1 month ago
40
The current data centre construction boom in the US is creating hitting problems when it comes to securing enough power. Facilities built to support artificial intelligence workloads require levels of electrical capacity that local grids won’t be able to supply at present levels of capacity and infrastructure. In response, several DC operators and commissioning companies have begun to secure dedicated power generation, often built locally to the data centre site and fuelled by natural gas purchased directly from producers.
An engineering and power consortium led by Babcock & Wilcox, Base Electron, and Applied Digital has agreed to construct gas-fired generators to supply new AI data centre campuses in North Dakota, with a reported value of $2.4 billion. Babcock & Wilcox will design and construct four gas turbine units, providing 1.2 gigawatts of on-site generation.
To put that in context, 1.2GW is comparable to the output of an average nuclear reactor. According to calculations by the US Energy Information Administration, sustaining such output will need nearly 5.6 million cubic metres of natural gas per day, an requirement comparable to a medium-sized US city.
Applied Digital’s chief executive described the project as supply that will feed directly into AI computing, and indicated the company is ‘evaluating’ a further 1.2GW of gas-fired generation (with the same partners) to support later development phases. AI processing ability is constrained primarily by electricity availability, so power supply levels determine revenue potential.
The International Energy Agency estimates that global electricity consumption by data centres will increase from roughly 460 terawatt hours in 2024 to more than 1,000TWh by 2030, with the current power used by DCs worldwide the same as that of a medium sized industrial economy.
The IEA expects natural gas to supply around 26% of DC demand by the end of this decade. Renewable energy capacity is also expanding quickly, although the preference for non-weather dependent supply makes renewables less suitable for facilities where computational loads are consistently high.
North America is currently the largest constructor of DCs, with more than 35GW of capacity under construction, according to real estate firm JLL. If all planned projects were completed and powered primarily by gas generation, Enverus Intelligence Research estimates that US natural gas demand could increase by roughly 0.12 billion cubic metres per day by 2030.
Energy producers are adjusting accordingly, with Comstock Resources stating it’s concentrating its drilling activity in northeast Texas, in part because of new data centre development planned in the region. According to JLL, Texas could surpass Northern Virginia as the largest global data centre conglomeration by the end of the decade.
Data centre operators are entering into direct supply agreements with gas producers. Comstock has contracted to deliver gas to NextEra Energy’s power plants to support AI data centres’ projected demands. The objectives in such agreements are to stabilise fuel supply and reduce exposure to regional electricity market price fluctuations.
US Public policy supports the tendency to keep it local and internal to any agreement’s partners, expressing concern that rapid data centre expansion could increase electricity prices for domestic households if new DCs draw primarily from the existing grid. The Trump administration is encouraging technology companies to finance dedicated generation capacity for their own facilities, adding the cost of power generation to tech companies’ operations, but minimising the impact on existing infrastructure.
Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI have signed a pledge committing themselves to financing new generation capacity and transmission infrastructure upgrades to supply the data centres they operate.
Commercial power suppliers are also contracting with technology giants directly: AES Corporation recently signed a 20 year agreement to provide power for Google’s data centres in Texas. Meta, OpenAI, and CloudBurst have separate arrangements for natural gas supply and gas-fired generation dedicated to their facilities.
At a smaller scale, Atlantic Energy is to supply electricity to a 150MW AI and cryptocurrency data centre operated by Viking Data Centres in Akron, Ohio, a facility that will occupy a former Goodyear tyre factory. The facilities at Akron will run primarily on gas sourced from fields in the northeastern US.
While access to computing hardware remains important, access to power is equally decisive. DC operators securing dedicated power generation will escape capacity constraints should AI demand meet AI vendors’ predictions.
Electricity supply has become part of the costs of AI, taking its place among loan repayments, training and inference costs, hardware, marketing and personnel, the sum of which currently outpaces income from AI services. Firms controlling compute capacity and energy supply may hold a structural advantage if artificial intelligence deployment by end-users hits the levels predicted by AI companies. Uncertainty remains about the long term fossil/sustainable energy mix, the pace at which renewable generation can scale, and where the cost burden for providing AI falls. Artificial intelligence infrastructure is beginning to resemble that of heavy industry – energy procurement being part of core investment.
(Image source: “Dual Power” by SPIngram is licensed under CC BY-NC-ND 2.0.)
Want to learn more about Cloud Computing from industry leaders? Check out Cyber Security & Cloud Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information.
CloudTech News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.