Most of the developers announcing data center projects across the country have never built one before.
That’s according to the analysis of Oliver Kerr, managing director of North America at Aurora Energy Research, which he explained onstage at the Transition-AI 2026 conference earlier this month. “Over half the planned capacity that we see in interconnection queues… is from data center developers with extremely large pipelines and funding ambitions to match, but with very little operational track record,” he said.
Newcomers account for 145 gigawatts out of 277 GW of the data center capacity planned in the U.S. over the coming years, as estimated by Aurora Energy based on projects that have secured site control, obtained permits, or have active interconnection requests. The company reached the number by taking the 780 GW of announced U.S. data center projects in the U.S., and first “weeding out” the more speculative, “phantom” projects that have been scrambling load forecasting across the country.
Of the remaining 132 GW, 51 GW have been announced by what Aurora describes as “aggressive scalers,” meaning operators whose whole pipeline is between 10 and 500 times bigger than their current operational capacity. Only 81 GW come from operators that are considered “proven,” with a reasonable pipeline that is less than ten times bigger than their current operational capacity.
Aurora’s numbers reflect a global trend. In December 2025, Bloomberg found that so-called “new players” are behind 32% of the global data center capacity planned for 2032.
It’s easy to assume these developers won’t be competitive with the old timers. Kerr noted, however, that the developers’ limited experience and grand expansion plans do not necessarily mean that they won’t be able to succeed — though it might inform their ability to get financing, and specifically “the effective discount rate you’d apply to the likelihood of these projects coming online when they say they will.”i.
Combined, 145 GW of new data center capacity would require over $6 trillion of investment for sites that are fully loaded, meaning they include everything from the basic infrastructure to cooling to GPUs. “That’s 12 times the GDP of Hungary, 15% of U.S. GDP, and 50% higher than the GDP of the U.K.,” Kerr said. “These are phenomenally large numbers that these folks would need to raise.”
There are also practical energy-related complications, like accessing enough generation and transmission, which together cast doubt over how many data centers can realistically get built. Newcomer Fermi, for instance, recently saw its CEO step down after struggling both to find an anchor tenant, and reportedly with the supply chains for especially the cooling systems needed for AI chips.
Physical logistics are a risk not only for newcomers. Even the “aggressive scalers” who have experience but are planning 51 GW of capacity, with pipelines hundreds of times larger than their current footprint, could struggle. Many developers are counting on building large and complex power projects they don’t necessarily have any experience with — especially in a moment when waits for both transmission and interconnection stretch to many years.
As Caroline Golin, Jigar Shah, Stephen Lacey discussed in a live taping of Open Circuit at the conference, many people involved in data center construction are underestimating power projects and applying the Silicon Valley philosophy of “move fast and break things” to an industry they don’t really know.
“They clearly know nothing about power,” Shah said. “I just think that one of the real tells… is [that] they talk about generation as if we’re short generation. We are not short generation in this country. We never have been. We are short wires.”
