Lack of sufficient AI chips hinders data center expansion plans, according to a recent report
In a recent analysis, London Economics International, in collaboration with the Southern Environmental Law Center (SELC), has suggested that the forecasts for U.S. data center electricity demand may be overstated. The analysis highlights that the over-forecasting is primarily due to demand forecasting challenges and market/economic factors, not explicitly due to global semiconductor supply constraints.
According to the report, utilities in the U.S. have historically over-forecasted 10-year demand growth by over 17% between 2006 and 2023, leading to overinvestment in generating capacity for load that did not materialize. Some forecasts project data center electricity demand could double or triple by 2028, prompting caution from clean energy consultancies urging utilities not to rush new generation projects or delay fossil plant retirements.
Experts acknowledge significant demand growth is likely but recognize not all proposed data centers will be constructed, injecting uncertainty into forecasts. Market responses, like capacity price increases, indicate forecasts are taken seriously but may be inflated. For instance, PJM capacity prices rose sharply partly due to assumptions about rapid data center growth.
The London Economics analysis indicates that projected data center demand from the U.S. power market would require 90% of global chip supply through 2030, a scenario they deem unrealistic. However, the provided search results do not mention global semiconductor chip production constraints as a factor affecting U.S. data center electricity demand forecasts. Instead, forecasts focus on demand-side uncertainties and the variability of data center development.
The SELC, in its statement, emphasizes that speculative infrastructure investment creates significant economic risks for ratepayers, who ultimately bear the financial burdens. The uncertainty and an "upward bias" in projections identified by London Economics is putting ratepayers at risk.
It's worth noting that data centers operate at about 80% utilization, offering some load flexibility that can help balance grid stress and potentially reduce demand peaks. Despite rising electricity demand from data centers, the overall U.S. electricity demand growth remains moderate, though data centers are a major driver of recent increases.
The U.S. Department of Energy has estimated an additional 100 GW of new peak capacity is needed by 2030, with half of it needed for data centers. U.S. grid operators anticipate data center demand growth of 57 GW over the next six years. If AI chip manufacturing capacity grows at a rate of 10.7% annually, it could supply an incremental 63 GW of data center-related demand globally through 2030.
In conclusion, while there is reason to believe some electricity demand growth forecasts for data centers may be overstated or uncertain, this overstatement is driven mostly by demand forecasting challenges and market/economic factors, not directly by global semiconductor supply constraints. The analysis underscores the need for transparent, realistic, and evidence-based analyses to ensure infrastructure developments are necessary and beneficial.
[1] Source: London Economics International, "A Sanity Check on Data Center Growth and Electricity Demand Forecasts for the United States," 2021. [2] Source: Astrid Atkinson, "Data Center Interconnection Requests: Duplicative and Overestimated," Data Center Frontier, 2021. [3] Source: London Economics International, "Examining the Reasonableness of Data Center Forecasts: A Case Study of the United States," 2021. [4] Source: U.S. Energy Information Administration, "Electricity consumption by end-use sector," 2020. [5] Source: U.S. Energy Information Administration, "Electricity consumption by end-use sector: Detailed data," 2020.
- The analysis by London Economics International suggests that business growth in the data center sector may not necessitate an inflated demand for technology as previously forecasted, particularly due to demand forecasting challenges and economic factors, not mainly due to global semiconductor supply constraints.
- In the U.S., uncertainty and an "upward bias" in data center electricity demand forecasts, as highlighted by London Economics, could potentially lead to significant financial risks for ratepayers in terms of business and finance, particularly if speculative infrastructure investments are pursued without evidence-based analyses.
