KTTC reports on data centers, starting with concise description of different types of data centers…
Cloud Data Centers are the facilities your personal data touches the most often. These data centers are very large facilities commonly run by cloud service providers such as Amazon Web Services (AWS), Google Cloud Platform (GCP), Microsoft (Asure), IBM Cloud, or other public cloud providers. They support massive databases, workloads, and global scalability that social media networks and platforms like Facebook, YouTube, and Netflix rely on.
Enterprise Data Centers are private facilities that companies and corporations either own or rent to handle their own sensitive data or host in-house services. These can be small or large facilities or rent out portions of larger data center facilities.
Colocation Data Centers are what we call facilities that share space with multiple companies. “Colo” data centers handle the infrastructure (the building, cooling, bandwidth, security, etc.) so companies can use the equipment as they need or store their specific equipment.
Hyperscalers are the massive facilities that AI companies primarily use because they are designed specifically to allow for the processing and handling of massive amounts of data and compute. These facilities are the ones that tend to house hundreds of thousands of GPUs that allow services like Chat GPT and Claude to function at the scale they do.
Edge Data Centers are the smallest form of data center and are typically found on location of local users or in a smaller, decentralized facility nearby. There are many cases when companies don’t need a whole data center and instead just need a few servers to handle their data; this is where Edge centers come into play.
The article outlines electricity use…
It’s estimated that the “global datasphere” in 2025 (data created, captured or replicated across global infrastructure) was around 175 zettabytes of data. That’s 175,000,000,000,000 gigabytes that all requires electricity to process.
According to a 2021 study, it’s estimated that U.S. data centers consumed about 1.8% of U.S. electricity. It’s worth noting this study uses data from 2018-2020, so just before the AI boom.
U.S. data centers consumed about 176 TWh of electricity in 2023 (4.4% of U.S. electricity), driven largely by AI workloads and accelerated data-center expansion, according to Lawrence Berkeley National Laboratory.
Including electricity impact on water use…
It’s estimated that about 75% of water impacts occur through electricity generation involved with data centers while the remaining 25% is actually used onsite.