Bank of America Merrill Lynch elaborates on AI power demand, cooling demand, and challenges facing the US power grid

Bank of America Merrill Lynch stated in its latest research report that with the rapid development of AI technology, the electricity demand of the U.S. power grid will face unprecedented growth. It is estimated that from 2023 to 2030, the electricity demand in the United States will grow at a compound annual growth rate (CAGR) of 2.8%.

Bank of America Merrill Lynch believes that this growth is mainly driven by the expansion of AI data centers, industrial growth, the adoption of electric vehicles, and the electrification of buildings. Among them, the demand for electricity in data centers is particularly prominent.

Bank of America Merrill Lynch predicts that by 2030, the electricity demand of data centers is expected to grow at a compound annual growth rate of 11%. This is equivalent to approximately 1-2% of global electricity demand, or 240-340 terawatt-hours (1 terawatt-hour equals 1 billion kilowatt-hours).

In addition, Bank of America Merrill Lynch predicts that the electricity consumption of AI will grow at a compound annual growth rate of 25-33% between 2023 and 2028, which will pose challenges to the design and operation of data centers.

Data center developers are also paying more attention to power issues. For example, Amazon Web Services (AWS) signed an agreement in March 2024 to purchase a data center next to a nuclear power plant in Pennsylvania, with the potential to obtain up to 960 megawatts of dedicated power from the plant. Microsoft Azure is building a 170-megawatt natural gas power plant in Dublin to provide basic load power for its €9 billion data center.

AI Cooling Demand

Bank of America Merrill Lynch points out in the report that with the increase in data center power density, traditional cooling methods may no longer be applicable. For example, 85% of data centers have a maximum rack power density of less than 30 kilowatts, and for power densities above 20 kilowatts, liquid cooling solutions may be required.

This technology involves directly delivering the coolant to the chip, with each chip having two pipelines (one hot and one cold).

Furthermore, Bank of America Merrill Lynch states that as data centers transition to higher power densities, the demand for power and thermal management content is also increasing. For example, Nvidia's liquid cooling supplier, Vitesco Technologies, expects opportunities for power and thermal management content per megawatt to increase by 18% and 740%, respectively.

Challenges Facing the U.S. Power Grid

Bank of America Merrill Lynch believes that in addition to data centers, industrial demand is also a key factor driving the growth of electricity demand in the United States. For example, spending on U.S. manufacturing facilities has increased by 60% in the past 12 months, reflecting growth in manufacturing construction In addition, there is also the popularization of electric vehicles and the electrification of buildings. For example, in 2023, electric passenger cars accounted for 7% of total sales in the United States, and it is expected that by 2030, the U.S. electric vehicle fleet will account for 13% of the total; the average power consumption of all-electric homes is 127% higher than that of regular homes.

Bank of America Merrill Lynch stated that as the demand for electricity in the United States increases, data center and grid operators need to find more efficient power and heat management solutions to support this growth.