Hyperscale datacenters are getting bigger, a direct response to the skyrocketing demand for generative artificial intelligence (GenAI) services. Synergy Research Group shares that the average size of these datacenters is set to double in just four years.
Historically, the IT load in hyperscale datacenters has grown over time. However, GenAI’s hungry appetite for power has accelerated this trend. With each datacenter handling more load, we’ll see the number of operational hyperscale datacenters steadily rise.
Currently, there are 1,103 hyperscale datacenters worldwide, and another 497 are on the way in the next four years. When you merge the increasing number of datacenters with their larger sizes, we can expect the total hyperscale datacenter capacity to nearly triple by the end of 2030.
Synergy’s insights come from evaluating the datacenter operations of 19 major cloud and internet service companies. John Dinsdale, the chief analyst at Synergy, notes that the number of operational hyperscale datacenters has doubled in the past five years. While this trend will continue, the real shift is in the growing capacity of newly launched datacenters.
The situation is complex, with various factors at play—old versus new facilities, regional considerations, and whether they’re owned or leased. Still, one thing is clear: GPU-oriented infrastructure will push new hyperscale datacenters towards double the capacity.
In recent days, Synergy has underscored how GenAI’s rising demand is transforming the datacenter landscape. Their latest findings reveal that this demand has driven datacenter hardware and software spending to record levels in 2024. Operators invested heavily last year to establish AI-ready server farms.
Dinsdale points out that while public cloud growth has spurred datacenter investments for over a decade, no one expected the 2024 market for datacenter equipment to exceed $280 billion.