OpenAI's recent aggressive expansion in the artificial intelligence sector has garnered widespread attention. Shortly after announcing a hundred-billion-dollar AI chip co-development agreement with Broadcom (AVGO.US), the company disclosed a five-year, up to $300 billion computing infrastructure procurement contract with Oracle (ORCL.US), further highlighting its unprecedented scale of capital investment and strategic ambition. These series of collaborations, each involving hundreds of billions of dollars, are not only among the largest cloud service contracts in the history of the tech industry but also signify OpenAI's bet on future computing power demands far exceeding its current revenue capabilities.
The partnership with Oracle is part of the company's previously announced "Stargate" data center construction plan. This deal is one of the largest cloud service contracts in history, with a value far exceeding OpenAI's current revenue. According to reports, OpenAI and Oracle will build 4.5 gigawatts of data center computing power, which is roughly equivalent to the power generation of two Hoover Dams or the electricity consumption of about 4 million U.S. households.
Initial signs of this deal emerged in June when Oracle stated in a regulatory filing that it had reached a cloud service agreement expected to generate over $30 billion in revenue by fiscal year 2027. As more data center infrastructure comes online, the revenue generated will increase annually. A month later, OpenAI publicly announced it had agreed to purchase 4.5 gigawatts of computing capacity from Oracle but did not disclose the specific scale of the contract at the time.
OpenAI's collaboration with Oracle may be seen by some as carrying certain risks. OpenAI is currently not profitable and is instead incurring continuous losses. According to its disclosure in June, its annual revenue is approximately $10 billion, which is less than one-fifth of its annual operating costs of $60 billion. These costs include building data centers and leasing computing resources from other data center operators.
Meanwhile, Oracle is betting a significant portion of its future revenue on just one customer. To meet the AI chip demands of its planned data centers, the company may have to take on debt. Compared to Microsoft, Amazon, and Meta, Oracle's debt burden is much heavier relative to its cash holdings. Its investment in AI infrastructure has already exceeded its cash flow, with a total debt-to-equity ratio as high as 427%, while Microsoft's ratio is only 32%.
OpenAI has already ventured into several seemingly high-risk projects. According to reports, OpenAI is also collaborating with Broadcom to develop its own custom AI chips and attempting to create a device that can compete with Apple's iPhone. The company, which has already received billions of dollars in funding, is consuming more resources than any startup in history. Last year, OpenAI's CEO told investors that he expects OpenAI to achieve profitability no earlier than 2029, and reaching this goal may require the company to spend $44 billion.
Behind these series of aggressive moves is OpenAI's long-term bet on ChatGPT's growth momentum: wagering that it can continue to attract billions of users and thousands of businesses worldwide while maintaining a lead in the fierce competition with rivals like Google and Anthropic. However, with growing tensions with its major investor Microsoft and continuously delayed profitability timelines, OpenAI's expansion model, supported by massive debt and binding partnerships, is facing increasingly severe sustainability tests. How it balances strategic investment and financial risks while pursuing technological leadership will become a key variable affecting the landscape of the AI industry.
