U.S. Data Center Growth Through 2030: A Critical Review of Grid Adequacy and Transmission Expansion Planning
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.
Rosa: We're looking at the paper: "U.S. Data Center Growth Through 2030".
Dev: Here's the gist of the paper. The growth of large data centers raises questions about both supply adequacy and the suitability of load representations used in connection studies This paper critically reviews the U.S.
Rosa: demand, generation, and transmission outlook through 2030 and examines the implications for transmission expansion planning (TEP) Demand Projections and Forecasting Discrepancies Public forecasts indicate substantial growth in data center demand through 2030, but annual average demand, peak demand, and grid connection capacity describe different quantities and cannot be used interchangeably National generating capacity totals cannot establish a dependable surplus because resource availability, regional adequacy, transmission constraints, and project timing determine supply capability Annual average demand is estimated at approximately 74.1 GW for the reference case in 2030.
Dev: Peak-demand forecasts project 45, 71, and 94 GW under low, medium, and high scenarios for EPRI.
Rosa: The market analyses reviewed by Grid Strategies suggest approximately 65 GW of additional datacenter demand through 2030.
Paper summary: Rosa: So wrapping up the "U.S. Data Center Growth Through two thousand thirty: A Critical Review of Grid Adequacy and Transmission Expansion Planning," what the authors are really telling us is that we need to be much more careful about how we measure demand and supply when looking at this growth <ref:2611.14759#pg1,Data Center Growth Through 2030: A Critical Review of Grid Adequacy and>.
Dev: They are arguing that the way public forecasts describe annual average demand, peak demand, and connection capacity are not interchangeable because they measure totally different things about the load.
Taro: And they’re emphasizing that national generating capacity totals don't give you a sure surplus because regional adequacy and transmission constraints play huge roles in whether power gets where it needs to go.
Rosa: The implication for planning is that we can’t rely on just one forecast number; we have to look at the specific customer requests and their location within the network, not just aggregate totals.
Dev: They also stress that dynamic behavior, like how demand shifts after a fault clears, has been poorly integrated into connection studies so we haven't fully understood the real performance of a system under stress.
Taro: So what this means for someone listening who isn’t in the lab—it means that when you hear about massive data center growth, you need to think beyond just the total power needed and consider how that power actually moves across the messy, dynamic grid.
Conclusion: Rosa: So we're wrapping up this look at data center growth through two thousand thirty specifically focusing on that paper by Rosa and Dev on grid adequacy and transmission planning.
Dev: It really boils down to how we measure demand versus how supply actually works across the whole system.
Rosa: Right. The authors are pointing out that you can't just use a single number for average demand or peak demand because those things describe totally different loads.
Dev: Exactly, and they’re showing that national generating capacity totals don't tell us if we have a surplus because regional weather and transmission lines matter too much.
Rosa: It means we need to be really careful when planning how much power the grid needs to handle.
Dev: They argue that the way we determine required connection capacity isn't just about the biggest equipment nameplate capacity or average consumption figures.
Rosa: It’s about what the customer actually asks for, and how that fits into the network right now.
Dev: And they bring up this thing about dynamic behavior, like how power shifts after a fault clears, which we haven't fully modeled in connection studies yet.
Rosa: So it’s not just about building bigger lines; it’s about modeling how everything interacts when things go wrong.
Dev: This suggests that future planning needs to include those fault sequences and demand recovery scenarios to see what's actually feasible.
Rosa: It makes you wonder what kind of realistic future network studies we need to get this right.
University of Texas at Dallas
eess.SY, cs.SY
Submitted: 2026-11-10
Updated: 2026-11-10
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
Importance score: 77/100
The gist: The growth of large data centers raises questions about both supply adequacy and the suitability of load representations used in connection studies This paper critically reviews the U.S.
Key concepts
- Annual Average Demand vs. Peak Demand
- These are different ways to measure data center needs. Annual average demand is a steady yearly figure, while peak demand represents the highest simultaneous load expected at any time. Using one interchangeably leads to inaccurate planning decisions.
- Supply Adequacy
- Simply comparing total national generating capacity against total projected demand is insufficient for determining if there will be a surplus of power. Actual supply depends on many variables like weather, generator availability, fuel levels, and transmission limits in specific regions.
- Dynamic Behavior in Connection Studies
- Current connection studies often ignore what happens *after* a fault is cleared. The paper stresses that planning must model the sequence: initial fault, the resulting power transfer to backup sources, and the subsequent recovery of data center demand. This sequence determines true performance.
- Grid Import Basis
- The most accurate measure for required connection capacity is not just equipment size or average use, but rather the maximum amount of power a customer requests to import from the grid. This focus on actual import limits provides a more realistic planning basis than other metrics.
Terminology
Summary
The growth of large data centers raises questions about both supply adequacy and the suitability of load representations used in connection studies This paper critically reviews the U. p. 0 edu
Abstract— The growth of large data centers raises questions
about both supply adequacy and the suitability of load
representations used in connection studies. This paper critically
reviews the U.Where the paper says it
Demand Projections and Forecasting Discrepancies
Public forecasts indicate substantial growth in data center demand through 2030, but annual average demand, peak demand, and grid connection capacity describe different quantities and cannot be used interchangeably National generating capacity totals cannot establish a dependable surplus because resource availability, regional adequacy, transmission constraints, and project timing determine supply capability
** Annual average demand is estimated at approximately 74.1 GW for the reference case in 2030. p. 0 The
report treats 2024 as its last historical year because its
equipment-shipment data extend only through the third quarter
of 2025, so the 2025 results already include projections. For
2030, the reference case reaches 649 TWh, equivalent to
approximately 74.Where the paper says it
** Peak-demand forecasts project 45, 71, and 94 GW under low, medium, and high scenarios for EPRI. p. 1 and coincidence with other loads [1]. The Electric Power
Research Institute (EPRI) estimates aggregate data-center peak
demand of 21–22 GW in 2024 and projects 45, 71, and 94 GW
under its low, medium, and high 2030 scenarios, respectively
[3].Where the paper says it
** The market analyses reviewed by Grid Strategies suggest approximately 65 GW of additional datacenter demand through 2030. p. 1 Although conceptually
closer to EPRI’s peak-demand estimates, it requires a common
baseline, coverage, and aggregation method for a meaningful
comparison. However, market analyses reviewed by Grid
Strategies suggest approximately 65 GW of additional datacenter demand through 2030, compared with approximately 90
GW included in utility forecasts.Where the paper says it
Supply Adequacy and Regional Constraints
National generating capacity totals do not establish a dependable surplus because resource availability, regional adequacy, transmission constraints, and project timing determine supply capability
** Comparing the projected 1,452 GW of net summer capacity with approximately 995 GW of aggregate summer peak demand gives an apparent difference of 457 GW.**
** This difference does not establish a dependable surplus or a valid planning reserve margin because the figures originate from separate forecasting frameworks. p. 1 Comparing the projected 1,452 GW of net summer
capacity with approximately 995 GW of aggregate summer peak
demand gives an apparent difference of 457 GW [4, 6]. However, this difference does not establish a dependable surplus
or a valid planning reserve margin: the figures originate from
separate forecasting frameworks, and net summer capacity does
not represent the accredited contribution of all resources during
critical hours.Where the paper says it
** Dependable supply depends on weather, generator outages, fuel availability, hydro conditions, storage duration and recharge, and transmission deliverability. p. 1 However, this difference does not establish a dependable surplus
or a valid planning reserve margin: the figures originate from
separate forecasting frameworks, and net summer capacity does
not represent the accredited contribution of all resources during
critical hours. Dependable supply depends on weather, generator
outages, fuel availability, hydro conditions, storage duration and
recharge, and transmission deliverability.Where the paper says it
Transmission Planning Framework Limitations
The relevant basis for determining required connection capacity is the maximum grid import requested by the customer rather than equipment nameplate capacity or average consumption alone
** An observed utilization level does not establish an enduring demand limit.**
** Conventional large-load connection planning evaluates alternative supply schemes using a future-year network model representing the study area and its interaction with the wider system. p. 2 Otherwise, the connection must accommodate the
requested maximum demand, including infrequent high-demand
periods. Conventional large-load connection planning evaluates
alternative supply schemes using a future-year network model
representing the study area and its interaction with the wider
system.Where the paper says it
Dynamic Behavior in Connection Studies
Recent grid events demonstrate that clearing the initiating fault does not necessarily restore prefault data-center demand from the grid, highlighting the need to represent fault-induced transfers to backup supply and subsequent demand recovery in connection studies
** Acceptable post-contingency voltages alone cannot establish acceptable performance throughout this sequence.**
** Dynamic studies should therefore assess the initiating fault, the resulting data-center demand transfer, and the subsequent recovery as a connected sequence. p. 3 These effects are not exclusive to data centers,
but their scale and concentration make them particularly relevant
to connection planning. Dynamic studies should therefore assess
the initiating fault, the resulting data-center demand transfer, and
the subsequent recovery as a connected sequence [17, 18, 20].Where the paper says it
** The principal technical gap identified in the selected literature is the limited integration of data-center dynamic models into the screening and economic comparison of connection alternatives. p. 4 The principal
technical gap identified in the selected literature is the limited
integration of validated data-center dynamic models into the
screening and economic comparison of expansion alternatives. Addressing this gap requires realistic future-year network
studies that establish whether these behaviors change feasible
connection capacity, required reinforcement, or the preferred
investment.Where the paper says it
Conclusion on Future Research Needs
Addressing this gap requires realistic future-year network studies that establish whether these behaviors change feasible connection capacity, required reinforcement, or the preferred investment
** Relevant assessments should cover peak and light-load conditions, applicable fault-clearing and reclosing sequences, and correlated demand transfers and recovery across nearby facilities.**
** Research should quantify whether these behaviors change feasible connection capacity, required reinforcement, or preferred investment within conventional planning. p. 4 Where unacceptable performance
occurs, planners should evaluate network reinforcement and
permissible changes to facility protection or controls under
consistent import requirements and reliability obligations. Research should quantify whether these behaviors change
feasible connection capacity, required reinforcement, or
preferred investment within conventional planning.Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Fault response modeling in connection studies should incorporate
fault-induced demand transfers to backup supply and subsequent demand recovery
to accurately represent grid interactions during disturbances. This allows planning models to test whetherclearing the initiating fault does not necessarily restore prefault data-center demand from the grid.
-
Connection planning frameworks must be updated to assess how dynamic behavior affects investment decisions, as studies should
test whether validated data-center models change the technical acceptability or economic ranking of connection alternatives.
This ensures that planning considers howfault-induced demand transfers and subsequent reconnection affect the selected plans.
-
Load modeling for connection studies needs to move beyond prescribed demands by incorporating dynamic behaviors such as
repeated disconnection and reconnection, termed flapping
when rapid demand restoration occurs. This requires planning models to account for sequences like those wherestaggering the return of ten UPS blocks mitigates this behavior, while soft starting further smooths demand recovery.
Abstract
The growth of large data centers raises questions about both supply adequacy and the suitability of load representations used in connection studies. This paper critically reviews the U.S. demand, generation, and transmission outlook through 2030 and examines the implications for transmission expansion planning (TEP). Comparison of public forecasts highlights distinctions among annual average demand, peak demand, and grid connection capacity. National generating-capacity totals cannot establish a dependable surplus because resource availability, regional adequacy, transmission constraints, and project timing determine supply capability. Recent grid events and published dynamic studies demonstrate that faults can trigger substantial data-center transfers to backup supply, followed by delayed or coordinated demand recovery. These behaviors require appropriate representation within the conventional planning framework, alongside assessment of the requested maximum grid import. The selected literature reveals limited integration of validated data-center dynamic models into the technical screening and economic comparison of expansion alternatives. The review identifies the need to establish whether fault response and recovery change feasible connection capacity, required reinforcement, or the preferred transmission investment.
Sources
- Impact of Data Center Load on Co-Optimized Generation and Transmission Expansion Planning with Embedded Unit Commitment
- Enhancing Data Center Low-Voltage Ride-Through
Related papers
- Industrial Load Modeling and Optimization for Market-Based Interaction with Power Systems
- Exact and Approximate Convex Reformulation of Linear Stochastic Optimal Control with Chance Constraints
- Certified Set Convergence for Piecewise Affine Systems
- Symbolic Control for Stochastic Systems via Finite Parity Games
- Multiobjective optimization-based design and dispatch of islanded, hybrid microgrids for remote, off-grid communities in sub-Saharan Africa
- Physiology-informed layered sensing for intelligent human-exoskeleton interaction