Goldman Sachs: Global Data Center Supply/Demand Update — Tight Conditions Likely to Extend Through 2026
Summaries are public. PDF access requires an active membership; all members have the same access. Sign in
Goldman Sachs has updated its global data center supply and demand model, concluding that current tight market conditions will likely persist through 2026. Hyperscale capital expenditure expectations have been broadly revised upward following 4Q25 earnings, driven by robust cloud and AI services demand. Enterprise AI adoption is emerging as an incremental demand driver, with Equinix reporting that 50% of its 4Q25 AI bookings came from non-cloud and non-IT providers, and Digital Realty offering similarly positive commentary.
The baseline forecast projects supply adequacy at 92% in 2025, rising to 94% in 2026, then declining to 93% in 2027 and falling to 87% by 2030, representing a two-percentage-point reduction from the prior model. Global data center demand stands at approximately 74 GW in 4Q25, with cloud workloads accounting for 57%, traditional workloads 26%, and AI workloads 17%. Demand is projected to grow 42% to 110 GW by 2030, with AI workloads expanding to nearly 40% of the total.
On the supply side, global online capacity is expected to exceed 170 GW by 2030, reflecting a five-year CAGR of 17%. Power density is forecast to increase from 159 watts per square foot to 190 by 2030. The report evaluates three scenarios: an AI upside case where 2030 occupancy could rise 21 percentage points implying supply shortages, a cloud downside case reducing occupancy by 4 points, and an oversupply scenario lowering occupancy by 7 points.
Key monitoring variables include hyperscale AI inference monetization progress, supply disruptions at large-scale AI infrastructure projects such as OpenAI, Oracle, and sovereign AI initiatives, and enterprise demand optimization through