Targeting 273GW by 2030! Goldman Sachs sharply raises global data center capacity forecast, with capital expenditure over the next five years potentially reaching US$8.5 trillion.

Goldman Sachs stated in its latest research report that, based on the latest data from the 451 Research database, it has updated its global data center capacity forecast. It now expects global data center supply to reach 273GW by 2030 (compared with 217GW in its previous update in July 2026), implying 170GW of new capacity after 2025, with net new capacity of 25GW, 32GW, and 35GW in 2026, 2027, and 2028, respectively. The upward revision is mainly because Goldman Sachs raised 451 Research's capacity estimates for the later part of the forecast period to incorporate projects that have not yet been announced. Goldman Sachs expects that the pace of net new data center capacity will not slow in the medium term. Based on the rule-of-thumb estimate of US$50 billion of investment per GW, this 170GW of new capacity implies capital expenditure of US$8.5 trillion over the next five years. At the same time, however, Goldman Sachs also noted that IT infrastructure costs continue to face upward pressure. For the U.S. market alone, Goldman Sachs expects data center capacity to reach 152GW by 2030 (compared with its previous forecast of 125GW), implying 103GW of new capacity after 2025. The upward revision is mainly due to Goldman Sachs raising estimates for the later part of the forecast period to incorporate projects that have not yet been announced. In addition, net new capacity from cloud services, emerging cloud service providers (neoclouds), cryptocurrency, and wholesale and retail data center operators also contributed to part of the revision (a cumulative 11.8GW of additions from 2026 to 2030). Goldman Sachs believes this reflects emerging cloud service providers rapidly scaling up to meet GPU demand. Goldman Sachs expects high-performance computing (HPC) data centers (typically associated with AI workflows) to drive most of the net new U.S. data center capacity growth, with projects of no less than 75MW accounting for 75% of net new capacity by 2028. A cumulative 103GW of new U.S. data center capacity over the next five years, based on the rule-of-thumb estimate of US$50 billion of investment per GW, would imply US$5.2 trillion of capital expenditure. Goldman Sachs believes that the current consensus forecast for U.S. hyperscaler capital expenditure is sufficient to cover this spending, at an average of about US$1.51 trillion per year. Historically, this forecast has been raised multiple times, and it has not yet included capital expenditure by private AI labs or capital expenditure by data center platforms, although it also does not exclude capital expenditure in international markets. Goldman Sachs pointed out that execution risks facing data center construction and delivery are rising, affected by factors such as power supply issues, regulatory uncertainty, and high leverage. Nevertheless, for projects with high certainty on power supply and permitting approvals, Goldman Sachs still believes data center capacity is expected to maintain sustained and above-trend growth. In addition, Goldman Sachs added that investors have recently raised increasing questions about possible project delays caused by data center-related directives issued by the governors of Texas and Pennsylvania. Both directives effectively paused fast-track approval processes in their respective states. Goldman Sachs believes the main impact will fall on speculative construction projects, but it is still necessary to watch for ongoing risks from related negative news and the possibility that project development progress may slow. Goldman Sachs' sustainability research team (GS Sustain) recently wrote a report analyzing in detail the main reasons for community opposition to data center construction and possible mitigation solutions. The team noted that these solutions may not necessarily fully eliminate community concerns. Community perceptions of the social and/or economic impact of AI, the transparency of the development process for individual data centers, and opposition driven by the broader "not in my backyard" (NIMBY) phenomenon may all override innovation and solutions. High-performance computing (HPC) data center construction typically carries relatively high leverage, with loan-to-cost ratios usually reaching 80% to 90%. This means that once debt yield spreads narrow, project returns will be highly sensitive. Goldman Sachs will continue to monitor further pressure on hyperscaler bond yields, as well as industry yield pressure brought by negative industry news and increased project delays. As credit costs rise, Goldman Sachs expects the market to focus more on the lease terms of data center assets and the credit quality of tenants. Debt financing backed by high-quality data center projects can obtain contracted cash flows from investment-grade tenants with terms of 15 to 30 years. In addition, as projects gradually mature, capital recycling provides developers with a way to monetize existing projects, thereby funding new project development at a lower cost of capital.