本文引述 Alphabet Inc. 于 2026 年资本支出（CapEx）预期范围为 1750 亿至 1850 亿美元（published Feb 5, 2026 6:30 PM GMT+8；UTC+8；original: Feb 5, 2026 6:30 PM GMT+8）。Bloomberg 共识显示分析师原预期约 1195 亿美元，该数字本身即被描述为较 2025 年约增加 30%；相较共识上缘 1850 亿美元高出 655 亿美元，约为 1.55 倍，且文中称投资人被要求接受「超过加倍」的支出增幅。
市场起初以股价反应放大成本疑虑：该预测一度使股价在盘后下跌逾 7%，类似 Microsoft Corp. 先前因 AI 建置成本上升而使「不错的业绩」仍不足以抵销成本压力的情境；文中并称 Microsoft 股价在过去 5 天下跌 14%。随后投资人检视其余财报指标后悲观情绪缓和，股价回补跌幅并一度转升，之后小幅走低。
支撑「高 CapEx 仍可被接受」的关键量化指标集中在 AI 与云端成长：Gemini 并入 Gmail、Chrome 等产品后，月活跃使用者（MAU）季增 1 亿，且每位使用者互动度上升；Google Search 被描述为使用量「前所未有」，Search 收入年增 17%。Google Cloud 年增 48%，其云端积压订单（backlog）年增 55%，并称由「多个客户」驱动；对照 Microsoft 表示其 backlog 有 45% 来自单一客户 OpenAI。公司亦以自研 AI 晶片降低对 Nvidia 稀缺且昂贵硬体的依赖，并以需求「全面且极强」作为 2026 年 1750 亿至 1850 亿美元支出规模的依据。
Alphabet Inc. projected 2026 CapEx of $175 to $185 billion (published Feb 5, 2026 6:30 PM GMT+8; UTC+8; original: Feb 5, 2026 6:30 PM GMT+8). Bloomberg consensus showed analysts had expected about $119.5 billion, a figure described as roughly a 30% increase from 2025; the $185 billion top end is $65.5 billion higher, about 1.55x the consensus, and the article characterizes the implied spending step-up as more than double.
The market initially priced the cost shock: the projection pushed shares down more than 7% after hours, echoing Microsoft Corp.’s recent earnings dynamic where solid performance did not offset rising AI build-out costs; Microsoft shares are described as down 14% over the past five days. As investors digested the rest of Alphabet’s results, the selloff faded, losses were retraced, and the stock briefly turned positive before settling slightly lower.
The metrics used to justify the higher CapEx focused on AI adoption and cloud momentum: Gemini’s integration into products such as Gmail and Chrome added 100 million monthly active users quarter-on-quarter, alongside higher engagement per user. Google Search was framed as seeing more usage than ever, with Search revenue up 17% year-on-year. Google Cloud grew 48% year-on-year and its backlog rose 55% year-on-year, described as driven by multiple customers, contrasted with Microsoft’s statement that 45% of its backlog came from a single client, OpenAI; in-house AI chips were also cited as reducing reliance on scarce and expensive Nvidia hardware, supporting the claim that demand across DeepMind and cloud is exceptionally strong.