标普将甲骨文的信用评级下调至BBB,距离垃圾级仅差一级。
S&P downgrades Oracle to BBB – only one notch above junk level

原始链接: https://www.heise.de/en/news/S-P-downgrades-Oracle-to-BBB-only-one-notch-above-junk-level-11363472.html

标普全球已将甲骨文(Oracle)的信用评级下调至BBB-,仅比“垃圾级”高出一档。此次下调源于该公司为拓展人工智能基础设施而进行的激进举债扩张,其2027财年的资本支出预测已激增至950亿美元。预计这将导致420亿美元的自由现金流赤字。 标普主要担忧甲骨文对OpenAI的高度依赖,后者的合同额约占甲骨文6380亿美元未来合同收入的一半。标普将其列为“核心信用风险”,因为一旦OpenAI面临财务动荡,甲骨文的长期数据中心承诺可能会变得难以为继。 随着甲骨文从软件供应商向超大规模云服务商转型,它正面临激烈的竞争,且财务灵活性不如微软或谷歌等大型竞争对手。尽管该公司裁减了13%的员工以抵消成本,但仍承受着巨大压力。这种情况反映了国际清算银行所提出的更广泛的行业担忧;该行曾警告称,大量债务融资的人工智能投资可能会引发类似于过去金融危机的系统性风险。尽管标普维持“稳定”的展望,但甲骨文仍处于非投资级地位的边缘。

甲骨文(Oracle)的信用评级被下调至 BBB-(仅比垃圾级高出一档),这引发了 Hacker News 上关于该公司未来及整个 AI 行业的一场重要讨论。 部分用户推测,如果甲骨文陷入财务困境,可能会面临政府救助或出售业务部门的情况;而另一部分用户则关注 AI 基础设施不断变化的格局。这场辩论的核心议题在于 AI 公司是否拥有“护城河”。评论者认为,由于用户更看重成本和性能,服务间的切换非常容易,这削弱了长期的用户忠诚度。 然而,也有用户指出,大型云服务提供商(微软、亚马逊和谷歌)通过将 AI 嵌入现有生态系统、利用企业锁定效应,以及规避数据迁移带来的“出口税”,维持着竞争优势。一些参与者对当前的 AI 热潮持怀疑态度,理由是担心数据中心可能存在产能过剩,以及硬件投资具有周期性。这场讨论最终反映了人们对企业债务水平以及当前 AI 基础设施淘金热可持续性的日益焦虑。
相关文章

原文

Rating agency S&P Global has lowered Oracle's creditworthiness from BBB to BBB- – this is the lowest notch in the so-called investment-grade area. A further downgrade would push the database company into speculative territory. However, the outlook remains stable according to S&P.

The rating agency attributes the downgrade, published on July 9, to Oracle's rapidly growing AI infrastructure business, which is massively increasing the company's debt and capital requirements. S&P had already set the outlook for Oracle to “negative” in July 2025, warning of precisely this scenario.

According to S&P, the core of the problem is Oracle's enormous investments in expanding AI data centers. S&P forecasts a deficit in free operating cash flow of almost 42 billion US dollars for the 2027 fiscal year. The rating agency expects Oracle to finance this deficit with a mix of debt and equity.

For the 2027 fiscal year, which ends in May next year, Oracle had raised its spending forecast to 90 to 95 billion US dollars – S&P had previously only assumed 60 billion. The analysts suspect rising component costs, such as for GPUs and network equipment, as the reason.

S&P views Oracle's strong dependence on a single major customer, OpenAI, as particularly critical. According to analyst estimates, about half of the contractually promised but not yet delivered service volume of 638 billion US dollars is attributable to OpenAI. S&P therefore explicitly describes OpenAI as a “central credit risk”.

Because if OpenAI were unable to meet its payment obligations, Oracle would be left with long-term data center rental agreements. These could neither be easily terminated nor transferred to other customers on comparable terms. And OpenAI's ability to service its contracts, according to S&P, depends on the AI boom continuing, the models remaining market-leading, and the company continuing to raise external capital -- which is not considered certain.

Oracle is currently undergoing a transformation towards a larger cloud infrastructure business. This accounted for about 27 percent of total revenue in fiscal year 2026. S&P expects this share to rise to nearly 60 percent by 2028. However, compared to other hyperscalers like Microsoft, Google, or Amazon, S&P sees Oracle in a weaker position: the company is more dependent on external customers and has less financial flexibility to weather an industry downturn. Furthermore, new competition is emerging – for example, from SpaceX, which rents computing capacity to Anthropic and Alphabet.

In parallel with the AI expansion, Oracle has cut over 21,000 jobs in the past twelve months – about 13 percent of the workforce. With this shift “from people to machines,” the company aims to finance AI infrastructure.

Oracle's situation fits a trend that international financial regulators are also warning about. The Bank for International Settlements (BIS) sees parallels between debt-financed AI investments, the dot-com bubble, and the financial crisis, and sees a “danger like in 2008”. The BIS warns of a system crash due to Nvidia & OpenAI debt.

(rie)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.

联系我们 contact @ memedata.com