科技巨頭在資料中心基礎設施上的巨額資本支出正引發投資者日益加劇的不安。Alphabet因歷史上首次出現自由現金流為負而遭到市場拋售，Meta也因堅持大規模投資計畫並拒絕轉售多餘資料中心容量而股價承壓。儘管微軟的AI收入開始加速成長，其股價仍較去年高點下跌約百分之十五，整體半導體股也在近一個月內大幅回落。
市場焦慮的根源有多重因素。首先，儘管短期支出依然強勁，華爾街可能預期資本支出即將見頂，但科技公司的言論並未支持這一判斷。其次，雖然微軟和Google的合併訂單積壓已從一年前不到五千億美元躍升至一點二兆美元，但其中高度集中於OpenAI和Anthropic等尚未證明自身客戶需求的公司，且中國模型的崛起和企業對AI過度使用的管控都為定價前景蒙上陰影。（關鍵數字：1.2）
更深層的結構性變化也令人擔憂。Meta的自由現金流在最新一季大幅下滑，Alphabet長期債務從一年前的一百一十億美元飆升至近一千億美元，並額外籌集了八百五十億美元的股權融資。信用評級機構穆迪指出，隨著大型科技公司轉變為「資產密集型」企業，其與資本、風險和信用的關係正在被重新定義。這些因素單獨來看或許不足以觸發拋售，但綜合在一起，足以解釋為何投資者的焦慮水平正持續攀升。
Massive capital expenditure on data centre infrastructure by major tech companies is provoking growing investor unease. Alphabet was punished by the market after burning through cash for the first time as a public company, while Meta's stock suffered after it doubled down on heavy investment and dashed hopes of reselling surplus capacity. Even Microsoft, which reported meaningful AI revenue acceleration, remains roughly fifteen per cent below its peak, and semiconductor shares have also fallen sharply over the past month.
Several factors underpin the anxiety. Wall Street may believe peak capex is approaching, yet nothing in the companies' guidance supports that view. Although combined order backlogs at Microsoft and Google surged to 1.2 trillion dollars, this demand is heavily concentrated on OpenAI and Anthropic, neither of which has proven its end-customer base can absorb all that capacity. Meanwhile, the rise of competitive Chinese models and corporate crackdowns on excessive AI usage—so-called tokenmaxxing—cast doubt on sustained pricing power and profitability.
A deeper structural transformation is also unsettling markets. Meta's free cash flow dropped sharply, and Alphabet's long-term debt ballooned from eleven billion to nearly one hundred billion dollars in a single year, supplemented by an eighty-five-billion-dollar equity raise. Moody's has warned that these technology giants are becoming asset-heavy businesses whose relationships with capital, risk, and credit are being fundamentally redefined. No single explanation is fully convincing on its own, but taken together these pressures explain the mounting unease over whether the AI-fuelled data centre boom can be sustained.