Hyperscale datacentres are increasing in capacity faster than number as operators push to meet the growing demand for compute capacity required by artificial intelligence (AI) workloads by building increasingly bigger facilities.
Data from market watcher Synergy Research Group confirms there are now 1,136 hyperscale datacentres in operation around the globe, as of the end of 2024, which is double the number that existed five years ago.
In addition, Synergy said its pipeline data shows there are at least a further 504 datacentres in the process of being planned, built and fitted out.
The company’s figures are based on an analysis of the datacentre footprint of 19 of the world’s largest cloud and internet service firms, including the likes of Amazon, Microsoft, Google, Meta, Tencent and ByteDance, to name a few.
Meanwhile, the market watcher’s data shows it has taken less than four years for the total capacity of operational hyperscale datacentres to double in size, as the average size of facilities continues to grow.
“Looking ahead, Synergy forecasts that it will take less than four years for the total hyperscale datacentre capacity to double once again,” said the analyst house, in a research note.
“Each year will see a reasonably steady 130-140 additional hyperscale datacentres coming online, but overall capacity growth will be driven more by the ever-larger scale of those newly opened datacentres [and] generative AI technology is the prime reason for that increased scale.”
More than half (54%) of the world’s hyperscale datacentre capacity is sited in the US, Synergy’s research shows, with Europe and China accounting for 15% and 16%, respectively.
Its data also confirms that Amazon Web Services (AWS), Microsoft and Google have the broadest datacentre footprint of all the providers tracked by Synergy, with each having a sizeable datacentre presence in the US and Europe.
Collectively, these three companies account for 59% of all hyperscale datacentre capacity around the world, followed by Facebook parent Meta, Alibaba, Tencent, Apple and TikTok owner ByteDance.
John Dinsdale, chief analyst at Synergy Research Group, said the demand for artificial intelligence workloads has had a transformative impact on datacentre capacity growth patterns.
“The big difference now is the increased scale of many of those new datacentres,” he said. “Historically, the average size of new datacentres was increasing gradually, but this trend has become supercharged in the last few quarters as companies build out AI-oriented infrastructure.”
While compiling its data, Dinsdale said the company discarded data pertaining to smaller points of presence that some operators “like to claim are cloud datacentres” and also discounted “marketing-orientated claims” that discuss potential builds to get an accurate picture of the state of the market.
“The resulting actual numbers show new datacentres that are a mix of owned versus leased, home country versus international, and large versus super-large, but in aggregate, the trend towards increased size is very clear,” he said.
“It is also very clear that the US will continue to dwarf all other countries and regions as the main home for hyperscale infrastructure.”
