Goldman Sachs estimates that data centers already account for 1-2% of global power demand and could rise to 3-4% by the end of the decade
In sum – what to know:
Data center demand to surge – Global power demand from data centers is forecast to rise 165% by 2030, with capacity expanding to 92 GW by 2027 at a 17% CAGR.
Hyperscale firms driving growth – The top five U.S. hyperscale companies are set to invest $736B in 2025–2026, fueling rapid build-outs in the U.S., China, and Europe.
Energy mix raises emissions – While 40% of new demand will be met by renewables and some nuclear, reliance on natural gas could add 215–220 million tons of CO2 by 2030.
Goldman Sachs Research forecasts global power demand from data centers to rise 165% by 2030 from 2023 levels, driven by the rapid adoption of artificial intelligence.
Current capacity is expected to expand by 50% to 92 GW by 2027, representing a compound annual growth rate (CAGR) of 17% between 2025 and 2028, the company said in a recent report.
Currently, global data centers consume approximately 55 GW, with AI accounting for 14%, cloud computing for 54%, and traditional workloads for 32%. By 2027, Goldman Sachs suggests that this could change significantly, with AI surging to 27%, cloud declining to 50%, and traditional workloads falling to 23%.
The report also noted that spending by U.S. hyperscale technology companies is surging, with the top five expected to invest $736 billion in 2025–2026 alone. This wave of construction spans major hubs such as Northern Virginia, Beijing, Shanghai, and Texas, as well as new projects across Europe, where energy providers are seeing a sharp rise in grid connection requests.
Goldman estimates that data centers already account for 1-2% of global power demand and could rise to 3-4% by the end of the decade. In the U.S., the weighting of power demand from data centers in the overall will increase even more, likely more than doubling by 2030 from 4% in 2023, the report added.
Goldman Sachs Research also estimates that 40% of the increase in power demand from data centers will be met by renewable energy, and there will be a modest amount of nuclear capacity that’s targeted for AI. The bulk of the remaining 60% is expected to be driven by natural gas. This will increase global carbon emissions by 215-220 million tons through 2030, equivalent to 0.6% of global energy emissions.
“Longer term, we see potential for a significant reduction of data center emissions intensity and potentially in absolute emissions as more nuclear power comes online and AI computing shifts to using AI models as opposed to training them,” Goldman Sachs analysts said.
While companies are racing to build AI-ready data centers, Goldman analysts caution that risks remain, including slower-than-expected monetization of AI or breakthroughs that lower power needs. The report also noted that demand could overshoot forecasts if GPUs grow even more energy-intensive or if AI adoption accelerates faster than anticipated.
