Goldman lifts 2030 data center capacity forecast to 273 gigawatts

Goldman Sachs's November 2 research note marks a significant upward revision to its global data center capacity outlook, now expecting 273 gigawatts by 2030 instead of 217 gigawatts. The change reflects stronger demand for artificial intelligence infrastructure and the expectation that more projects will be announced. The revised forecast implies 170 GW of additional capacity from 2025, when worldwide supply was approximately 103 GW.

The U.S. is expected to account for roughly 60% to 70% of worldwide additions through 2030, with its capacity estimate rising to 152 GW from 125 GW previously. Expanding cloud, neocloud and related facilities supporting GPU demand contributed to this revision, adding 11.8 GW to cumulative capacity forecasts for 2026 through 2030. Large high-performance computing facilities associated with AI workloads are expected to dominate construction, with projects of at least 75 megawatts potentially representing 66% of global additions and 75% of U.S. additions by 2028.

However, the outlook also highlights mounting construction risks, including limited power availability, permitting delays, community opposition and rising financing costs. High-performance computing developments typically rely on debt financing covering 80% to 90% of project costs, and projects with secured electricity supplies and clear regulatory approvals are seen as better positioned to proceed. These risks could shape which projects move forward as developers face tighter financing conditions and heightened scrutiny.