许多早期科技巨头淡出舞台，但甲骨文（Oracle）81 岁共同创办人赖瑞・艾里森（Larry Ellison）成为 AI 热潮核心人物。Oracle 几乎把未来押在 AI 上，为 OpenAI 等客户以前所未见的规模扩建资料中心，预期将使现金流在未来数年转为负值。公司文件显示，截至 11 月的三个月内签下约 1,500 亿美元租赁（多为资料中心）；若计划顺利，营收可在数年内「超过翻倍」。
华尔街最初被云端业务订单暴增所吸引，股价在 9 月大涨，艾里森甚至一度在彭博富豪榜超越好友马斯克。然而疑虑随后升温：债务攀升、如此规模的资料中心是否真能高获利，以及最大客户 OpenAI 同时对亚马逊与 CoreWeave 等供应商做出巨额承诺，让外界担心款项是否能全数到位。Oracle 股价自 9 月 10 日历史高点以来已下跌 44%，信用违约交换（CDS）价格也走高，反映市场对偿债能力与财务前景的担忧。
在剧烈波动中，艾里森仍牢牢掌控决策：他提高了直接控制的持股比例，Oracle 也加速以两位联合执行长取代长期搭档执行长 Safra Catz 的安排；自 11 月起，最高财务主管改为直接向他汇报、绕过两位 CEO。尽管如此，Oracle 表示已累积 5,230 亿美元销售待履约（backlog），分析师认为合约量足以安抚投资人，但市场对履约「能力或机率」已受损；公司将在 2026 年证明能否把承诺化为实际交付与现金回收。
Many original tech titans have faded, but Oracle co-founder Larry Ellison, 81, has become a central figure in the AI boom. Oracle is betting its future on AI infrastructure, launching an unusually large data-center build-out for customers including OpenAI that is expected to push cash flow negative for years. In the three months ending in November, Oracle signed about $150 billion in leases, mostly tied to data centers; management says revenue could more than double within a few years if plans hold.
Wall Street initially embraced the surge in cloud bookings that helped drive Oracle’s stock sharply higher in September, briefly lifting Ellison past Elon Musk on Bloomberg’s wealth rankings. Since then, investor anxiety has grown around a rising debt load, whether data centers at this scale can sustain high profitability, and whether customers can meet all commitments—especially as OpenAI has also made large deals with other vendors such as Amazon and CoreWeave. Oracle’s stock is down 44% from its Sept. 10 record high, and its credit-default swap prices have been rising, signaling greater perceived credit risk.
Ellison is tightening control amid the turbulence: he has increased the stake he directly controls, Oracle accelerated a shift to two co-CEOs replacing long-time partner CEO Safra Catz, and as of November the top finance executive reports directly to Ellison, bypassing both CEOs. Oracle says it already has enough contracted business to fuel an AI-era reinvention, disclosing $523 billion in backlog. The key test shifts to execution and delivery, with 2026 framed as the year Oracle must prove it can convert those promises into results and cash.