Artificial intelligence will change everything from booking flights to brewing beer. Financial analysts estimate that AI will raise global GDP by about 10% to 20% over the coming decade. Just as most businesses became internet businesses in the 1990s and 2000s, nearly every business will be an AI business a decade from now.

人工智能将改变从预订航班到酿造啤酒的一切。金融分析师估计，人工智能将在未来十年内将全球 GDP 提高约 10% 至 20%。正如大多数企业在 1990 年代和 2000 年代成为互联网企业一样，十年后几乎每家企业都将是 AI 企业。

To deliver the compute power fueling this transformation, companies, countries, and capital markets have mobilized resources at an unprecedented scale. When Chinese AI firm DeepSeek released its latest open-source model at a fraction of the expected training cost, it called these investments into question, inviting a tech selloff. Yet shares have rebounded. As economists would point out, for products with strong demand like AI, technological developments that lower costs through resource efficiency drive greater overall usage. Not surprisingly, Meta and Microsoft have since doubled down on their AI investments.

为了提供推动这一转型的计算能力，公司、国家/地区和资本市场以前所未有的规模调动了资源。当中国人工智能公司 DeepSeek 以预期训练成本的一小部分发布其最新的开源模型时，它对这些投资提出了质疑，从而引发了科技股抛售。然而，股市已经反弹。正如经济学家所指出的那样，对于像人工智能这样需求强劲的产品，通过资源效率降低成本的技术发展会推动更高的整体使用。毫不奇怪，Meta 和 Microsoft 此后加倍投入了他们的 AI 投资。

While cost reductions and open-source models make AI development more accessible to startups, supporting demand and pushing the frontier of AI will remain costly. Four major U.S. tech firms in particular—Meta, Alphabet, Microsoft, and Amazon—are spending massively on AI investments. And while returns to these investments have yet to materialize in their bottom lines, the sums needed in this development race have already rewritten the traditional rules of finance.

虽然降低成本和开源模型使初创公司更容易进行 AI 开发，但支持需求和推动 AI 前沿的成本仍然很高。特别是美国四家主要科技公司——Meta、Alphabet、Microsoft 和 Amazon——正在大量投资 AI 投资。虽然这些投资的回报尚未体现在其底线中，但这场发展竞赛所需的资金已经改写了传统的金融规则。

Running on cloud computing credits

使用云计算积分运行

While venture deal volume trended flat to down since the heady days of 2021 and early 2022, the rise of AI brought about two significant shifts. First, it crowded out other areas of investment, with 62% of all North American startup funding headed toward AI companies in the fourth quarter of 2024. Second, while deal volumes are down, deal values are up.

虽然自 2021 年和 2022 年初的高峰期以来，风险投资交易量呈持平下降的趋势，但人工智能的兴起带来了两个重大转变。首先，它排挤了其他领域的投资，2024 年第四季度，北美所有创业资金的 62% 流向了人工智能公司。其次，虽然交易量下降，但交易价值上升。

The staggering costs of compute needed to drive these AI companies sparked not just sky-high funding rounds but also solidified Big Tech’s dominance—few companies have sufficiently deep pockets to place these bets. OpenAI found its patron in Microsoft, Anthropic followed with Amazon. Both eschewed standard investment terms. Only a small part of Microsoft’s $10 billion 2023 investment went directly to OpenAI. Instead, most of the money was allocated as credits for OpenAI to purchase computing power from Microsoft's cloud services.

推动这些 AI 公司所需的惊人计算成本不仅引发了天价融资，还巩固了大型科技公司的主导地位——很少有公司拥有足够雄厚的财力来押下这些赌注。OpenAI 在 Microsoft 找到了它的赞助人，Anthropic 紧随其后是亚马逊。两者都避开了标准的投资条款。Microsoft 2023 年的 100 亿美元投资中，只有一小部分直接流向了 OpenAI。相反，大部分资金被分配为信用额度，供 OpenAI 从 Microsoft 的云服务购买计算能力。

In the age of AI, cloud computing credits have become currency. Access to GPUs and compute has made AI giants like Nvidia the investors of choice for most startups. As Elon Musk told investors at a conference in 2023, “GPUs at this stage are much more difficult to obtain than drugs.” AI companies spent an average of 22% of their expenses on computing costs in the first three months of 2024—more than double the amount spent by traditional software companies. Musk himself put his prodigious fundraising talents to work, closing a $6 billion fundraising round to build the largest supercluster in the world.

在 AI 时代，云计算积分已成为货币。对 GPU 和计算的访问使 Nvidia 等 AI 巨头成为大多数初创公司的首选投资者。正如埃隆马斯克在 2023 年的一次会议上告诉投资者的那样，“现阶段的 GPU 比药物更难获得。在 2024 年前三个月，AI 公司平均将 22% 的支出用于计算成本，是传统软件公司支出的两倍多。马斯克本人将他惊人的筹款才能付诸实践，完成了 60 亿美元的融资，以建造世界上最大的超级集群。

Through strategic funding arrangements offering cloud compute credits, Microsoft, Google, and Amazon have made the up-and-comers of the AI world reliant on the giants of the internet’s Gilded Age. These deals offer a host of benefits beyond the investment itself. The cloud providers not only gather new customers and acquire market share, they also benefit from the high margins on cloud computing services. Valuations increase in turn.

通过提供云计算积分的战略融资安排，Microsoft、Google 和 Amazon 使 AI 世界的后起之秀依赖于互联网镀金时代的巨头。这些交易提供了超出投资本身之外的许多好处。云提供商不仅可以吸引新客户并获得市场份额，还可以从云计算服务的高利润率中受益。估值反过来会增加。

Fueling the AI future  为 AI 未来提供动力

While Big Tech firms can provide access to enough compute to support the growth of the industry, they also need to source sufficient power to actually run the GPUs, which are massively energy-intensive. AI-powered searches require as much as 10 times more electricity than a standard Google search. And usage is expected to grow dramatically over the next five years: Data center demand will likely take up 8% of total U.S. energy demand by 2030.

虽然大型科技公司可以提供足够的计算来支持行业的发展，但他们也需要获得足够的电力来实际运行具有大量能源的 GPU。AI 驱动的搜索需要的电力是标准 Google 搜索的 10 倍。预计未来五年的使用量将大幅增长：到 2030 年，数据中心需求可能占美国能源总需求的 8%。

At the same time, giants like Google and Amazon have been under careful scrutiny for their energy consumption for years, and have responded with bold climate targets over the last decade. Historically, the vast majority of data centers were run off the grid, buying power like everyone else. Meanwhile, long-term power purchase agreements (PPAs) have been common in the energy industry for decades. They provide an energy purchaser with a stable rate and the supplier with guaranteed demand and have been common tools for corporations looking to advance their climate goals. Only in the last few years have tech giants begun using similar agreements as a solution to their new energy demand.

与此同时，谷歌和亚马逊等巨头多年来一直受到对能源消耗的仔细审查，并在过去十年中以大胆的气候目标作为回应。从历史上看，绝大多数数据中心都是在电网上运行，像其他人一样购买电力。与此同时，几十年来，长期购电协议 （PPA） 在能源行业一直很普遍。它们为能源购买者提供稳定的费率，为供应商提供有保证的需求，并且一直是希望推进其气候目标的企业的常用工具。直到最近几年，科技巨头才开始使用类似的协议来满足他们的新能源需求。

In this vein, Microsoft has revived the dormant Three Mile Island nuclear plant in Pennsylvania, the scene of the United States’ worst nuclear accident. Amazon recently purchased a data center campus near a 960-megawatt nuclear power plant for $650 million; the center’s power will move off-grid via an agreement with plant-operator Talen Energy. Google has signed an agreement to source energy from a fleet of mini nuclear reactors, while OpenAI’s Sam Altman and many other tech moguls are investing heavily in nuclear fusion. Energy markets have taken note, with the S&P 500 Utilities Index almost doubling its growth rate in 2024.

本着这种精神，Microsoft 恢复了宾夕法尼亚州休眠的三哩岛核电站，那里是美国最严重的核事故现场。亚马逊最近以 6.5 亿美元的价格购买了一座 960 兆瓦核电站附近的数据中心园区;该中心的电力将通过与发电厂运营商 Talen Energy 达成协议而离网。谷歌已经签署了一项协议，从一组小型核反应堆中获取能源，而 OpenAI 的山姆·阿尔特曼 （Sam Altman） 和许多其他科技大亨正在大力投资核聚变。能源市场已经注意到了这一点，标准普尔 500 公用事业指数在 2024 年的增长率几乎翻了一番。

Gathering AI talent and technology

聚集 AI 人才和技术

Faced with aggressive antitrust actions, Big Tech firms have also adopted creative mechanisms toward industry consolidation. Last year, joint ventures and partnerships increased by 40%. As talent becomes an increasingly valuable resource, companies have returned to “acquihires” to grow their research teams while steering clear of antitrust restrictions.

面对激进的反垄断行动，大型科技公司也采用了创造性的行业整合机制。去年，合资企业和合伙企业增加了 40%。随着人才成为一种越来越有价值的资源，公司纷纷回归“收购”，以扩大其研究团队，同时避开反垄断限制。

In 2023, Microsoft hired Mustafa Suleyman, the cofounder of Inflection AI, a company that raised over a billion dollars to support its own AI model. Along with him came hundreds of key employees and a $650 million licensing deal. In 2024, Google secured a licensing agreement with Character.ai, acquiring the company’s chatbot technology at the same time as it rehired its CEO and other key technologists. Both Microsoft and Google gained access to critical talent while distributing payments to employees and investors with sophisticated deal structures.

2023 年，Microsoft 聘请了 Inflection AI 的联合创始人 Mustafa Suleyman，该公司筹集了超过 10 亿美元来支持自己的 AI 模型。与他同行的还有数百名关键员工和一份价值 6.5 亿美元的许可协议。2024 年，谷歌与 Character.ai 达成许可协议，在重新聘请其首席执行官和其他关键技术人员的同时收购了该公司的聊天机器人技术。Microsoft 和 Google 都获得了关键人才，同时通过复杂的交易结构向员工和投资者分配付款。

While talent-centric acquisitions have been a tool in the tech finance landscape for years, the clever coupling of acquihires with technology licensing is a new phenomenon. The popularity of these deals reflects the vigorous competition over AI talent and the market's shift to consolidation in vanguard frontier models. With the prospect of artificial general intelligence and recursive self-learning creating a flywheel effect of continuous progress, the most valuable asset isn’t a product or service—it is the researchers and engineers driving the revolution.

虽然多年来，以人才为中心的收购一直是科技金融领域的一种工具，但收购与技术许可的巧妙结合是一种新现象。这些交易的受欢迎程度反映了对 AI 人才的激烈竞争以及市场转向先锋前沿模型的整合。随着通用人工智能和递归自我学习的前景产生持续进步的飞轮效应，最有价值的资产不是产品或服务，而是推动革命的研究人员和工程师。

The U.S. tech giants, which now represent a fifth of the S&P 500, will find new ways to adapt the world of finance to deliver on the promise of AI. AI’s productivity promises may still be in their infancy, but the race to procure AI has already delivered us a new era of technology finance.

目前占标准普尔 500 指数五分之一的美国科技巨头将找到新的方法来适应金融世界，以实现 AI 的承诺。 AI 的生产力承诺可能仍处于起步阶段， 但采购 AI 的竞赛已经为我们带来了一个技术金融的新时代。

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

Fortune.com 评论文章中表达的观点仅代表其作者的观点，并不一定反映其观点和信念 财富。

相关标签： 10 2024 AI Microsoft their Google will power companies OpenAI