Morgan Stanley (NYSE:MS) analysts warned of a challenge to the electric vehicle (EV) industry's push to reduce CO2 emissions: the surging power demands of data centers.
"Total projected US data center power usage (base data center power + GenAI power) to reach 337 TWh by 2027," they note, adding that it represents a 20.5% compound annual growth rate (CAGR).
This includes a 105% CAGR for power used by Generative Artificial Intelligence (GenAI) from 2023 through 2027. Extrapolating this growth, Morgan Stanley estimates US data centers could consume nearly 450 TWh by 2030, representing a significant chunk (10.6%) of the total US power grid.
"US data center power usage may be equivalent to the power used by 150 million electric cars by 2030," Morgan Stanley adds. Their analysis assumes a typical EV's range efficiency of 4 miles/kWh by 2030 and an annual mileage of 12,000, leading to a consumption of roughly 3 MWh annually.
This translates to a significant near-term impact. Morgan Stanley estimates the 2023-2027 increase in US data center power usage is equivalent to adding 59 million EVs – a 21% jump in total US vehicles.
The analysts point out the contrast with EV production timelines. " It took the past 16 years for the US car population to add 59 million cars," says Morgan Stanley. "At Tesla (NASDAQ:TSLA)'s current FY24 production rate, it would take nearly 35 years to deliver an incremental 59 million units."
The report also highlights the power-hungry nature of AI chips. "An NVIDIA (NASDAQ:NVDA) H100 Hopper running at 700 Watts and 70% utilization would consume approximately 4.3 MWh of power annually," Morgan Stanley calculates, emphasizing the need for efficiency improvements.