Data centers, those often-maligned, architecturally unremarkable concrete boxes surrounded by power containers, diesel generators, and air-cooled chillers, are evolving rapidly. Some say the changes will be permanent and irreversible. Whether this is true or not, one thing is certain. These buildings that many people feel are ugly, are likely to undergo a shift and become even uglier. I say this as a true lover of mechanical and electrical systems, after spending the last 25 years immersed in the data center business.
Data centers have always been one of the clearest examples of 20th century architecture’s guiding principle: form follows function. The “form,” since the emergence of colocation and the cloud, has been football-field-sized rooms filled with server, storage, and network cabinets. Today, however, billions of dollars in capital are flowing into our industry to fund gigawatts of artificial intelligence (AI) infrastructure, which requires a different infrastructure that is also far denser and more complex. The data centers of tomorrow may look less like buildings and more like industrial plants.
ChatGPT served as a major catalyst for this transformation, but OpenAI was merely the first of many creators to develop headline-worthy AI applications. Though AI already existed prior to ChatGPT 3.5, the marvel of AI flipped the switch on a floodlight that is spurring innovation and creativity around the world, igniting an unprecedented surge in interest and investment throughout the IT industry.
However, this surge has brought a new set of challenges, particularly for the data center industry. The swift rise of AI, fueled by advancements in high-performance computing (HPC) and the use of Graphics Processing Units (GPUs), has dramatically altered the landscape of data center design, deployment, and operations.
