巴克莱表示，这一增长预测远高于许多机构预测的15%，也高于巴克莱在此前报告Powering AI: Calibrating US Data Center Energy Demand中预期的14-21%。不过，巴克莱在这一报告中指出，由于数据中心部署、设备类型、运营实践和冷却系统的快速变化，预测存在不确定性。Considering the growing demand for AI Servers, the US Energy Department predicts that from 2023 to 2028, the Electrical Utilities demand of Datacenters in the United States is expected to grow by 13-27% each year, higher than the 15% forecasted by many Institutions.
The market still underestimates the energy-consuming monster AI.
On January 3, Barclays Analyst William Thompson and his team released a report analyzing and summarizing the approximately 70-page congressional authorization research report released by the US Department of Energy (DOE). The DOE predicts that starting in 2023, the electrical utilities demand of data centers in the USA (excluding cryptos) will increase by about 13-27% annually, reaching 325-580 TWh by 2028 (1 terawatt-hour = 1 billion kilowatt-hours), accounting for 6.7%-12% of the total electrical utilities demand in the USA. According to this forecast, the electrical utilities demand of data centers in the USA is expected to grow 2-3 times from 2023 to 2028.
Barclays stated that this growth forecast is much higher than the 15% predicted by many institutions and higher than Barclays' previous report "Powering AI: Calibrating US Data Center Energy Demand,' which expected 14-21%. However, Barclays pointed out in this report that there is uncertainty in the forecast due to rapid changes in data center deployment, equipment types, operation practices, and cooling systems.
In addition, the US Department of Energy also predicts that in the coming years, the water consumption of data centers in the USA will grow faster than the increase in electrical utilities demand. Currently, the direct water consumption has soared from 21 billion liters in 2014 to 66 billion liters in 2023, and it is expected to grow by approximately 2-4 times by 2028.
The energy demand growth of data centers exceeds market expectations.
The report from the US Department of Energy shows that the electrical utilities consumption of data centers in the USA was 176 TWh in 2023, accounting for 4.4% of the total electrical utilities demand in the USA. It is projected that by 2028, the electrical utilities demand will increase to 325-580 TWh, accounting for 6.7%-12% of the total electrical utilities demand in the USA.
According to the Department of Energy's research, AI servers are the main driver of the growth in electrical utilities demand.
According to the report, the electrical utilities demand for AI Servers is expected to increase from about 40 TWh in 2023 to about 165 TWh (low estimate) to 325 TWh (high estimate) by 2028, representing a growth of approximately 4-8 times. Research shows that by 2028, the consumption for AI training will exceed the demand for AI inference, accounting for 50-53% of the total energy consumption of AI servers, as more high-power GPUs will be utilized for AI training.
As for the electrical utilities demand for traditional servers, the US Department of Energy expects little change, increasing from about 60 TWh in 2023 to approximately 60 TWh (low estimate) to 70 TWh (high estimate) by 2028.
The water consumption of datacenters is expected to grow faster than the increase in electrical utilities demand.
The Department of Energy's research also indicates that the direct water usage of US datacenters (primarily related to cooling) has soared from 21 billion liters in 2014 to 66 billion liters in 2023, and is expected to grow by about 2-4 times to around 145-275 billion liters by 2028.
Interestingly, the report also states that the direct water intensity (water usage per kilowatt-hour) is expected to rise from about 0.38 L/kWh in 2023 to approximately 0.45-0.48 L/kWh in 2028. Barclays assumes that this reflects the change in the proportion of water-cooled chillers — used to cope with the increasing energy density of AI datacenters.
It is worth noting that the report's forecast does not include nearly 80 billion liters of indirect water usage, which is the water used for generating electricity, such as water used for nuclear energy, steam for coal plants, etc. If this portion of water is included, it would add approximately 4.5 liters of water consumption per kilowatt-hour.
Editor/ping
