Oracle confident OpenAI can pay $60b yearly for cloud services

Oracle CEO Clay Magouyrk said he is confident OpenAI can pay US$60 billion per year for cloud infrastructure.

He made the remarks at Oracle’s AI World conference in Las Vegas. OpenAI, which reported a US$5 billion net loss in 2024, has a cloud services agreement with Oracle valued at over US$300 billion across five years.

The company said its ChatGPT chatbot now has 800 million weekly active users.

Oracle CEO Mike Sicilia said the company is integrating OpenAI models into a patient portal for electronic health records, following its US$28 billion Cerner acquisition in 2022.

OpenAI uses Nvidia chips rented from Oracle, CoreWeave, Google, and Microsoft, and is developing a custom AI processor with Broadcom.

Oracle shares rose 5% on October 13, bringing its market capitalization close to US$900 billion.

🔗 Source: CNBC

🧠 Food for thought

Implications, context, and why it matters.

OpenAI’s hypothetical $60 billion yearly cloud bill outpaces its revenue

OpenAI generated about $4.3 billion in revenue in the first half of 2025 1.
At an annualized $12 billion in July 2025 2, a $60 billion cloud bill equals roughly five times revenue just to cover infrastructure, before salaries, research, or other operating costs.
The company posted a $5 billion net loss in 2024, and third-party analyses peg its gross margins at roughly 40% 1. At that margin, funding $60 billion in cloud spend needs well over $100 billion in annual revenue before operating costs. A sustainable level could be around $150 billion.

Real estate, infrastructure developers can pursue Oracle and OpenAI’s nearly 7 gigawatt Stargate buildout across Texas and the Midwest

Stargate is a planned network of hyperscale AI data center campuses. The group added sites in Texas, New Mexico, Ohio, and another Midwest location. The OpenAI, Oracle, and SoftBank (a Japanese investment conglomerate) effort targets nearly 7 gigawatts with more than $400 billion of investment over the next three years. The flagship Abilene, Texas site is designed to scale past 1 gigawatt 3.
The buildout will create over 6,000 construction jobs. It also brings about 1,700 long-term roles for real estate developers, construction firms, and infrastructure suppliers in these regions 3.
OpenAI and Broadcom plan to deploy 10 gigawatts of new OpenAI chips. Some compute capacity will not come online until 2026, starting with Nvidia’s next-generation Vera Rubin chips (Nvidia’s forthcoming data center AI processors) 3.