Computing power and the scale of data centers are growing rapidly, driven by the rise of generative artificial intelligence. Construction projects for these centers are underway in France and other countries. Data and computing centers are essential parts of the digital world, but their profitability is being questioned, and some face challenges in securing enough electricity. This raises basic but important questions: what exactly are data and computing centers, and what do they rely on? Data centers are the physical backbone of the digital world and artificial intelligence. They enable secure storage of financial transactions and customer data, the streaming of movies and series, online gaming with minimal response delays, real-time social media interactions, archiving of photos and videos, digitized government services, and more. These centers are not just about storing data—they also handle massive processing tasks, especially with the rise of AI and high-performance computing needs. Strictly speaking, these critical infrastructures should be called "data and computing centers" (CDC), or in English, Compute and Data Center. These are real, physical installations that contrast with the often intangible perception of the digital world, which is partly shaped by terms like "cloud computing" and "fog computing." A CDC is a complex industrial facility that includes machine rooms with servers, data storage units, cooling systems to maintain optimal temperatures, electrical infrastructure to ensure continuous operation, and areas dedicated to information systems, cybersecurity, and logistics. Before installing a CDC, it is essential to understand the specific needs of the location, as the type of equipment, computing methods, and cooling systems will impact the resources required, such as electricity and water, and affect the surrounding communities. The evolution of computing power has been remarkable. In 1946, the first electronic computer, ENIAC, consumed 150 kilowatts of power and performed 5,000 operations per second. Today, supercomputers can perform over a billion billion operations per second using less than 20 megawatts of power. Modern data and computing centers use various models to meet different needs, including internal centers for exclusive use, outsourced centers, cloud-based centers, and hyperscale centers for handling massive data volumes. The use of GPU accelerators has become crucial for generative AI, as they handle complex matrix calculations more efficiently than traditional CPUs. However, this increased computing power requires significant energy, which poses challenges in terms of sustainability, security, and resource management. As the demand for AI infrastructure grows, the economic and physical limits of this expansion are becoming more apparent.