Increasing community opposition to data center development is compelling hyperscalers to rethink their decades-old practices.
Enterprises deploying AI inference need to choose carefully to limit costs and protect their data.
Concerns over rising electricity costs and environmental impact are driving local opposition to new data center projects in the US, prompting a growing number of cancellations.
Elon Musk's merger of his SpaceX aerospace company with his AI firm xAI has relit the thrusters under the concept of building big data centers in space. However, the technical difficulties involved may ultimately thwart his ambitions.
Simple arithmetic shows that newly constructed, large-scale private data centers with high occupancy rates can sometimes undercut colos on unit costs, but in most cases colos remain significantly cheaper.
Nvidia's DGX-Ready certification for colocation facilities has been around for nearly six years, yet what the program actually entails remains obscure.
Giant data centers are being planned and built across the world to support AI, with successful projects forming the backbone of a huge expansion in capacity. But many are also uncertain, indicating risks and persistent headwinds.
Uptime Institute's 2025 Service Providers and Capacity Survey (n=872) benchmarks the industry in the areas of public cloud, capacity in owned data centers, and capacity in colocation facilities.The attached data files below provide full results of…
The US Cloud Act lets US authorities access cloud data stored overseas. Encryption offers protection only when keys are firmly controlled - creating challenges for enterprises but an opportunity for colocation providers.
The remote nature of cloud computing makes it vulnerable to extraterritorial legal reach. Colocations, by contrast, only manage infrastructure - not data - shielding them from the same level of foreign governmental access.
Financial data suggests that hyperscalers' use of colocation facilities has grown substantially over the past few years. Their investments in colocations also show no signs of slowing down.
AWS has recently cut prices on a range of GPU-backed instances. These price reductions make it harder to justify an investment in dedicated AI infrastructure.
The terms "retail" and "wholesale" colocation not only describe different types of colocation customers, but also how providers price and package their products, and the extent to which customers can specify their requirements.
Current geopolitical tensions are eroding some European organizations' confidence in the security of hyperscalers; however, moving away from them entirely is not practically feasible.
Although the share of processing handled by the corporate or enterprise sector has declined over the years, it has never disappeared. But there are signs that it may reclaim a more central role.
Peter Judge
Max Smolaks
Andy Lawrence
Dr. Owen Rogers
John O'Brien
Paul Carton
Anthony Sbarra
Laurie Williams