Southern Company Unit Signs 25-Year Power Supply Agreement with OpenAI

A Southern Company subsidiary has signed a 25-year power purchase agreement with OpenAI, in the latest of a series of major, long-term energy commitments driven by the boom in artificial intelligence infrastructure.

The deal illustrates the massive electricity demands of the artificial intelligence technologies, which require electricity-hungry data centers to operate around the clock while training and deploying models. As the pace of artificial intelligence innovation quickens, tech companies are increasingly lining up long-term power supply agreements for the energy capacity that will support their future compute needs.

The pact is one of the latest signals of an increasing convergence of the energy and technology sectors.

Artificial intelligence companies are investing billions in constructing new data centers packed with sophisticated graphics processors, dedicated artificial intelligence chips, and high-speed networking gear. These facilities demand enormous quantities of electricity to power the silicon, as well as robust cooling systems needed to keep the equipment from overheating.

Southern Company, one of the largest electric utilities in the United States, supplies power through its subsidiaries to millions of customers across the Southeast. It has been expanding its infrastructure to accommodate a rise in industrial and commercial demand. For utilities, long-term contracts with large technology companies provide predictable revenue, while helping justify new generation and transmission investments.

For OpenAI, a 25-year contract helps it achieve certainty around power for an operation that has experienced rapid growth as generative artificial intelligence applications have gained widespread traction. The explosion in demand has driven needs for cloud computing, massive training clusters, and inference infrastructure to serve millions of users globally.

Industry experts say that the growth in artificial intelligence has been fundamentally reshaping electricity markets. Utilities throughout North America have reported increased calls from data center developers to be connected to the grid, especially those with facilities dedicated to AI. In some regions, energy demand growth has accelerated to levels not seen in decades, following years of tepid increases.

Power is an indispensable element of the artificial intelligence equation. Training large language models and other artificial intelligence technologies requires thousands of advanced processors to run for weeks at a time, consuming large amounts of energy. Long-term agreements are part of a strategy for managing supply risk and for the large investments in new computing technologies to be funded.

The deal also illustrates broader trends in the corporate energy market. Numerous large technology companies have signed multi-decade contracts to acquire electricity, supporting development of new power-generating facilities that in some cases might blend new renewable resources and existing generation.

Utilities are responding to evolving market demands. Powering facilities that are packed with the equipment to run artificial intelligence programs requires significantly increasing generation, transmission and other aspects of electricity delivery. A customer with a long-term commitment can support funding for such infrastructure projects.

The collaboration highlights the deepening entanglement of artificial intelligence with the energy system. As AI finds its way into almost every facet of life - business, government, consumer use - electricity has become as crucial to tech companies as the next generation of high-speed silicon and networking hardware.

The 25-year pact between the Southern Company subsidiary and OpenAI demonstrates how planning for electricity availability has become a critical element in scaling artificial intelligence. With data center growth anticipated to continue, arrangements like these are expected to become increasingly central to meeting the power demands of tomorrow’s artificial intelligence applications.