New data and forecasts provided by Synergy Research Group indicate a nearly threefold increase in the average capacity of hyperscale data centers within the next six years.
While it's been a longstanding trend for the critical IT load of hyperscale data centers to expand over time, the influence of generative AI technology and services has added an extra layer of urgency in the demand for significantly more potent data facilities.
Concurrently, as the typical IT load of individual data centers continues to rise, the count of operational hyperscale data centers will continue its gradual ascent. Moreover, there will be a degree of retrofitting existing data centers to enhance their capacity.
This hyperscale research stems from an analysis of the data center operations and footprint of 19 prominent cloud and internet service companies globally, meeting Synergy's criteria for classification as hyperscale operators. These encompass major operators in SaaS, IaaS, PaaS, search, social networking, e-commerce and gaming. As of mid-2023, these companies collectively operated 926 significant data centers worldwide.
The composition of hyperscale data centers is undergoing shifts, varying by region and ownership status (owned vs. leased). However, when considered collectively, the total count of data centers worldwide has doubled over the past five years.
The impact of recent generative AI advancements doesn't so much pertain to an increase in the number of data centers but rather significantly raises the power requirements for operating these data centers. With the proliferation of GPUs in hyperscale data centers, primarily due to AI-driven applications, the power density of the racks and data center facilities must also undergo substantial augmentation. This shift is prompting hyperscale operators to reevaluate aspects of their data center architecture and deployment strategies.