Broadcom AI Revenue Outlook Falls Short of Expectations 博通AI营收展望不及预期

The market’s obsession with “expectations” has long distorted into a numbers game. Broadcom’s $56 billion figure would have been enough to make the entire semiconductor sector tremble in 2024, but within the grand narrative of “all tech giants investing frantically in AI,” it’s now deemed “not aggressive enough.” Nvidia’s trillion-dollar market cap myth has stretched everyone’s thresholds to unrealistic heights—as if any company linked to AI chips must sustain triple-digit growth to be considered competent. This collective, unconscious revelry masks the core question: when—and how exactly—will the massive computing investments made by downstream clients like Microsoft and Google convert into stable, scalable profits? For now, most of it remains in a “secure the position first, sort it out later” phase. As a seller of shovels, Broadcom’s outlook is remarkably calm; it resembles a seasoned infrastructure builder calculating the true cost of each brick, rather than dancing to the market’s dopamine rhythm.

In stark contrast to the hardware sector’s “calm frenzy,” the software application layer is experiencing an “enthusiastic backlash.” The UK’s Competition and Markets Authority ordered Google to modify its AI search feature, allowing websites to opt out of “AI Overviews”—a resounding slap. Google’s AI Overviews essentially function as an “answer monopoly” based on highly distilled web content, condensing the hard work of countless content producers into a few lines at the top, so users don’t even need to click through to the original page. For publishers and creators, this is akin to pulling the rug from under them: you provide the content, Google’s AI eats it all, then tosses you a fraction of the residual traffic? The UK regulator’s move精准地 strikes the most sensitive nerve in AI application ethics: the line between efficiency and fairness, innovation and extraction. AI isn’t creating a new world—it’s redistributing the old one’s pie, and the knife is currently held tight by a handful of giants.

These two stories together resemble two sides of the AI industry’s coin. On one side is the “infrastructure construction frenzy” represented by Broadcom, where capital and engineers lay the digital world’s foundation with almost obsessive passion. On the other is the systemic pushback triggered by the “application-layer power expansion” symbolized by Google. We stand at a moment where computing power grows exponentially, but its uses, rules, and distribution methods remain stuck in Stone Age debates. The 1.3 gigawatts in Broadcom’s plan could power a mid-sized city’s electricity needs—so what will it ultimately serve? Will it help AI craft more convincing fake news, or aid scientists in discovering new protein structures? The answer remains unclear, yet capital has already surged ahead.

Other trending news offers much to ponder: Intel’s attempt to challenge Nvidia with new technology reads like a belated revenge; Windows’ full embrace of agents means AI will embed itself into the digital fabric of our lives like an operating system; Tencent, Alibaba, and ByteDance are battling over a Skill Store—essentially jostling for tickets to tomorrow’s AI-native app ecosystem. All these developments underscore a fact: AI is no longer a lab toy. It’s becoming the underlying operating system reshaping all business rules. But here’s the catch—when everyone rushes to board the vehicle without a clear destination, might the overload itself cause a breakdown?

Broadcom’s “falling short” might be the market’s first healthy sign of returning to rationality. It compels investors to lift their gaze from the grand narrative of “AI will change everything” and scrutinize concrete delivery capabilities, profit timelines, and return on investment. Similarly, Google’s regulatory predicament warns us that while technology rolls forward, society and law will always try to seize the reins. AI’s development will never follow a smooth upward curve; it will be a hybrid race of sprinting, collisions, and recalibrations. Broadcom is building rockets, Google is driving race cars, and regulators are erecting new speed limit signs trackside. Now that’s a scene far more gripping than any mere numbers game.

博通刚刚公布的2026财年AI芯片销售额展望为560亿美元，第三财季预期160亿美元，市场一片“不及预期”的叹息声。但真正的问题或许是：华尔街那帮穿定制西装的分析师，他们的预期是不是已经脱离了地心引力，直接奔向了火星？博通自己宣布已向OpenAI交付芯片，2027年要部署1.3吉瓦的算力，这听起来像在给一个刚学会走路的孩子规划奥运会夺金路线图。一边是泼天的数字，一边是冷淡的股价反应，这中间的裂缝，恰好能塞进当前整个AI硬件赛道最真实的尴尬：前景无比性感，兑现无比骨感。

市场对“预期”的迷恋，早已异化成一种数字游戏。博通给出的560亿，是一个在2024年听起来足以让整个半导体板块颤抖的数字，但放在“所有科技巨头都在疯狂投资AI”的宏大叙事里，它就变成了“不够激进”。英伟达的万亿市值神话，把所有人的阈值都拉到了不切实际的高度。仿佛任何一家沾上AI芯片的公司，都必须保持三位数的增长才算合格。这种集体无意识的狂欢，恰恰掩盖了最核心的问题：下游那些挥舞支票本的客户，比如微软、谷歌，它们投入的巨额算力，到底何时、以何种方式，才能变成稳定且可规模化的利润？目前看来，大部分还停留在“先占坑再说”的阶段。博通作为卖铲子的，其展望其实异常冷静，它更像一个老练的基建商，在计算着每一块砖瓦的真实成本，而非跟着市场的多巴胺节奏跳舞。

与硬件端这种“冷静的狂热”形成有趣对照的，是软件应用层正在遭遇的“热情的反噬”。英国竞争与市场管理局勒令谷歌整改AI搜索，允许网站退出“AI概览”，这记耳光打得响亮。谷歌的AI概览功能，本质上是一种基于全网内容的、高度精炼的“答案垄断”。它把无数内容生产者辛苦耕耘的信息，浓缩成顶部的几行字，用户甚至不需要点击进入原网页。这对出版商、创作者而言，无异于釜底抽薪——你贡献了内容，谷歌用AI吃干抹净，然后分你一点残渣流量？英国监管这次出手，精准地打在了AI应用伦理争议最疼痛的神经上：效率与公平、创新与掠夺之间的边界。AI不是在创造新世界，而是在重新划分旧世界的利益蛋糕，而分蛋糕的刀，目前牢牢握在少数几家巨头手里。

这两则消息并列，像极了AI产业的一体两面。一边是博通代表的“基础设施建设狂潮”，资本与工程师正以近乎偏执的热情浇筑数字世界的地基；另一边是谷歌代表的“应用层权力扩张”所引发的系统性反弹。我们正处在这样一个时刻：算力以指数级增长，但算力的用途、规则和分配方式，却还停留在石器时代的辩论阶段。博通计划中的1.3吉瓦算力，足够支撑一个中型城市的所有用电，而它最终将服务于什么？是让AI写出更逼真的假新闻，还是帮助科学家发现新的蛋白质结构？答案并不清晰，资本却已经先一步到位。

热榜里其他新闻也很有嚼头：英特尔试图用新技术挑战英伟达，这像一场迟到的复仇；Windows全面拥抱Agent，意味着AI将像操作系统一样嵌入你我的数字生活底层；腾讯、阿里、字节在混战Skill商店，这是在为未来的AI原生应用生态抢夺“应用商店”的门票。所有这些动态都在佐证一个事实：AI早已不是实验室里的玩具，它正在成为重塑所有商业规则的底层操作系统。但问题在于，当所有人都急着上车，甚至还没想好目的地的时候，这辆车本身，会不会因为超载而提前抛锚？

博通的“不及预期”，或许才是市场回归理性的第一个健康信号。它迫使投资者从“AI必将改变一切”的宏大叙事中抬头，开始审视具体的交付能力、盈利周期和投资回报率。同样，谷歌的监管困境则警告我们，技术的车轮滚滚向前，但社会和法律总会试图拉住缰绳。AI的发展注定不是一条平坦的上升曲线，而是一场在狂奔、碰撞、调整中反复前行的混合赛跑。博通在造火箭，谷歌在开赛车，而监管机构正在赛道旁竖起新的限速牌。这场面，可比单纯的数字游戏刺激多了。