Goldman Sachs: Data Center Power Demand to Surge 170%, AI Innovation Cycle Approaching Efficiency Inflection Point
TechFlow reports, according to TechFlow Research, Goldman Sachs' August 6 research report significantly raised the 2030 global data center power demand growth from 117% to 170%, with cumulative new electricity consumption expected to exceed Japan's total national electricity consumption in 2023. Hyperscalers' combined capex and R&D are expected to reach $1.2 trillion in 2026, further increasing to $1.7 trillion in 2027. Goldman Sachs estimates that if the U.S. green reliable power premium ($40 to $48 per megawatt-hour) is applied to global data center new demand, total industry expenditure in 2030 will be approximately $45 to $53 billion, accounting for 2.9% to 3.4% of hyperscalers' expected EBITDA in 2028.
Goldman Sachs believes the AI innovation cycle is shifting from the "evaluation and dream stage" to the "execution and efficiency stage", and community opposition, labor shortages, and physical environmental constraints are becoming new bottlenecks restricting data center construction. By 2030, data center carbon emissions will increase by approximately 370 million tons (+210%) compared to 2023, equivalent to 1% of global energy emissions. AI has increased the success rate in drug discovery by 370 basis points, discovering 28 additional new drugs annually, reducing the time from R&D to first revenue from 13 years to 10 years, creating corporate present value of approximately $84 to $93 billion. Goldman Sachs is optimistic about structural opportunities in fields such as power equipment, cooling solutions, grid infrastructure, and key materials.





