up Standard Oil under the Sherman Antitrust Act. Today, Big Tech’s use of non-acquisition acquisitions, exclusive infrastructure deals, and strategic talent consolidation echoes Rockefeller’s playbook。

讲述者: 这些策略最终导致了1911年最高法院依据《谢尔曼反垄断法》（Sherman Antitrust Act: 美国第一部也是最重要的反垄断法）作出的拆分标准石油的判决。如今，科技巨头们所采用的“非收购式收购”、排他性基础设施交易以及战略性人才整合，无不呼应着洛克菲勒当年的策略手册。

案例分析：Meta与Scale AI的“人才收购”交易

Narrator: Meta's – recently announced $14.8 billion investment in Scale AI which involved hiring the data-labeling startup's CEO will test how the Trump administration views acquihire deals – where a company buys a startup as a means of hiring its most important staff, which many have criticized as an attempt to evade regulatory scrutiny.

讲述者: Meta最近宣布对Scale AI进行148亿美元的投资，其中包括雇佣这家数据标注初创公司的CEO。这笔交易将考验特朗普政府如何看待“人才收购”（Acquihire：指公司收购另一家公司的主要目的是为了其员工，而非产品或业务）——许多人批评这种做法是逃避监管审查的企图。

Narrator: Meta executed a $14.3 billion dollar investment in Scale AI, a leading data-labeling firm earlier this month. On paper, Meta bought a 49% non-voting stake in Scale. But in practice, they gained exclusive access to Scale’s human-in-the-loop data pipeline, the migration of Scale’s founder Alexandr Wang and key staff to Meta’s new “superintelligence” unit, and a web of contractual arrangements that effectively embedded Scale’s capabilities within Meta’s AI infrastructure.

讲述者: 本月早些时候，Meta对领先的数据标注公司Scale AI执行了143亿美元的投资。从纸面上看，Meta购买了Scale公司49%的无投票权股份。但实际上，他们获得了对Scale的人机协同（Human-in-the-loop：指在人工智能模型训练或决策过程中引入人类干预的系统）数据管道的独家使用权，Scale的创始人Alexandr Wang及核心员工也转移到了Meta新成立的“超级智能”部门，此外还有一系列复杂的合同安排，这些安排有效地将Scale的能力整合进了Meta的AI基础设施中。

Narrator: The structure of this deal meant that Meta could avoid the formalities of a merger while achieving many of its strategic benefits. In an excellent Substack article – which I’ll link to in the description Drayton D'Silva described the deal as being like the way medieval masons would move entire cathedrals from one site to another, brick by brick - by numbering each brick, transporting it to a new site, and then rebuilding the entire cathedral exactly as before. The reconstructed cathedral – he says - is in a new location but in a spiritual and metaphysical sense, he argues – it’s the same cathedral.

讲述者: 这种交易结构意味着Meta可以在避免正式合并程序的同时，实现诸多战略利益。Drayton D'Silva在一篇精彩的Substack文章中（我会在描述中附上链接）将这笔交易比作中世纪的石匠搬迁大教堂：他们将每块砖石编号，运到新址，然后完全按照原样重建。他表示，重建后的大教堂虽然位于新的地点，但从精神和形而上的意义来看，它仍然是同一座大教堂。

Narrator: Scale AI, which was once an independent supplier to the entire AI ecosystem – once this deal is done - now functions as a de facto internal division of Meta. D’Silva argues that there are multiple elements in the deal structure that make it vulnerable to FTC scrutiny.

讲述者: Scale AI曾是整个AI生态系统的独立供应商，而一旦这笔交易完成，它实际上就成为了Meta的一个内部部门。D’Silva认为，该交易结构中的多个元素使其极易受到FTC的审查。

Narrator: From an antitrust perspective, the deal raises red flags across multiple dimensions: de facto control, input foreclosure, talent consolidation, coordination risk, and dominance entrenchment. Even if regulators ultimately block the deal, Meta may have already achieved a strategic victory by disrupting rival operations and gaining insight into Scale’s operations.

讲述者: 从反垄断的角度来看，这笔交易在多个方面都亮起了红灯：事实上的控制权、投入品封锁（Input foreclosure：指上游企业利用其市场力量，限制或阻止下游竞争对手获取关键生产要素的行为）、人才整合、协调风险以及主导地位的巩固。即使监管机构最终阻止了这笔交易，Meta可能已经通过扰乱竞争对手的运营并洞察Scale的业务，取得了战略性胜利。

Narrator: Regular viewers will notice that I'm not filming from my usual location as I'm traveling right now and today's sponsor Saily makes it so much easier to travel and stay connected without having to pay ridiculous roaming charges find a local SIM card or search around for public Wi-Fi Saily makes it incredibly easy and affordable to stay connected while traveling you just download it once and use it in over 180 countries you choose how much data you want and for how long it's much cheaper than roaming and it takes seconds to set up you just select the country and plan and activate an ESIM before you leave home and when you land your phone automatically connects to a local network with no hidden charges there's no fussing around at the airport swapping SIM cards you can just book an Uber access GPS or make a call if you're traveling this summer scan this QR code to download the app choose a plan in the country you're going to and use my code boil at checkout to get an exclusive 15% off your first purchase

讲述者: 熟悉我的观众可能会注意到，我现在正在旅行，所以拍摄地点与往常不同。今天的赞助商Saily让旅行和保持网络连接变得异常轻松，无需支付高昂的漫游费、寻找本地SIM卡或到处搜索公共Wi-Fi。Saily提供了一种极其便捷且经济实惠的旅行上网方式，您只需下载一次，即可在超过180个国家使用。您可以自由选择所需的数据量和使用时长，这比漫游便宜得多，而且设置过程只需几秒钟。您只需选择国家和套餐，在家出发前激活ESIM（嵌入式SIM卡：一种直接嵌入设备的可编程SIM卡，无需物理插拔），当您落地时，手机就会自动连接到本地网络，没有任何隐藏费用。您无需在机场手忙脚乱地更换SIM卡，可以直接预订Uber、使用GPS或拨打电话。如果您今年夏天有旅行计划，可以扫描屏幕上的二维码下载该应用，选择您目的地的套餐，并在结账时使用我的代码“boil”，首次购买可享受15%的独家折扣。

科技巨头的普遍困境与AI增长策略

Narrator: In a funny piece about Mark Zuckerberg’s new “bro look” in the FT, Hannah Murphy argues that Facebook feels desperately dated today. Instagram has been losing ground to TikTok for quite some time, and the “metaverse” – after which Mark Zuckerberg renamed the entire company - is a ghost town – but without even ghosts…

讲述者: 在《金融时报》一篇关于马克·扎克伯格新“兄弟范”形象的有趣文章中，Hannah Murphy指出，如今的Facebook给人一种极度过时的感觉。Instagram在与TikTok的竞争中早已节节败退，而扎克伯格为之更名整个公司的“元宇宙”（Metaverse：一个集体的、虚拟的共享空间，由虚拟现实、增强现实和互联网融合而成）则是一座连鬼魂都没有的鬼城。

Narrator: The FTC is already pursuing an antitrust case against the firm, saying that its acquisitions of Instagram and WhatsApp were anticompetitive and aimed at monopolizing the social media market. The case is currently “in trial”, with arguments focused on whether Meta's acquisitions were part of a broader strategy to eliminate competition and maintain its dominance.

讲述者: FTC已经对Meta提起了反垄断诉讼，指控其收购Instagram和WhatsApp的行为具有反竞争性，旨在垄断社交媒体市场。该案件目前正处于审理阶段，辩论的焦点在于Meta的收购是否是其消除竞争、维持主导地位的更广泛战略的一部分。

Narrator: Meta has struggled from the very beginning to innovate internally and has mostly had to rely on acquisitions for growth. If Meta loses this antitrust case - the company could be forced to sell off Instagram and WhatsApp. They would – most likely – still be allowed to hold on to the metaverse and those creepy AI glasses that Zuckerberg wears with the built-in cameras – just because nobody else actually uses them.

讲述者: 从一开始，Meta就在内部创新方面举步维艰，其增长主要依赖于收购。如果Meta输掉这场反垄断官司，公司可能被迫出售Instagram和WhatsApp。不过，他们很可能仍被允许保留元宇宙和扎克伯格戴的那副带有内置摄像头的、令人毛骨悚然的AI眼镜——仅仅因为根本没有别人在用。

Narrator: Meta has been very aggressive in its attempts to grow in generative AI and the new investments that Meta – and almost all of the other big tech firms have been making in AI startups - are not just financial investments—they are strategic - and designed to secure early access to cutting-edge models, proprietary data, and elite research talent.

讲述者: Meta在生成式AI领域的增长尝试一直非常激进。而Meta以及几乎所有其他大型科技公司对AI初创公司的投资，不仅仅是财务投资，更是战略布局，旨在确保能优先获得前沿模型、专有数据和顶尖研究人才。

Narrator: These deals – it could be argued - are driven by a belief amongst big tech CEO’s that the next wave of AI breakthroughs will be shaped not just by algorithms, but by control over the infrastructure and inputs that power them. Data-labeling pipelines, cloud compute access, and model training workflows have become the new battlegrounds.

讲述者: 可以说，这些交易的驱动力源于科技巨头CEO们的一个信念：下一波AI突破将不仅由算法决定，更取决于对驱动这些算法的基础设施和生产要素的控制。数据标注管道、云计算资源和模型训练工作流已经成为新的战场。

深度剖析：微软、亚马逊、谷歌、英伟达的AI布局

Narrator: Microsoft’s relationship with OpenAI is a masterclass in contractual leverage – but the deal still has its flaws. Since 2019, Microsoft has invested over $13 billion dollars into OpenAI’s for-profit subsidiary, securing exclusive commercial rights to its models and integrating them deeply into its Azure cloud platform and enterprise software products. Yet, despite the scale of this investment, Microsoft holds no formal equity stake in OpenAI’s nonprofit parent and no voting control over its board.

讲述者: 微软与OpenAI的关系是合同杠杆运用的典范，但这项交易也存在缺陷。自2019年以来，微软向OpenAI的营利性子公司投资超过130亿美元，获得了其模型的独家商业权利，并将其深度集成到Azure云平台和企业软件产品中。然而，尽管投资规模巨大，微软在OpenAI的非营利母公司中并无正式股权，对其董事会也没有投票控制权。

Narrator: This structure was carefully designed to avoid the appearance of a merger. Microsoft didn’t report the transaction to the Federal Trade Commission at the time because the investment didn’t amount to control of the company under U.S. law.

讲述者: 这种结构是精心设计的，旨在避免看起来像一次合并。当时，微软并未向FTC报告这笔交易，因为根据美国法律，该投资并未构成对公司的控制。

Narrator: OpenAI remains - an independent entity—at least on paper. But in practice, Microsoft has a level of influence that rivals or exceeds that of a majority shareholder. The FTC is now investigating whether this partnership constitutes a de facto acquisition.

讲述者: 至少在纸面上，OpenAI仍然是一个独立的实体。但在实践中，微软的影响力堪比甚至超过了控股股东。FTC目前正在调查这种合作关系是否构成事实上的收购。

Narrator: A year and a half ago when OpenAI's board of directors ousted Sam Altman over concerns about AI safety - Microsoft received just minutes’ notice of what was going on, and its executives were not consulted in the decision. At the time Microsoft offered Altman and the core research team roles within Microsoft – which would have likely drawn regulatory scrutiny.

讲述者: 一年半前，当OpenAI董事会因对AI安全的担忧而罢免Sam Altman时，微软仅在事发前几分钟才得到通知，其高管并未参与决策过程。当时，微软为Altman和核心研究团队提供了在微软内部的职位，此举很可能会引起监管机构的审查。

Narrator: The bizarre structure of Open AI which includes a non-profit entity, and a capped-profit subsidiary has caused all sorts of conflicts. Now that the company is trying to convert into a for profit company, The FT reports that Microsoft has considered halting discussions if the two sides can’t agree on critical issues, like the size of Microsoft’s future stake in OpenAI.

讲述者: OpenAI奇特的结构——包含一个非营利实体和一个有利润上限的子公司——已经引发了各种冲突。据《金融时报》报道，现在该公司正试图转型为一家营利性公司，而微软已考虑，如果双方在关键问题上（如微软未来在OpenAI中的持股规模）无法达成一致，将中止相关讨论。

Narrator: If Microsoft was to walk away from these discussions, it could still rely on the existing commercial contract to retain access to OpenAI’s technology until 2030.

讲述者: 即使微软退出这些讨论，它仍然可以依据现有的商业合同，在2030年之前继续使用OpenAI的技术。

Narrator: The Wall Street Journal recently reported - that OpenAI executives have discussed levelling antitrust accusations against Microsoft as a last-ditch option for loosening the firm’s control under their deal – so while Microsoft’s deal may have been structured to avoid FTC scrutiny – the fact that they don’t have full control is causing them plenty of problems.

讲述者: 《华尔街日报》最近报道称，OpenAI的高管们曾讨论过，作为最后的手段，可以对微软提出反垄断指控，以削弱其在协议下的控制力。因此，尽管微软的交易结构可能是为了规避FTC的审查，但其未能实现完全控制的事实，正给他们带来诸多麻烦。

Narrator: Amazon and Google have each invested billions of dollars in Anthropic, the AI startup behind the Claude family of models. In return, Claude is hosted on AWS and Google Cloud, raising concerns about cloud dependency and preferential access. Nvidia, meanwhile, has taken equity stakes in dozens of AI startups, bundling those investments with early access to hardware, developer support, and co-marketing opportunities.

讲述者: 亚马逊和谷歌都向Anthropic投资了数十亿美元，这家AI初创公司是Claude系列模型的开发者。作为回报，Claude托管在AWS和Google Cloud上，这引发了关于云依赖和优先访问权的担忧。与此同时，英伟达则入股了数十家AI初创公司，并将这些投资与硬件的优先使用权、开发者支持以及联合营销机会捆绑在一起。

Narrator: In one of the stranger deals – Nvidia invested in Coreweave – a business that buys and rents out access to Nvidia graphics processing units for training artificial intelligence models.

讲述者: 在一笔较为奇特的交易中，英伟达投资了Coreweave——这家公司的业务是购买并出租英伟达的GPU，用于训练人工智能模型。

Narrator: Nvidia’s investment strategy creates a self-reinforcing loop: it funds the companies most likely to buy its products, and in doing so, ensures continued demand for its hardware. This dual role raises questions about conflicts of interest and fair market practices. Preferential treatment can distort competition—not because Nvidia is denying access outright to certain competitors, but because it is subtly shaping the playing field through capital and supply chain influence.

讲述者: 英伟达的投资策略形成了一个自我强化的循环：它资助那些最有可能购买其产品的公司，从而确保了对其硬件的持续需求。这种双重角色引发了关于利益冲突和公平市场实践的疑问。这种优惠待遇可能会扭曲竞争——并非因为英伟达直接拒绝向某些竞争对手提供产品，而是因为它通过资本和供应链的影响力，在暗中重塑竞争格局。

“人才收购”策略的进一步审视

Narrator: The term "acquihire" is used in finance to describe the acquisition of a company primarily to hire its employees, rather than for the operating business. It's a strategy used by larger companies to quickly acquire specific talent, particularly in industries where skilled teams can be hard to find through traditional recruitment.

讲述者: “人才收购” (acquihire) 是一个金融术语，指主要为了雇佣其员工而收购一家公司，而非看重其经营业务。这是大公司用来快速获取特定人才的一种策略，尤其是在那些难以通过传统招聘方式找到熟练团队的行业。

Narrator: The Information reported last week that Meta is in advanced talks to hire the prominent AI investors Nat Friedman and Daniel Gross to help lead its AI efforts, and that as part of those talks, Meta might partially buy out Friedman and Gross’ venture capital fund, which holds stakes in top AI startups and is worth billions of dollars on paper.

讲述者: The Information上周报道，Meta正在进行深入谈判，计划聘请知名的AI投资人Nat Friedman和Daniel Gross来帮助领导其AI业务。作为谈判的一部分，Meta可能会部分收购Friedman和Gross的风险投资基金，该基金持有顶尖AI初创公司的股份，账面价值达数十亿美元。

Narrator: If the talks are successful, Gross would leave Safe Superintelligence, which he co-founded with the former OpenAI chief scientist last year and work mostly on AI products at Meta.

讲述者: 如果谈判成功，Gross将离开他去年与前OpenAI首席科学家共同创立的Safe Superintelligence公司，转而在Meta主要负责AI产品。

Narrator: As part of the deal, Meta could end up owning minority stakes in the startups that the venture capital firm has invested in but allegedly won’t get information about and control over these startups according to the article.

讲述者: 根据该报道，作为交易的一部分，Meta最终可能持有该风投基金所投资的初创公司的少数股权，但据称不会获得有关这些初创公司的信息和控制权。

监管视角：FTC如何审查新型交易

Narrator: Drayton D’Silva’s Substack article shows how Meta broke its non-acquisition of Scale AI into four main parts, each of which could raise red flags under U.S. antitrust law: Meta Bought 49% of Scale’s shares (but with no voting rights). They got exclusive access to Scale’s data-labeling pipeline (a key resource for training AI models). They Hired Scale’s CEO and key staff (meaning that competitors could no longer access them) and they kept Scale’s CEO on Scale’s board while he leads Meta’s AI lab.

讲述者: Drayton D’Silva的Substack文章揭示了Meta如何将其对Scale AI的“非收购”分解为四个主要部分，每一部分都可能触犯美国反垄断法：1. Meta购买了Scale 49%的股份（但无投票权）；2. 他们获得了Scale数据标注管道的独家使用权（这是训练AI模型的关键资源）；3. 他们雇佣了Scale的CEO和核心员工（意味着竞争对手无法再接触到他们）；4. 他们让Scale的CEO在领导Meta AI实验室的同时，继续留在Scale的董事会。

Narrator: Each of these moves, on its own, might seem harmless. But together, they give Meta almost complete control—without calling the deal a merger.

讲述者: 单独来看，这些举措中的每一步似乎都无伤大雅。但合在一起，它们在不称之为“合并”的情况下，赋予了Meta近乎完全的控制权。

Narrator: Drayton points out that traditional antitrust enforcement was built around clear ownership and control, but in the world of technology, influence can be exerted through contracts, infrastructure, 和 informal arrangements. The Federal Trade Commission and Department of Justice have recently started to adopt a “substance over form” approach, focusing on the economic reality of deals rather than their legal structure.

讲述者: Drayton指出，传统的反垄断执法是围绕明确的所有权和控制权构建的，但在科技领域，影响力可以通过合同、基础设施和非正式安排来施加。FTC和司法部最近已开始采取“实质重于形式”的审查方法，关注交易的经济实质而非其法律结构。

Narrator: The FTC has published 11 Merger Guidelines, and Drayton focuses on guidelines 3, 4, 5, 6 and 11 in his piece - which are most relevant to this structured deal. Guideline 11 is about how minority stakes can still give control, and under this deal Meta owns 49% of Scale and might be able to influence decisions with that leverage. Guideline 5 is about how blocking rivals from access to key resources is bad – and under this deal Meta gets exclusive access to Scale’s data. Guideline 4 is about how hiring away key people can kill competition and – in this deal -Meta is hiring away Scale’s CEO and other top talent. Guideline 3 is about how shared board members can lead to collusion, and in this deal Scale’s CEO is on both sides. Guideline 6 is about how deals that make a dominant firm even stronger are risky, and Meta weakens its AI rivals by disrupting their access to Scale.

讲述者: FTC发布了11项《合并指南》，Drayton在他的文章中重点分析了与此类结构化交易最相关的第3、4、5、6和11条指南。指南11涉及少数股权如何仍能赋予控制权，在此交易中，Meta拥有Scale 49%的股份，可能利用此杠杆影响决策。指南5阐明了阻止竞争对手获取关键资源是有害的，而此交易中Meta获得了Scale数据的独家使用权。指南4指出挖走关键人才可能扼杀竞争，此交易中Meta正是挖走了Scale的CEO及顶尖人才。指南3说明了共享董事会成员可能导致串通，此交易中Scale的CEO身兼两职。指南6强调了使主导企业更强大的交易具有风险，而Meta通过扰乱竞争对手对Scale的访问，削弱了其AI竞争对手。

Narrator: He argues that even if the FTC or the courts block the deal - Zuckerberg already has a win. OpenAI, xAI, Google, and Microsoft have already scaled back or paused their work with Scale AI, citing data-privacy risks and conflict of interest issues. Anthropic has not announced anything publicly but it also works with Scale. Together these five labs plan to spend over $250 billion on AI-related capex in 2025. Meta has already induced months-long delays and disruptions to their critical workflows, likely costing them billions. Drayton estimates that setting up and announcing this deal cost Meta approximately $40 million dollars across external advisers and internal resources over 6 months, which is a very effective asymmetric attack on their rivals.

讲述者: 他认为，即使FTC或法院最终阻止了这笔交易，扎克伯格也已经赢了。OpenAI、xAI、谷歌和微软已经因数据隐私风险和利益冲突问题，缩减或暂停了与Scale AI的合作。Anthropic虽未公开发表声明，但它也与Scale合作。这五家实验室计划在2025年在AI相关的资本支出（Capex：企业为获取或升级固定资产的投资）上花费超过2500亿美元。Meta已经成功地对它们的关键工作流程造成了长达数月的延误和中断，可能使其损失数十亿美元。Drayton估计，策划并宣布这笔交易在6个月内花费了Meta约4000万美元（包括外部顾问和内部资源），这对竞争对手来说是一次非常有效的不对称攻击。

结论：AI竞赛的未来格局

Narrator: He finishes up his piece with an amusing playbook for non-acquisitions for Tech CEO’s hoping to avoid regulatory scrutiny – I’ll leave a link to his Substack in the video description – it’s worth checking out.

讲述者: 他在文章结尾为希望规避监管的科技公司CEO们提供了一份有趣的“非收购”操作手册——我会在视频描述中留下他Substack的链接，值得一看。

Narrator: The deals being struck by Meta, Microsoft’s - OpenAI partnership, and Nvidia’s customer investment loop all challenge regulators to look beyond formal mergers in assessing competition within an industry. While the deals that are being struck might not trigger traditional review thresholds, their cumulative effect can without a doubt be expected to reshape competition within an industry.

讲述者: Meta的交易、微软与OpenAI的合作关系，以及英伟达的“客户即投资”循环，都在挑战监管机构，促使他们在评估行业竞争时，必须超越传统的正式合并框架。虽然这些正在达成的交易可能不会触发传统的审查门槛，但它们的累积效应无疑将重塑行业内的竞争格局。

Narrator: Big tech CEOs don’t seem to just expect AI to be a technological shift— but a structural shift too. They believe that as foundational models become the engines of the next digital economy, the companies that control their development, deployment, and distribution could end up wielding extraordinary influence. Unlike past industrial consolidations, this one is being shaped less through traditional mergers and more through strategic entanglements.

讲述者: 科技巨头的CEO们似乎不仅仅将AI视为一次技术变革，更将其看作一次结构性变革。他们相信，随着基础模型成为下一个数字经济的引擎，那些控制其开发、部署和分销的公司最终将拥有非凡的影响力。与过去的产业整合不同，这一次的整合更多是通过战略性的“纠缠”而非传统合并来形成的。

Narrator: While regulators may not have noticed this in the past – the sheer volume of these deals has brought this issue to their attention. At present it would appear that there is plenty of competition in the AI space – There are a number of competitors and they are all racing to develop better and better models. Some of the huge investment in the space has (quite likely) been driven by a belief that AI will have - winner take all economics - like internet search did – where google built the best search engine many years ago - and then dominated that sector for the next few decades.

讲述者: 尽管监管机构过去可能没有注意到这一点，但这些交易的庞大规模已经引起了他们的关注。目前看来，AI领域竞争激烈——众多竞争者都在竞相开发更优秀的模型。该领域的部分巨额投资很可能是基于一种信念，即AI将具有“赢家通吃”的经济效应，就像当年的互联网搜索一样——谷歌在多年前打造出最好的搜索引擎后，主导了该领域长达数十年。

Narrator: Since the DeepSeek breakthrough earlier this year – there has been growing concern amongst investors - that while it might be extremely expensive to achieve AI breakthroughs – the new models – once built - might be quite easy to replicate – meaning that there may not be a dominant AI company in the near future and that the industry could instead remain extremely competitive with very little pricing power.

讲述者: 自今年早些时候DeepSeek取得突破以来，投资者中日益增长的一种担忧是：尽管实现AI突破的成本可能极其高昂，但新模型一旦建成，可能很容易被复制。这意味着短期内可能不会出现占主导地位的AI公司，整个行业反而可能保持极度竞争的状态，几乎没有定价权。

Narrator: Meta, Microsoft, Amazon, Google, and Nvidia are – without a doubt - trying to build moats around AI—one contract at a time – but regulators do appear to already be “on to what’s going on” and could shut these deals down if they appeared to be harming competition in the industry.

讲述者: 毫无疑问，Meta、微软、亚马逊、谷歌和英伟达正在试图通过一份份合同，为AI构建护城河。但监管机构似乎已经“洞悉了一切”，如果这些交易被认为损害了行业竞争，他们就可能叫停这些交易。

Narrator: We have all seen the sudden transformation of Mark Zuckerberg from the left leaning tech nerd who wrote an open letter to his child about his goal of building inclusive and welcoming communities while pouring millions into social justice causes – into the gold chain wearing MMA bro who told Joe Rogan that Corporate America had been “culturally neutered” and workplaces needed more “masculine energy.” This all happened right around the time of Trumps election victory, and this metamorphosis could be explained as much by Zuckerberg’s need to align himself with the new administration as an actual change in beliefs.

讲述者: 我们都目睹了马克·扎克伯格的突然转变：从一个左倾的科技宅男——他曾给自己的孩子写公开信，谈论建立包容和友善社区的目标，并向社会正义事业投入数百万美元——变成了一个戴着金链子的综合格斗爱好者，向Joe Rogan表示美国企业界已被“文化阉割”，工作场所需要更多“男性气概”。这一切都发生在特朗普胜选前后，这种蜕变既可以解释为扎克伯格信仰的真实改变，也可以看作是他为了与新政府保持一致的需要。

Narrator: Should any of these deals be deemed anticompetitive – being aligned with the current administration (along with the purchase of the correct meme coins) could solve a lot of Zuckerberg’s problems.

讲述者: 如果这些交易中的任何一笔被认定为反竞争——那么与现任政府保持一致（或许再加上购买正确的模因币(Meme Coins: 源于互联网模因或笑话的加密货币，通常波动性极大)）可能会解决扎克伯格的许多问题。

Narrator: As an example, the US Supreme Court unanimously upheld a federal law requiring TikTok to either be sold by its Chinese parent company or be banned in the U.S. more than six months ago but enforcement has been repeatedly delayed by executive orders from the US president – who even invited it’s CEO to his inauguration.

讲述者: 举个例子，六个多月前，美国最高法院一致支持一项联邦法律，要求TikTok要么由其中国母公司出售，要么在美国被禁用。但该法律的执行却因美国总统的行政命令而一再推迟——总统甚至邀请了TikTok的CEO参加他的就职典礼。

Narrator: It is not obvious that any of the big tech firms have anything like a monopoly on AI at present –it appears to be an extremely competitive space. While the technology is driving massive investment and strategic growth, profitability remains elusive for most AI companies other than NVIDIA. According to the Wall Street Journal - Microsoft loses around $20 per user per month on GitHub Copilot.

讲述者: 目前尚不明显有任何一家科技巨头在AI领域形成了垄断——这似乎是一个竞争异常激烈的领域。尽管该技术正推动着巨额投资和战略增长，但除了英伟达之外，大多数AI公司的盈利能力仍然遥不可及。据《华尔街日报》报道，微软在GitHub Copilot上每位用户每月亏损约20美元。

Narrator: The expectation for these companies – whose core businesses are very profitable – appears to be that long-term gains will come as AI becomes more efficient, cheaper to run, and more deeply integrated into enterprise workflows. At present they have a lot of money – and are spending some of it on generative AI to make sure that they are still competitive in the future. They are also spending this money as a moonshot investment – where maybe they will develop a dominant product – which will matter as much to the future internet as search did – to the internet - twenty-five years ago.

讲述者: 对于这些核心业务非常盈利的公司来说，他们的期望似乎是，随着AI变得更高效、运行成本更低，并更深度地融入企业工作流程，长期收益将会显现。目前，他们拥有大量资金，并将其一部分投入生成式AI，以确保未来仍具竞争力。他们也将这笔钱视为一次“登月式”投资——或许他们能开发出一款主导性产品，其对未来互联网的重要性，将不亚于二十五年前搜索对互联网的意义。

Narrator: The Economist made a good argument in a recent article - that the FTC case calling for the breakup of Meta is without merit. They say that the FTC’s argument against Meta wasn’t compelling when it was launched in 2020, and in the years since the complaint was filed it has only become less persuasive.

讲述者: 《经济学人》在最近一篇文章中提出了一个有力的论点，认为FTC要求拆分Meta的诉讼缺乏依据。他们表示，FTC在2020年提起诉讼时其论点就不具说服力，而在诉讼提出后的几年里，其说服力更是日益减弱。

Narrator: The FTC’s claim in the case is that network effects create insurmountable barriers to entry, but the Economist points out that the last few years have seen a rush of new successful social-media startups. Some have fizzled, like Clubhouse and BeReal, others remain small, like Bluesky and Truth Social. But the arrival of TikTok, made a mockery of the idea that the market was closed to new entrants.

讲述者: FTC在案件中的主张是，网络效应构成了不可逾越的进入壁垒。但《经济学人》指出，过去几年见证了一批新兴社交媒体初创公司的成功。有些昙花一现，如Clubhouse和BeReal；有些规模尚小，如Bluesky和Truth Social。但TikTok的出现，则彻底颠覆了“市场对新进入者关闭”的说法。

Narrator: They point out that - partially because of TikTok, social media apps now revolve around short form video, which has put Meta’s Instagram and Facebook in competition with YouTube shorts and Twitter who all are pursuing the same short form video audience that TikTok drew in.

讲述者: 他们指出，部分由于TikTok的影响，社交媒体应用现在都围绕短视频展开，这使得Meta的Instagram和Facebook与YouTube Shorts和Twitter（现为X）展开了竞争，它们都在争夺TikTok吸引来的同一批短视频观众。

Narrator: They highlight that Spotify, which was originally a music service, is pushing into video too. Roblox, a gaming platform, acts as a social network for many teenagers. They say that as competition spreads across previously separate categories, the FTC’s narrow market definition looks ever sillier.

讲述者: 他们还强调，最初是音乐服务的Spotify也在向视频领域进军。游戏平台Roblox则充当了许多青少年的社交网络。他们认为，随着竞争跨越了以往界限分明的类别，FTC狭隘的市场定义显得愈发荒谬。

Narrator: There is no doubt that all of these AI deals are being specifically designed to work around the FTC rulebook – and it does appear that some of the deals are anticompetitive in nature – in that they are being done to block access to key AI services – slowing down competitors – but this scramble to invest and build better and better models, while keeping prices low - doesn’t appear to be harming consumers and reflects how competitive the space is at present.

讲述者: 毫无疑问，所有这些AI交易都是为了规避FTC的规则手册而精心设计的。而且，确实有些交易本质上是反竞争的——因为它们旨在阻止竞争对手获取关键的AI服务，从而拖慢其发展。但是，这种争相投资、构建更优模型、同时保持低价的竞争态势，目前看来并未损害消费者利益，反而反映了该领域的激烈竞争程度。

Narrator: While these deals are worth keeping an eye on – it is not obvious to me that they will be stopped anytime soon.

讲述者: 虽然这些交易值得密切关注，但在我看来，它们短期内不太可能被叫停。

Narrator: If you found this video interesting – you should watch my video on Deepseek next. Don’t forget to check out our sponsor Saily using the link in the video description. Have a great day and talk to you in the next video – bye.

讲述者: 如果您觉得本期视频有趣，接下来可以观看我关于DeepSeek的视频。别忘了通过视频描述中的链接了解我们的赞助商Saily。祝您有美好的一天，我们下期视频再见！