Hyperscale datacentre capacity doubling every four years – report

• There are now more than 1,000 large hyperscale datacentres in operation worldwide, according to Synergy Research Group
• The total capacity of hyperscale datacentres has doubled in the past four years
• And it’s due to double again in the next four years as more capacity is needed to support generative AI (GenAI) workloads

The total capacity on offer from the world’s hyperscale datacentres has doubled over the past four years and is set to do so again over the next four years, primarily to support growing demand related to generative AI (GenAI) workloads, according to Synergy Research Group.

New data collated by the research firm shows that the number of large hyperscale datacentres in operation worldwide hit 992 at the end of 2023 and passed the 1,000 mark early this year. Not only that but the capacity of those datacentres has been increasing, and there’s more to come.

“Despite the super-massive size of the datacentre network, its capacity is doubling every four years,” according to the Synergy Research Group team, which expects that that growth rate will continue throughout the rest of the decade. “While there are many different deployment scenarios at play, in aggregate the average capacity of hyperscale datacentres continues to grow,” explained John Dinsdale, chief analyst at Synergy Research Group.

The company’s forecast further suggests that in the next five years, there will be around 120 to 130 additional hyperscale datacentres coming online per year, while capacity growth is expected to be driven increasingly by the even larger scale of newly opened datacentres. This will mainly be due to generative AI [GenAI] technology, according to Synergy’s findings. The company bases its forecast on the future datacentre plans of the hyperscale operators – currently, 440 new facilities are at various stages of being planned, developed or fitted out.

As the graph above shows, the US accounted for more than half (51%) of the world’s total hyperscale datacentre capacity – measured by megawatt (MW) of critical IT load – at the end of last year. Europe and China followed, with each of them representing around a third of the balance (17% and 16%, respectively).

Within the US, the “state of Virginia alone accounts for a third of US capacity,” noted Dinsdale in comments shared by email with journalists. “Outside of the US, the next biggest countries ranked by megawatt of critical IT load are China, Ireland, Germany, Japan, Netherlands and Australia,” he added. 

Amazon, Microsoft and Google (the cloud services giants) are the companies with the broadest datacentre footprints globally. Each of them not only has a “huge” number of facilities in their home market of the US but also multiple datacentres in many other countries. “In aggregate, the three now account for 60% of all hyperscale datacentre capacity,” according to the research firm.

That trio were followed by Meta, Alibaba, Tencent, Apple, ByteDance and other relatively smaller hyperscale operators.

“Generally speaking, self-owned datacentres are much bigger than leased datacentres and datacentres in the home country of a hyperscale company are much bigger than its international facilities, though there are plenty of exceptions to these trends. We’re also seeing something of a bifurcation in datacentre scale: While the core datacentres are getting ever bigger, there is also an increasing number of relatively smaller datacentres being deployed in order to push infrastructure nearer to customers. Putting it all together though, all major growth trend lines are heading sharply up and to the right,” explained Dinsdale.

In a previous forecast, the research firm forecast that the installed base of hyperscale datacentres would hit 1,200 in 2026 – see How many hyperscale data centres does the world need? Hundreds more, it seems.

- Yanitsa Boyadzhieva, Deputy Editor, TelecomTV