Hyperscale Data Centers to Triple Capacity in 6 Years, Fueled by AI

AWS booth

The average capacity of hyperscale data centers that will launch over the next six years will shortly more than double that of hyperscale data centers that are now in operation, according to new statistics and projections from Synergy Research Group. The need for much more powerful facilities has increased due to the influence of generative AI technologies and services, even if the key IT load of hyperscale data centers has always tended to expand with time.

The number of operational hyperscale data centers will continue to rise rapidly as the average IT load of individual data centers increases. To increase their capacity, some retrofitting of already-existing data centers will also be done. As a consequence, during the next six years, the combined capacity of all active hyperscale data centers is expected to almost triple.

The hyperscale study was conducted based on an examination of the operations and data center footprint of 19 of the largest cloud and Internet Service Providers (ISPs) globally that satisfy Synergy’s requirements to be classified as hyperscale operators. The biggest players in search, social networking, ecommerce, SaaS, IaaS, and PaaS are included in this. These businesses operated 926 significant data centers globally by the middle of 2023. One of the primary inputs into Synergy‘s projection model is the additional 427 facilities that are part of Synergy’s pipeline of future data centers.

– story continues below he graphic –


Generative AI

Over the last five years, the overall number of data centers globally has doubled, however the composition of hyperscale data centers varies by location and whether they are owned or leased.

Recent developments in generative AI have not so much increased the number of data centers – which will still expand by more than a hundred each year – as they have significantly increased the electricity needed to power those data centers. The power density of related racks and data center facilities must also rise significantly as the number of GPUs in hyperscale data centers soars, mostly due to AI. Hyperscale operators are having to reconsider parts of their deployment strategies and data center architecture as a result, according to Synergy Research Group.