Energy Institute at Haas

Artificial intelligence is driving unprecedented investment in data centers, yet the local costs and benefits of these facilities remain largely unmeasured. We estimate how households value nearby data center development using a newly constructed database of U.S. data centers and housing transaction data. Our statistical analysis uses a matched stacked event study design that compares housing transactions near newly announced large data centers with transactions in observably similar areas near data centers that are announced later. Large data centers reduce housing values within 3 miles by 3% on average over the following 5 years. The effect is highly localized: prices within 1 mile fall by 7.2% on average, with smaller effects beyond 1 mile. Price declines persist for at least 5 years and are larger for hyperscale and high-capacity data centers, urban locations, and the first large data center in an area. Consistent with increasing public awareness of data center externalities, before 2021, prices responded primarily when data centers became operational, while for more recently announced data centers, the effect is capitalized almost entirely at announcement. Nationwide, these price declines imply $20.6 billion in housing wealth losses.