The Trump administration recently launched a new policy for AI data center power supply. In fact, prior to the implementation of this policy, several leading AI companies in the United States had already taken the lead in expanding into the power sector through various means such as acquisitions and technology research and development. However, their main actions in the energy sector were mostly focused on "buying electricity" and optimizing electricity, rather than independently "producing electricity".

On December 23, 2025, Alphabet, the parent company of Google, announced the full acquisition of Intersect Power LLC, a data center and clean energy solutions provider, for $4.75 billion in cash and debt. It is reported that this is the largest energy acquisition in Google's history. Its core goal is to break through the power supply bottleneck caused by the surge in AI computing power, and to build an exclusive power supply system for Google's global AI data centers through the "energy computing power" vertical integration model, in order to cope with the increasingly fierce competition in the AI industry.

Unlike Google's acquisition and integration model, in February 2026, Nvidia announced its new layout in the power sector, clearly stating the concept of "AI computing power expansion will crush the power grid in the short term, but will drive down energy costs in the long term". In response to this concept, Nvidia has launched a dedicated AI solution for the power industry, focusing on optimizing the operation and maintenance of nuclear power plants, helping power companies reduce energy consumption and improve power supply efficiency; At the same time, the "Open Power Artificial Intelligence Alliance" led by it is collaborating with multiple industry organizations to develop open-source AI models, focusing on industry pain points such as grid dispatching and load forecasting. Currently, relevant technologies have been piloted and applied in some US power companies.

Microsoft has chosen to collaborate with energy companies to optimize power supply. In December 2025, Microsoft and Duke Energy of the United States reached a cooperation agreement to assist Duke Energy in optimizing energy storage resource allocation and grid load forecasting, jointly developing exclusive AI agent tools, and ultimately achieving early warning of power facility failures, significantly improving power operation and maintenance efficiency, indirectly providing guarantee for the stable power supply of its own AI data center.

However, these diversified energy strategies adopted by top AI companies in the United States, while capable of effectively bridging the power gap for AI data centers in the short term, struggle to fundamentally meet the exponentially growing long-term demand for AI computing power. Industry analysis even suggests that the "self-built power generation" model, where AI companies construct their own power plants, may not entirely resolve this predicament. Achieving a sustainable long-term balance between computing power expansion and energy supply remains a significant challenge.