Amid the rapid expansion of artificial intelligence technologies, discussions around data center power requirements have intensified. A China-based company, DeepSeek, has drawn attention with its cost-effective AI models, suggesting that advanced technology might decrease energy demands. However, industry experts argue that such innovations could paradoxically lead to increased usage. As tech giants expand their data center footprints and invest heavily in AI infrastructure, the construction of power facilities is accelerating to meet rising needs. Despite challenges like skilled labor shortages and sustainability goals, the urgency to address these energy demands has reached federal levels, prompting executive actions and explorations into alternative energy sources.

While some speculate that enhanced AI efficiency may lower power consumption, others believe it will stimulate further growth. John Medina from Moody’s Ratings highlights how reduced computing costs can inspire new applications and companies. Similarly, Mitchell Osborne of Adolfson & Peterson Construction asserts that despite more efficient models, we are still in the early stages of AI and quantum computing. Consequently, data center and power infrastructure projects continue to surge, driven by tech giants like Amazon, Microsoft, and Meta. Their expansions coincide with a joint venture among OpenAI, Softbank, and Oracle investing significantly in AI infrastructure, which could reach up to $500 billion.

This escalation in demand is mirrored by an increase in power construction. Ryan Wobbrock from Moody’s Ratings notes that utilities and power producers are ramping up expenditures to accommodate data centers' energy needs. New projects and renovations are driving local grid work, as evidenced by specific substation constructions for data centers and stability upgrades subsidized by these corporations. Yet, older power plants, once marked for retirement, are being refurbished instead of decommissioned due to the inability of new projects to match the pace of data center growth in key regions.

Data center energy consumption has skyrocketed over the years, reaching 176 terawatt-hours in 2023, approximately 4.4% of total U.S. electricity use. Projections estimate this figure to rise between 325 and 580 terawatt-hours by 2028. Theodore Paradise from CTC Global emphasizes that the current U.S. power grid was not designed to handle such rapid demand growth. Interconnection costs have escalated dramatically, sometimes reaching over a billion dollars. Despite these challenges, data center construction remains robust, with some developers proceeding without secured power contracts.

Tech giants are increasingly taking matters into their own hands by constructing their own power facilities adjacent to data centers. According to Medina, this trend involves building large data centers alongside power plants, creating self-sufficient islands powered by gas, solar, and battery systems. However, skilled labor shortages pose significant challenges to construction activities. Byron Sarhangian from Snell & Wilmer warns that strict environmental requirements could hinder U.S. progress in AI development compared to other nations, emphasizing the need for all available energy infrastructure.

Federal intervention has become necessary to address these escalating challenges. The recent executive order declaring an energy emergency aims to expedite new power generation and transmission construction. Interest in nuclear energy as a solution is growing, particularly among hyperscalers like Microsoft experimenting with small modular reactors. State and federal regulators may also influence construction expansion through regulatory changes. As power consumption increases, potential prohibitions or encouragements regarding data center expansion and AI usage could impact the speed at which these projects commence.