As AI and related technology companies become increasingly ubiquitous in the United States and beyond, data centers have started popping up everywhere, sometimes against the wishes of local businesses and residents. It’s specifically data centers serving AI technology that have sparked anger and controversy, but data centers were in business long before the generative AI boom. It is through data centers that most of the basic functions of the wireless ecosystem are carried out, from file sharing to communication across various applications.
There are data centers all over the world, with some of the largest providing services for exchanging wireless network connections and storing data in the cloud. These centers operate using a combination of physical data storage servers and access to cloud storage to circulate data within their respective countries and communities. Data centers are not problematic on their own, especially those powered entirely by renewable sources, but new data centers specifically aimed at serving the AI industry, increasing utility costs and inconveniencing local residents, are what are generating the majority of the anger.
A data center stores and circulates data physically and virtually
panumas nikhomkhai/Shutterstock
A data center is, as its name suggests, a data center. In other words, it is a central hub through which large amounts of data can be stored and distributed. In the same way that your personal computer needs internal storage and memory to store and run programs, data centers also use similar types of components. The difference is that these facilities handle data streams from entire counties, regions, states or more, and therefore need a massive amount of physical servers, storage drives and network infrastructure.
Modern data centers use a combination of physical and virtual storage and computing methods, both to save space and to better connect with other data centers. The combined data needs of a single service can be networked across multiple different data centers via cloud computing, which also helps protect the functions of that service in the event that a data center is damaged or compromised. In addition to IT components, data centers are typically also equipped with various supporting hardware elements, including power subsystems, backup generators, and ventilation and cooling equipment. Many servers running at full capacity get very hot, although many data centers offset cooling and power costs by locating in cooler locations like deserts or high altitude areas. China has even gone so far as to build a fully submerged data center.
Data centers can be used for practical purposes (but not always)
VisualAviation/Shutterstock
According to current standards, data centers are divided into two broad categories: traditional data centers and AI data centers. It’s largely thanks to traditional data centers that the seemingly seamless Internet experience you have today is facilitated; they manage web hosting services that keep websites running, store data for cloud services, run virtual applications that external users can access without local installation, and manage important communications services like email and digital calling.
One of the largest data centers in the world, the China Telecom Information Park located in Hohhot, Inner Mongolia, is responsible for all of these types of services, as well as government systems and business services. There is a similarly sized data center located in Tahoe Reno, Nevada, USA, owned by Switch, which provides a myriad of telecommunications services to government, businesses and individual technology customers. While traditional data centers serve these many practical purposes, it is specifically AI data centers that have come under public scrutiny.
Much like traditional centers, AI data centers contain large quantities of physical servers, although they are also loaded with GPUs, TPUs, NVME storage drives, and other high-end components, all for the exclusive purpose of running and training AI models, as well as performing adjacent tasks such as data analysis and chatbot management. Because these particular centers are full of sophisticated equipment, they have higher power and cooling requirements, which is why they have been controversial for monopolizing local resources. Amazon’s AI data centers alone used more than 2.6 billion gallons of water for cooling in a single year.
