Barclays has published a list of Power-as-a-Service companies it believes are well positioned to supply the behind-the-meter power capacity data centers will require as artificial intelligence infrastructure spending rises. The bank's analysis notes that AI-related infrastructure investment from Western hyperscalers and AI laboratories could top $1 trillion before peaking in 2028, and that annual AI infrastructure spending could run as much as $300 billion higher than current consensus estimates.
Those spending projections are tied to growth in hyperscaler capital expenditures. Barclays' view anticipates that recursive self-improvement in AI - a factor in the pace of training activity - should slow training demand growth by 2028. In the near term, however, data center operators and their customers are locking in large, behind-the-meter power arrangements with pick-and-shovel providers that can deliver multi-gigawatt solutions.
Companies highlighted by Barclays
-
Williams Company - Barclays identifies Williams as executing roughly $7 billion of behind-the-meter power projects for data centers. The company has a 5 gigawatt agreement with Siemens Energy covering various turbine types to expand its power generation portfolio. Williams is reported to be in advanced talks to acquire Momentum Midstream for about $5.5 billion. The stock has seen bullish signals from analysts, with UBS raising its price target on the basis of growth in Williams' power innovation business, while Jefferies has maintained a Buy rating.
-
Halliburton - The oilfield services firm has formed a strategic collaboration with VoltaGrid to deploy modular natural gas power systems for data centers, initially targeting 400 megawatts of capacity in the Middle East. UBS has increased its price target for Halliburton, citing an improved outlook for North American operations, and Wolfe Research has initiated coverage with a Peerperform rating.
-
Fermi America - Barclays notes Fermi America's proposed development of an 11 gigawatt, behind-the-meter campus in Amarillo, Texas, intended to support data center demand. The project would comprise 4.5 gigawatts of gas-fueled power and 6 gigawatts of nuclear capacity.
-
Liberty Energy - Described as an energy services provider targeting delivery of 1 gigawatt of capacity by the end of 2027. Included in that pipeline is a 400 megawatt project with Vantage and a 330 megawatt agreement with another data center developer.
-
Solaris Energy Infrastructure - Barclays cites Solaris as aiming to reach 2.2 gigawatts of capacity by early 2028 across multiple commercial end markets, with more than 1 gigawatt secured at two data center projects.
-
Kodiak Gas Services - Through its acquisition of Distributed Power Solutions, Kodiak added approximately 380 megawatts of gas-fired power capacity that serves data center and microgrid customers.
-
ProPetro - Identified for providing utility-like power services using gas-to-power solutions that serve both Permian Basin operations and the data center sector.
-
Atlas Energy Solutions - The company is targeting 400 megawatts deployed by early 2027, with most of that capacity already contracted under long-term agreements.
Barclays frames these firms as essential suppliers - the pick-and-shovel providers - enabling the physical expansion of compute-heavy AI infrastructure. Their work centers on behind-the-meter installations, modular or utility-like gas generation, and larger buildouts that combine different generation technologies to meet the continuous and high-density power needs of data centers.
Investors and market participants will likely watch execution on announced projects, contract terms and the progression of hyperscaler capex plans as signals of near-term demand for PaaS offerings. Barclays' selection spans established energy infrastructure companies and newer power-services providers, reflecting an industry response to the scale of projected AI-related energy demand.
Note on scope - The companies above were enumerated in Barclays' review of Power-as-a-Service providers supporting data center power needs tied to AI infrastructure buildout. The analysis and company descriptions reflect the information presented by Barclays and related analyst commentary; this article does not expand beyond those reported facts.