This week, OpenAI and Oracle surprised the markets with a new five-year agreement valued at three hundred billion dollars. This deal is part of a surge of new business that sent the cloud provider’s stock skyrocketing. However, the markets should not have been so surprised. The agreement is a clear reminder that, despite its legacy status, Oracle still plays a major role in AI infrastructure.
For OpenAI, the agreement is more revealing than the lack of details might suggest. The startup’s willingness to pay such a significant amount for compute power provides a strong measurement of its ambitions. It remains unclear where the electricity to power this compute will come from or how OpenAI will pay for it.
An industry analyst noted that the deal makes sense for both sides. For OpenAI, working with multiple infrastructure providers diversifies its risk and gives it a scaling advantage compared to competitors. This strategy appears to be building one of the most comprehensive global AI supercomputing foundations for extreme scale, which is quite unique and exemplary of what a model ecosystem should look like.
Some industry watchers expressed surprise that Oracle was involved, citing the company’s diminished role in the AI boom compared to cloud rivals like Google, Microsoft Azure, and AWS. However, observers should not be so surprised. Oracle has a history of working with hyperscalers and provides the infrastructure for sizable operations, such as TikTok’s U.S. business. Over decades, the company has built core infrastructure capabilities that enable it to deliver extreme scale and performance.
Even as the stock market celebrates the deal, key details are missing and questions around power and payment remain. OpenAI has made a string of major infrastructure investment announcements over the past year, each with an eye-popping price tag. The company has committed to spend around sixty billion dollars a year for compute from Oracle and another ten billion to develop custom AI chips with Broadcom.
OpenAI also recently reported hitting ten billion dollars in annual recurring revenue, up from five and a half billion last year. This figure includes revenue from consumer products, business products, and its API. Despite this growth and a rosy picture of future prospects, the company is burning through billions of dollars in cash each year.
Power is another critical question. Specifically, where will the companies source the enormous energy required to run this level of compute? Industry observers have predicted a near-term boost for natural gas, though solar and batteries are better positioned to deliver power sooner and at lower cost in many markets. Tech companies are also making significant bets on nuclear energy.
The energy impact of OpenAI’s anticipated growth is not entirely unexpected. Data centers are projected to consume fourteen percent of all electricity in the U.S. by 2040. Compute has always been a constraint for AI companies, leading investors to buy thousands of Nvidia chips to ensure their startups have the necessary power.
However, compute is worthless without electricity. To keep their data centers powered, large tech companies have been acquiring solar farms, purchasing nuclear power plants, and making deals with geothermal startups. So far, OpenAI has been relatively quiet on this front. While its CEO has placed personal bets in the energy sector through companies like Oklo, Helion, and Exowatt, OpenAI itself has not invested directly in power infrastructure like Google, Meta, or Amazon.
With a four and a half gigawatt compute deal, that may soon change. The company may take an indirect role, paying Oracle to handle the physical infrastructure. This would leave OpenAI asset light, a strategy that would please its investors and help keep its valuation in line with other software-centric AI startups rather than legacy tech firms burdened with expensive infrastructure.