| Datacentre power
Energy and Location Strategy at data center operator CyrusOne.
AI’s insatiable power demands Omdia predicts that by 2030, AI will drive more than 50% of installed MW of power generation capacity and AI usage will increase by up to 100 times current levels. Despite so many datacentres coming online each year, available datacentre space for leasing is at an all-time low. “The pace of datacentre investment is accelerating and we expect that to continue for some time to come,” said Howard. “Datacentre developers can’t get enough power as there is so much demand. The amount of compute power being consumed by AI data centres and applications has increased by two orders of magnitude in the last six months.” By 2030, hundreds of datacentres will be operating that require 1 GW or more of generating capacity. Hyperscalers such as Google, Microsoft, Oracle, Meta, and Amazon, along with chipmaker Nvidia, are driving the AI datacentre investment boom. Oracle, for example, is scrambling to get power from anywhere it can. It is a senior partner in the vast Stargate datacentre complex that is being built in Abilene, Texas using a potpourri of power assets including GE Vernova LM2500XPRESS gas turbines, Solar Turbines Titan 350 units, Dynamis DT35 turbines, and fuel cells from Bloom Energy. It will eventually take up eight buildings across 4 million square feet, requiring 1.2 GW. And that’s just the start. Oracle is adding another 60 or more facilities to the 147 it currently operates. In addition, Oracle is adding a lot of battery energy storage systems (BESS). Storage is becoming an increasingly vital part of datacentre projects as one element of a sophisticated array of power electronics being added to cope with the volatility of AI computing workloads, which can swing from 300 MW to 1 GW within a few milliseconds. “Gas turbines can’t cope with massive power swings produced from AI datacentres that can happen several times a second,” said Ram Nagappan, Vice President of AI Infrastructure at Oracle Cloud Infrastructure. “BESS can act as a buffer to dampen these power swings, which is beginning to emerge as a new requirement for AI datacentres.”
Better grid citizens
Such arrangements are increasingly being mandated by regulators and utilities as part of datacentre permitting processes. Storing energy in battery energy storage systems is only one part of it. Other facets include consenting to load shedding and demand response agreements during peak periods or outages. Voltage ride through capabilities, having enough onsite backup generators and batteries to operate for extended periods during a grid emergency, and willingness to ramp down AI processing when the grid needs support are becoming common
Datacentre developer investment in battery energy storage (blue) and gas turbines (green). It is projected to rise steadily each year until at least the end of the decade. Gas turbines are widely used where grid power is scarce. In the face of supply constraints, a market for refurbished machines is emerging. Battery energy storage systems complement turbines and also enable peak shaving capability. Graph: Omdia
Stargate datacentre campus in Abilene, Texas, is obtaining power from wherever it can get it – including these GE Vernova LM2500XPRESS gas turbines. Photo: GE Vernova
features in new designs. Google is even designing systems that can switch traffic from a datacentre in one utility zone to a different region during grid events.
“Datacentres are going to have to provide AI smoothing, more flexibility, and adhere to new grid requirements such as low voltage right through,” said Sean James, a datacentre designer and engineer at Nvidia.
Another design change is the introduction of 800 V DC power. Grids have traditionally operated on AC power and then a complicated AC to DC to AC switching goes on once that power is delivered to the datacentre. As well as higher efficiency, and being able to bring in a lot more power to the datacentre, 800 V DC significantly reduces the amount of cabling needed to
support AI – hence more space is made available for computing. Google is one of those in the forefront of introducing 800 V DC into the datacentre.
Dealing with local opposition There is no doubt that some of the planned AI facilities won’t see the light of day. Some projects are being cancelled, some scaled back, and local opposition groups sometimes stall projects that have already been approved. Nevertheless, the volume of construction, groundbreakings, announcements, and investment continues to rise. “The pace of AI construction and investment has accelerated in the past year,” said Howard. “The volume of datacentre activity planned is unprecedented.”
www.modernpowersystems.com | October 2026 | 25
USD billion
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47
