Strategic Data Center Load Shifting: Implications for Market Efficiency and Transmission Value

Data center electricity use may reach 12% of U.S. demand by 2030, alongside growing ability to shift workloads geographically in response to prices or carbon signals. We examine the system-level implications of such strategic flexibility using a bilevel two-zone model that couples economic dispatch with consumer cost minimization. Two market failures emerge. First, discontinuous price changes at generator capacity limits can induce flexible consumers to shift load in socially inefficient directions; for example, toward a higher-cost region to trigger a price drop elsewhere. Second, by positioning near capacity boundaries, consumers can counteract the marginal benefit of transmission expansion: although shadow prices suggest additional capacity is valuable, strategic consumers reoptimize to offset resulting flow changes, leaving dispatch and costs unchanged. We derive conditions under which these effects arise and show that conventional price signals can misrepresent system value in the presence of large spatially flexible loads.

💡 Research Summary

The paper investigates how large, geographically flexible loads—exemplified by modern data centers—can behave strategically in electricity markets and thereby undermine both market efficiency and the perceived value of transmission expansion. Using a bilevel optimization framework with two zones (A and B), the upper level models a flexible consumer who chooses a net load shift δ to minimize its own procurement cost, while the lower level represents the system operator’s economic dispatch that minimizes total generation cost subject to generation capacities and a transmission limit F.

Key findings are organized around two market failures. First, because locational marginal prices (LMPs) are piece‑wise constant and jump discontinuously when a marginal generator reaches its capacity, a consumer can exploit these jumps. Theorem 1 (Externalities at basis changes) identifies three sufficient conditions under which a consumer will shift load toward the higher‑cost zone to trigger a de‑commitment of a cheaper marginal generator in the lower‑cost zone. This results in a reduction of the consumer’s own cost (ΔP < 0) while increasing total system cost (ΔG > 0), i.e., a misalignment of private and social incentives. The paper illustrates this with a numerical example where shifting 20 % of total demand causes the marginal generator in zone A to drop from $25/MWh to $0/MWh, yet forces more expensive generation in zone B, raising overall generation cost.

Second, the paper shows that the conventional shadow price of a transmission constraint no longer reflects its true marginal value when flexible loads act strategically. Theorem 2 (Zero bilevel marginal value of transmission) proves that, when the consumer positions its load exactly at a basis‑change point that keeps the cheap generator in zone A marginal, any increase in transmission capacity F is offset by a one‑for‑one increase in the optimal load shift δ*. Consequently, the system‑level objective G remains unchanged (−∂G/∂F = 0) even though the shadow price μ⁺ stays positive. Meanwhile, the consumer’s cost P rises because more load is forced into the higher‑cost zone B (−∂P/∂F < 0). This demonstrates that transmission expansion can appear valuable in market clearing models but deliver no real cost savings once strategic load shifting is accounted for.

The authors discuss the implications for market design and transmission planning. Relying solely on LMP‑based price signals may be insufficient to curb strategic behavior of large flexible loads; additional mechanisms such as non‑price incentives, capacity‑based payments, or explicit market‑power limits may be required. Moreover, transmission investment decisions that use shadow prices as the sole metric risk over‑investing, because strategic load responses can neutralize the expected welfare gains. The paper calls for incorporating bilevel or game‑theoretic models into planning tools and for empirical validation with real data‑center workload and market data.

Overall, the study provides a rigorous analytical foundation showing that as data‑center demand grows to a significant share of total electricity consumption, its spatial flexibility can generate counter‑intuitive outcomes—price‑driven load shifts that increase system costs and render transmission upgrades ineffective—unless market rules are adapted to anticipate and mitigate such strategic actions.