To Defer or To Shift? The Role of AI Data Center Flexibility on Grid Interconnection

Overview

Integrating AI data centres into the power grid is among the most complex challenges for energy systems, and the traditional grid-planning paradigm of treating data centres as rigid, inflexible loads is becoming economically, mathematically and operationally untenable. A cited projection says U.S. AI data-centre demand could grow more than 30-fold to 123 GW by 2035 from 4 GW in 2024, and median time from interconnection request to commercial operation is over five years. The work tries to understand the large-load interconnection bottleneck by modelling AI load flexibility.

Depending on data-centre location, flexibility range and grid load conditions, flexible AI load reduced grid investment and operational costs by 3-21%. In one case flexibility reduced operating cost enough that the planner spent more on generation: total cost fell 3.5% while investment rose 7.1%. On the WECC 240-bus system the benefit of deferral was almost the same for delay windows of 3 hours or more, so short-term flexibility already captures most of the grid value. In the IEEE 14-bus study, data centres on buses connected to congested lines benefit most from geographic load shifting.

The finding is that flexibility lowers total system cost but does not always reduce the generation capacity that has to be built: its effect is not felt consistently, and more flexibility does not necessarily mean less generation capacity. Deferral windows longer than about 3 hours bring diminishing returns.

Limits: two test systems (IEEE 14-bus and WECC 240-bus) and a capacity-expansion model with DC power-flow constraints; results depend on these systems and the assumed flexibility participation rates. It is a preprint.

Key points

1. Depending on data-centre location, flexibility range and grid load conditions, flexible AI load reduced grid investment and operational costs by 3-21%.
2. Cited projection: U.S. AI data-centre demand could grow more than 30-fold to 123 GW by 2035 from 4 GW in 2024, and median time from interconnection request to commercial operation is over five years.
3. In one case flexibility reduced operating cost enough that the planner spent more on generation: total cost fell 3.5% while investment rose 7.1%.
4. On the WECC 240-bus system the benefit of deferral was almost the same for delay windows of 3 hours or more, so short-term flexibility already captures most of the grid value.
5. In the IEEE 14-bus study, data centres on buses connected to congested lines benefit most from geographic load shifting.

Using it in design

• In large-load interconnection negotiation, present AI load flexibility (a deferral window of up to about 3 hours) as a lever for interconnection capacity and timing; returns diminish beyond 3 hours.
• Test in planning scenarios that flexibility does not always reduce generation investment (total cost fell 3.5% while investment rose 7.1%).
• Assess the value of geographic load shifting separately for buses connected to congested lines.

Who should read it

Planners and grid specialists evaluating large data-centre load interconnection, grid capacity planning and load flexibility.

Document contents

1. 1 Introduction
2. 2 Capacity Expansion Model with Data Centers
3. 3 Test systems and scenarios (IEEE 14-bus, WECC 240-bus)
4. 4 Results
5. 5 Conclusion and Future Works

Read the document

Publication details

Authors

Yize Chen; Xiaogui Zheng

Publisher

arXiv preprint (University of Alberta)

Year

2026

Document number

arXiv:2604.05376

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.48550/arXiv.2604.05376

Licence

CC BY 4.0 — share with attribution

Licence evidence

arxiv.org

Citation