Goldman Sachs Group Inc. is co-leading financing for a Texas project to build private power campuses for artificial intelligence.
The New York banking giant is working with real estate advisory Newmark Group Inc. to raise both equity and credit for the project building modular natural gas-fired generation serving a cluster of south Dallas data centers, according to developer GridFree AI. Newmark has also been tapped as the exclusive marketer and advisor for the data hub, called South Dallas One.
The aim in this initial round is to raise “hundreds of millions of dollars” with more rounds as the first project is scaled up, according to Philip Krim, chief executive officer of Montauk Capital, which incubated GridFree AI.
Goldman declined to comment.
The US power system, which is based on Thomas Edison’s 19th century grid, is struggling to meet the electricity needs of the modern economy. Aging infrastructure is being strained by extreme weather and unprecedented demand growth. Consumers are faced with a double whammy of rising power bills and a greater threat of blackouts in large parts of the US.
“Edison’s grid was not built for the scale and size of these data centers,” said Krim. “We ended up saying, ‘Let’s start a company for energy experts to build a data center — knowing how energy-dense and energy—hungry data centers are — to scale up the AI buildout.’”
GridFree AI was formed about a year ago with Ralph Alexander as CEO. Alexander, a nuclear engineer, previously served as the top executive at Talen Energy Corp., where he led the development of a data center campus next to the independent power producer’s Susquehanna nuclear plant in Pennsylvania. Amazon Web Services bought that campus early last year, which was expanded this year into a 1.92 gigawatt supply contract with Talen. This co-location deal quickly became the major flash point in making sure tech giants pay for their fair share of power plants and grid infrastructure without shifting those costs to other consumers.
“South Dallas One represents the future of AI-ready infrastructure: fast to deploy, resilient by design, and insulated from the grid constraints holding back the industry,” Alexander said in an statement.
With land for two of the three sites secured, GridFree said initial power can be brought online in less than 24 months from the lease being signed, significantly faster than waiting to connect to the Texas grid. Each site will have about 1.5 gigawatts of power with structures to house data centers. The plan is to build 100-megawatt modules — 10 gas turbines providing around-the-clock power and two for back-up — so they can be quickly plugged into projects, Krim said.
This set—up can be done on 500 acres and will use the one household’s worth of water because waste heat from power generation will be captured to help with cooling needs, he said.
While these AI campuses will be self-reliant, the goal is to ultimately connect to the Texas grid to supply spare power, Krim said. Fuel for the gas plants will be supplied by two pipelines, including one owned by Energy Transfer LP, and no back-up diesel generators will be needed, Krim said. GridFree AI plans to build campuses around the world and eventually that will include nuclear, he said.