Research from Goldman Sachs Research warns that electricity shortages could cause a construction bottleneck.
According to a study by Goldman Sachs Research, data center energy consumption will more than double by 2030 due to the rising demand for AI technology. As companies continuously develop new and better AI models, in addition to new data centers, more and more energy is needed.
Strong increase
According to the study, data centers currently consume an estimated 55 gigawatts of electricity worldwide. Fourteen percent of that goes to AI systems, 54 percent to cloud computing and 32 percent to traditional business processes such as mail or storage. By 2027, consumption would rise to 84 gigawatts, 27 percent of that for AI, according to the study. By 2030, it would rise as high as 122 gigawatts.
Amazon has already spent $75 billion in capital investment by 2024, largely for AWS, and this amount is likely to increase. Microsoft predicts spending $80 billion, Meta around $60 billion. Power networks will also continue to expand, according to Goldman Sachs. That would cost an estimated $720 billion. Without those investments, data center expansion may be delayed.
Once the peak in data center utilization is reached, the pressure will ease due to more efficient AI technology. “If AI models become more efficient, that could reduce capital investment and market fluctuations in the long run,” stated James Schneider, senior analyst at Goldman Sachs.
Data centers remain large energy consumers that can triple CO2 emissions. Experts say that can be avoided by switching more quickly to renewable energy sources and using more efficient technologies to make data centers more energy efficient.