U.S. tech giant Google (GOOGL) is actively considering building a data center in Lea County, New Mexico. This latest AI infrastructure plan reveals the increasingly important site selection logic of tech giants in AI infrastructure expansion: in addition to chip procurement, stable power supply, energy costs, water resources, and community support also determine whether a project can land. Lea County is located in the western Permian Basin, and its abundant natural gas resources and local gas-fired power generation infrastructure may provide energy conditions for large-scale computing facilities. Google is still evaluating the project and conducting community communication, and has not yet disclosed the construction scale, commissioning time, or specific power supply plan.
From the perspective of AI computing power industry layout, as AI inference and agent applications increase and continuously drive computing demand, large cloud computing vendors urgently need to plan chip, data center, and power supply capacity in sync. From the perspective of data center site selection economics, abundant fuel resources and existing gas-fired power generation infrastructure may provide an energy-related attraction for Google’s proposed project, which is also the reason Google has committed to bearing the full cost of its own energy use and the infrastructure required. Natural gas spot prices in the region once fell to negative values due to an oversupply of associated gas, reflecting the area’s abundant natural gas supply. Google management also regards community communication as an important part of project development, stating that it recognizes that many communities are resistant to data center development and therefore intends to first engage in dialogue with local residents.
Financial data disclosed by Google show that in the second quarter, Google Cloud revenue surged 82% year over year to about $24.8 billion; the Gemini model processes 22 billion API tokens per minute, demonstrating the substantial growth in computing power brought by AI applications sweeping the globe. Strong demand related to the AI computing power industry chain has already been significantly reflected in the performance of industry leaders. Data from South Korea’s customs show that from September 1 to 10, semiconductor exports reached $16.5 billion, up 270% year over year; Nvidia’s revenue for the second quarter of fiscal year 2027, ended July 26, reached $96.2 billion, up 106% year over year, of which data center revenue was $89 billion, up 117% year over year. Anthropic has also expanded computing power supply through long-term agreements, including investing more than $100 billion over the next decade in AWS-related technologies to obtain up to 5 gigawatts of new capacity, as well as signing agreements with Google and Broadcom involving several gigawatts of next-generation TPU capacity. These long-term commitments increase the visibility of future chip, data center, and power demand.
Google’s consideration of building a data center in Lea County highlights the importance that AI computing power expansion places on energy supply conditions. The area’s abundant natural gas resources and existing gas-fired power generation infrastructure provide potential energy support for additional data centers. From the perspective of underlying technical logic, models with computer operation and complex task execution capabilities may expand AI demand from single-turn question answering to continuously running agent workflows, and inference demand thereby forms a new driving force for expansion. AI inference infrastructure must simultaneously solve computing, memory, and service latency issues, which means that Google’s economic goal in expanding data centers is to make accelerators, high-bandwidth memory, networking, and power supply systems work in coordination. The potential advantages of energy-resource-rich regions need to be converted into profits through long-term reliable power supply resources, reasonable construction costs, and relatively high utilization rates, which is exactly the investment logic corresponding to the saying that “the end of AI is power.”
