Battery Storage Sizing for Data-Center Flexibility Under Uncertain Grid Hosting Capacity: Firm Supply Versus Load Curtailment

The rapid growth of data-center (DC) demand is challenging transmission networks, where available capacity may accommodate new loads during most hours but not at all times. DC flexibility can enable earlier grid connections by reducing network withdrawals during congestion, while improving the utilization of existing transmission infrastructure. A DC whose bus hosting capacity varies hourly and is uncertain can become flexible with behind-the-meter storage, self-curtailment, or both. We formulate the least-cost sizing of battery energy and power capacity as a two-stage stochastic linear program with network-imposed charging limits, a transmission-usage (MUST) charge on the monthly peak recharge and optional curtailment at a penalty cost. For a 450 MW DC at a candidate bus in São Paulo, Brazil, with 100 hourly scenarios, full firm supply requires an 866 MWh/345 MW battery costing R$ 111.7 million per year to cover gaps in about 1.2% of the hours; energy and power capacities are set by different rare events. Curtailment at 8000 R$/MWh, of only 0.1% of DC consumption, reduces battery energy by 79% and total cost by 49%. Storage pays off above a curtailment cost of about 3700 R$/MWh, and firm supply is infeasible above 500 MW.