Table of Contents
New data from Synergy Research Group shows hyperscale companies accounted for 48% of worldwide data center capacity at the end of 2025
In sum – what to know:
Hyperscale dominance – Hyperscalers will control 67% of global capacity by 2031, driven by large-scale AI infrastructure expansion.
AI-driven scale – New data centers are significantly larger, with AI accelerating capacity growth and reshaping infrastructure design.
Enterprise shift – On-premise declines continue despite modest AI-driven growth, while colocation gains relevance as a hybrid infrastructure layer.
Hyperscale operators are set to control the majority of global data center capacity within the next decade, as artificial intelligence workloads accelerate a long-term shift away from enterprise-owned infrastructure.
New data from Synergy Research Group shows hyperscale companies accounted for 48% of worldwide data center capacity at the end of 2025, with that share projected to rise to 67% by 2031. Over the same period, enterprise on-premise capacity is expected to decline from 32% to 19%, continuing a multi-year decline from 56% in 2018.
The shift reflects both the scale of hyperscale investment and the changing nature of compute demand. Total data center capacity is still growing rapidly, but nearly all of that expansion is being driven by hyperscale deployments, which are forecast to more than triple by 2031.
A key factor behind this acceleration is AI. While traditional cloud workloads continue to grow, the rise of generative AI and GPU-intensive computing is reshaping the scale and design of new facilities.
In an interview with RCR Wireless News, John Dinsdale, chief analyst at Synergy Research Group, said: “If you look at new hyperscale data centers that have come or are coming online in the 2025-27 period, the average size of those new data centers is about twice the size of the average hyperscale data center that was live at the end of 2024. Data center size was trending upwards, but the increase was rather gentle until AI supercharged the trend. Over the last three years we have had cause to increase our five-year forecasts of hyperscale data center capacity due to the speed of AI adoption.”
This surge in capacity is also changing the balance between owned and leased infrastructure. Nearly 60% of hyperscale capacity is now in owned facilities, reflecting a strategic shift toward greater control over high-performance environments required for AI workloads.
At the same time, AI is creating a more nuanced impact on enterprise infrastructure. While on-premise capacity is seeing some renewed growth—largely tied to GPU deployments and specific AI use cases—this is not expected to reverse the broader trend toward centralization.
Dinsdale added: “I think we’ll continue to see some growth in on-premise capacity over the next five years, but the growth will be rather muted while hyperscale capacity growth will be shooting upwards. Ultimately it makes more sense to push the majority of workloads onto public clouds, or at least locate most enterprise-owned computing gear in colocation facilities. Running your own in-house data centers does not make much sense for most organizations, so that seachange will continue to play our over the coming years.”
Colocation providers remain an important part of this evolving ecosystem. Although their share of total capacity is expected to decline slightly, absolute capacity continues to grow at near double-digit rates, supported by both hyperscale demand and enterprise offloading strategies.
“There is a profitable and sustainable place in the world for colocation companies. An ever-increasing share of their revenues is coming from hyperscale clients, and that is a good thing. Hyperscale operators will continue to need colocation partners, both for large-scale wholesale leasing and for smaller edge deployments. Meanwhile, in the non-hyperscale world, enterprises will continue to push more of their IT hardware offsite and into colocation facilities,” Dinsdale said.
Neocloud providers are scaling rapidly as demand for AI infrastructure accelerates, with recent data from Synergy Research Group pointing to a market on track to approach $400 billion by 2031.
According to the research firm, neocloud revenues reached $9 billion in the fourth quarter of 2025, marking a 223% year-on-year increase, and surpassed $25 billion for the full year. Growth is being driven primarily by demand for GPU-accelerated compute, which continues to outpace the capacity available from traditional hyperscale cloud providers.
