One question.
Two explanations.

How should electricity law allocate scarce capacity and interconnection risk when AI data-center demand grows faster than new supply and transmission?

Internal source tag USLP-001

Authors

Alexandra B. Klass; Dave Owen

Publication

The George Washington Law Review, vol. 94, pp. 60-120

Year

2026

Publication status

PUBLISHED

Peer review

STUDENT_EDITED

Method

Doctrinal and comparative legal analysis. The article examines utility duty-to-serve premises and emerging large-load arrangements, then draws cautiously from federal natural-gas curtailment/capacity-release law and western prior-appropriation/water-market doctrine.

Finding recorded in corpus

The authors propose a contracts- and trading-based 'demand-side connect-and-manage' framework under which large loads can connect subject to priority, curtailment, and transfer rules; they contend this could accelerate connections and place more overbuild and innovation risk on large data-center companies.

Scope limit

United States federal electricity law; United States state utility law; western United States water law

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Can existing rate structures, special contracts, federal/state allocation gaps, and co-location arrangements shift data-center power costs to other ratepayers, and what should state institutions do?

Internal source tag USLP-003

Authors

Eliza Martin; Ari Peskoe

Publication

Harvard Electricity Law Initiative, Environmental & Energy Law Program

Year

2025

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Doctrinal and institutional analysis plus review of nearly 50 regulatory proceedings; the special-contract subsection reports a review of 40 state PUC proceedings.

Finding recorded in corpus

The authors report that, in their review of 40 state PUC proceedings involving data-center special contracts, regulators frequently approved the contracts in short, conclusory orders, limiting the public record supporting cost-isolation claims.

Scope limit

United States state public-utility regulation; United States federal transmission regulation

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What state legislative, executive, tariff, interconnection, and federal cost-allocation approaches were emerging for large electric loads through 2025?

Internal source tag USLP-004

Authors

Erik N. Anderson; Prathit Dave; Adrienne L. Rackley; Devyn W. Powell

Publication

Pacific Northwest National Laboratory, PNNL-38455

Year

2025

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Descriptive taxonomy built from 61 state bills/executive orders in 40 states proposed or enacted during 2013-2025 (43 enacted, 18 proposed), ten selected 2024-2025 utility rate cases in ten states, and relevant federal orders.

Finding recorded in corpus

Seven of ten selected rate cases sought a new data-center or large-load rate class; the utilities' stated rationale in most such cases was protecting other classes from infrastructure costs attributed to new large loads.

Scope limit

United States federal; 40 United States states

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Which large-load tariff elements are established, stable, emerging, or declining, and how do utilities allocate planning, operational, credit, and stranded-cost risk?

Internal source tag USLP-005

Authors

Natalie Mims Frick; Peter Cappers; Goksin Kavlak; Long Lam; Ryan Hledik; Jadon Grove; Sebastian Rotella

Publication

Lawrence Berkeley National Laboratory technical brief

Year

2026

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Codes a purposive sample of 55 tariffs derived from Halcyon's Large Load Tariff Tracker as of March 2026 and supplemental research through April 2026; consolidates duplicates and removes superseded/non-primary entries under stated exclusions.

Finding recorded in corpus

Among tariffs with standardized minimum terms, the range was one to twenty years and the sample median increased from five years for pre-2025 proposals to twelve years for proposals since 2025; the authors interpret longer terms as protection against cost-recovery and stranded-investment risk.

Scope limit

United States state and local retail tariffs; United States federal tariffs where applicable

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Under what conditions can commercial/industrial demand growth lower or raise retail electricity prices?

Internal source tag USLP-007

Authors

Eric O'Shaughnessy; Ryan H. Wiser; Peter Cappers; Jeffrey Deason; Natalie Mims Frick; Ryan Hledik

Publication

Joule, article 102528

Year

2026

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Conceptual economic synthesis and commentary drawing on historical and empirical literature; organizes the relationship around capacity utilization, system-expansion cost, and rate-design cost allocation. No original causal estimate.

Finding recorded in corpus

The authors argue the retail-price effect is not predetermined: higher utilization and low per-unit expansion costs can lower average prices if incremental costs are allocated to cost causers, while constrained or costly expansion and imperfect allocation can raise them.

Scope limit

United States electricity sector

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What was the causal effect of state-level data-center capacity growth on average residential retail electricity prices during 2015-2024?

Internal source tag USLP-008

Authors

Asa Watten; John Bistline; Geoffrey Blanford

Publication

arXiv:2606.19777v1

Year

2026

Publication status

PREPRINT_OR_WORKING_PAPER

Peer review

WORKING_PAPER

Method

State-year instrumental-variables design using length of the 1947 planned Interstate Highway System within each state, conditional on GDP and population; links S&P Global data-center capacity with FERC Form 1 cost data for 366 IOUs and EIA Form 861 price data. Preferred first-stage F-statistic is 7.8; Anderson-Rubin weak-IV inference is reported.

Finding recorded in corpus

The authors estimate that a 10% increase in data-center capacity was associated, under their IV design, with about a 0.4% decrease in average residential retail prices; they caution that future supply constraints could reverse the sign.

Scope limit

United States states and investor-owned utilities, 2015-2024

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How should the traditional utility obligation to serve change when uncertain data-centre load threatens resource adequacy and cost allocation?

Internal source tag USLP-501

Issued / announced

2026-10-14

Evidence status

PUBLISHED · independent policy report; not a binding rule or peer-reviewed journal article

Authors

Sarah Freeman; Louisa Eberle; Camille Kadoch

Publisher

Regulatory Assistance Project (RAP)

Method / lens

Normative and doctrinal regulatory policy analysis

Arguments

Four strategies: evaluate duty-to-serve impacts; clarify extension limits; strengthen interruptible service; define conditions for expedited connections

Contribution

Identifies choices available to state utility regulators, distinct from FERC's federal PJM tariff decision

Limitation

Normative policy proposals; not causal evidence, statute or commission order

Source record: approved research update · 2026-10-19 · primary/official source; original 2026-09 Excel corpus preserved

Which legal bottlenecks constrain carbon-free electricity supply during renewed demand growth, and what targeted reforms would an energy-abundance approach support?

Internal source tag USLP-002

Authors

Alexandra B. Klass; Matthew Appel

Publication

North Carolina Law Review, vol. 104, starting p. 63

Year

2025

Publication status

PUBLISHED

Peer review

STUDENT_EDITED

Method

Doctrinal and policy synthesis using legal authorities, project examples, electricity-market material, and an analogy to housing-abundance reforms.

Finding recorded in corpus

The authors define a clean-energy-only 'Law of Energy Abundance' and propose targeted changes to federal environmental review, state permitting, local siting, and high-voltage transmission governance.

Scope limit

United States federal law; United States state law; United States local land-use law

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Can geographically distributed Internet services route workload toward locations with temporarily cheaper electricity without unacceptable performance or bandwidth cost?

Internal source tag TE02

Authors

Asfandyar Qureshi; Rick Weber; Hari Balakrishnan; John V. Guttag; Bruce M. Maggs

Publication

ACM SIGCOMM Computer Communication Review 39(4), 123-134

Year

2009

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

trace-driven counterfactual simulation

Finding recorded in corpus

Under the paper's strict operational constraints, dynamic electricity-price-aware routing reduced modeled energy cost by at least about 2%; with more elastic constraints savings exceeded 30%, and removing distance constraints produced modeled savings up to about 45%.

Scope limit

electricity expenditure, not necessarily energy or emissions reduction

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How should a data center combine workload shifting and local generation to reduce coincident-peak charges under prediction uncertainty?

Internal source tag TE03

Authors

Zhenhua Liu; Adam Wierman; Yuan Chen; Benjamin Razon; Niangjun Chen

Publication

Performance Evaluation 70(10), 770-791

Year

2013

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

stochastic and robust optimization with trace-driven simulation

Finding recorded in corpus

In the modeled cases, combining workload shifting and local generation reduced energy cost by roughly 35-40% and emissions by roughly 10-15% relative to then-current practices; policy performance depended on forecast quality and robust design.

Scope limit

one utility/tariff setting

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Can a live AI GPU cluster follow utility-style multi-hour power-reduction profiles while completing heterogeneous AI jobs within declared SLAs?

Internal source tag TE14

Authors

Philip Colangelo; Ayse K. Coskun; Jack Megrue; Ciaran Roberts; Shayan Sengupta; Varun Sivaram; Ethan Tiao; Aroon Vijaykar; Chris Williams; Daniel C. Wilson; Zack MacFarland; Daniel Dreiling; Nathan Morey; Anuja Ratnayake; Baskar Vairamohan

Publication

Nature Energy 11, 254-261 (2026 issue)

Year

2025

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

field demonstration plus simulator validation and hardware power-performance sweeps

Finding recorded in corpus

The cluster delivered a 25% power reduction for three hours with 15-minute ramps and no snapback, while all 212 jobs completed within their declared experimental SLAs.

Scope limit

one 256-A100 cluster and researcher/partner-defined workloads and SLAs

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How would different shares and time horizons of data-center temporal flexibility affect least-cost generation/storage/transmission investment, dispatch, cost, and CO2 in three US regions?

Internal source tag TE17

Authors

Juan Ramon L. Senga; Shen Wang; Christopher R. Knittel

Publication

iScience 29(7), 116497

Year

2026

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

capacity-expansion and dispatch scenario modeling

Finding recorded in corpus

Flexibility reduced modeled system cost by up to about 5% in Texas, 4% in the Mid-Atlantic, and 2% in WECC, but emissions depended on the resource mix: Texas emissions fell by up to about 40%, whereas Mid-Atlantic flexibility increased coal utilization and could increase emissions.

Scope limit

three stylized US regions under specified 2030 scenarios

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What was US data-center electricity use through 2014, what trajectory was plausible to 2020, and what efficiency scenarios could change it?

Internal source tag TE05

Authors

Arman Shehabi; Sarah J. Smith; Dale Sartor; Richard Brown; Magnus Herrlin; Jonathan G. Koomey; Eric R. Masanet; Nathaniel Horner; Inês Lima Azevedo; William Lintner

Publication

Lawrence Berkeley National Laboratory, LBNL-1005775

Year

2016

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

bottom-up installed-base energy model with scenarios

Finding recorded in corpus

The report estimated US data centers used about 70 billion kWh in 2014 (1.8% of US electricity) and projected about 73 billion kWh in 2020 under then-current trends; efficiency scenarios reduced the modeled 2020 level by as much as 33 billion kWh relative to that trajectory.

Scope limit

all US data centers, not AI-specific

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How much electricity did US data centers use through 2023 and what range could result by 2028 under uncertain accelerated-server growth?

Internal source tag TE11

Authors

Arman Shehabi; Sarah Josephine Smith; Alex Hubbard; Alexander Newkirk; Nuoa Lei; Md AbuBakar Siddik; Billie Holecek; Jonathan G. Koomey; Eric R. Masanet; Dale A. Sartor

Publication

Lawrence Berkeley National Laboratory, LBNL-2001637

Year

2024

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

bottom-up historical estimate and scenario projection

Finding recorded in corpus

The report estimated US data centers used about 176 TWh in 2023 (4.4% of US electricity) and modeled a 2028 range of 325-580 TWh (6.7-12.0%), equivalent to roughly 74-132 GW at the report's assumed 50% capacity utilization.

Scope limit

US national total; 2028 range is conditional, not observed

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What electricity consumption, generation mix, and attributable CO2e emissions are associated with U.S. data centers at balancing-authority and state scales?

Internal source tag USLP-011

Authors

Gianluca Guidi; Francesca Dominici; Jonathan Gilmour; Kevin Butler; Eric Bell; Scott Delaney; Falco J. Bargagli-Stoffi

Publication

arXiv:2411.09786v1

Year

2024

Publication status

PREPRINT_OR_WORKING_PAPER

Peer review

WORKING_PAPER

Method

Facility dataset of 2,132 data centers (authors estimate 78% coverage), combining public web/satellite validation and proprietary Baxtel data; 337 missing capacity values imputed with gradient boosting (R-squared 0.77); assumes uniform 0.75 utilization; attributes annual balancing-authority generation shares using EPA eGRID plants.

Finding recorded in corpus

Under the authors' 75% utilization and attributional generation-weighting assumptions, the sample is estimated at 192.64 TWh and 105.59 million metric tons CO2e; its energy-weighted carbon intensity is estimated at 548 gCO2e/kWh, 48% above the national average used by the paper.

Scope limit

contiguous United States, September 2023-August 2024

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How do facility types and siting/operational demands create local energy, water, land, fiscal, and nuisance effects, and why do existing local/state governance systems struggle to oversee them?

Internal source tag USLP-014

Authors

Corey Young

Publication

Oxford Open Energy, vol. 5, oiag006

Year

2026

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

Integrative interdisciplinary literature review drawing on technical, environmental, economic, planning, infrastructure, and governance literature; constructs an infrastructure-centered conceptual framework.

Finding recorded in corpus

The review identifies rapid industry change, information asymmetry/operational opacity, fragmented authority, and limited municipal capacity as recurrent oversight problems, and reframes data centers as privately owned critical infrastructure with public place-based obligations.

Scope limit

United States local and state governance, with international examples

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Where and how large were US data centers' operational carbon and water footprints, including indirect electricity-sector water use?

Internal source tag TE07

Authors

Md Abu Bakar Siddik; Arman Shehabi; Landon Marston

Publication

Environmental Research Letters 16, 064017

Year

2021

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

spatial bottom-up footprint accounting

Finding recorded in corpus

For 2018 the study estimated about 5.13e8 m3 of operational water consumption, roughly three quarters indirect through electricity, and about 3.15e7 tonnes CO2e; about one fifth of direct water footprint occurred in moderately to highly stressed watersheds, while location created carbon-water tradeoffs.

Scope limit

operational footprint only

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Which server, workload, cooling, climate, and electricity factors determine total water consumed per unit of data-center work?

Internal source tag TE13

Authors

Nuoa Lei; Jun Lu; Arman Shehabi; Eric R. Masanet

Publication

Resources, Conservation and Recycling 219, 108310

Year

2025

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

literature review, parametric model, and global sensitivity analysis

Finding recorded in corpus

Modeled water per workload varied by more than four orders of magnitude; total-effect sensitivity ranked server efficiency first, grid water-consumption factor second, and utilization third, so no single cooling or siting intervention dominates all cases.

Scope limit

server workload archetype, not measured national AI water use

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What institutional and technical-policy options could reduce large-load connection bottlenecks across forecasting, interconnection, procurement, operations, and ratemaking?

Internal source tag USLP-006

Authors

Fredrich Kahrl; Natalie Mims Frick

Publication

Lawrence Berkeley National Laboratory report

Year

2026

Publication status

PUBLISHED

Peer review

EDITORIALLY_REVIEWED

Method

Extensive document and literature review identifying 41 illustrative solutions in 13 solution areas grouped into five functions; includes federal, state, ISO/RTO, utility, and tariff examples.

Finding recorded in corpus

The review identifies 41 potential solutions and expressly presents them as an illustrative menu, not recommendations or a roadmap; some are region-specific or mutually exclusive and federal-state overlap is especially difficult.

Scope limit

United States federal and state electricity systems

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How can distribution feeders and substations serve rapidly growing data-center/large-load requests, and where do timelines, redundancy requirements, and planning tools create constraints?

Internal source tag USLP-012

Authors

Gayathri Krishnamoorthy; Vincent Westfallen; Killian McKenna

Publication

National Laboratory of the Rockies, NLR/TP-6A40-96783

Year

2026

Publication status

PUBLISHED

Peer review

EDITORIALLY_REVIEWED

Method

Combines a Baxtel dataset of more than 3,600 operational, under-construction, and proposed sites with national datasets, utility filings, and industry studies; classifies facilities by ultimate demand and develops infrastructure buildout estimates.

Finding recorded in corpus

The report estimates data-center development typically unfolds in 12-24 months while utility reinforcement often takes 3-5 years, creating a need to evaluate phased energization, temporary feeders, onsite supply, and other bridging arrangements while permanent facilities are built.

Scope limit

United States distribution systems

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How should U.S. electricity governance adapt grid-access, interconnection, transmission, rate, co-location, and participation rules to large-load growth while protecting reliability and other consumers?

Internal source tag USLP-015

Authors

Hannah Wiseman; Matthew McHale

Publication

Harvard Environmental Law Review, vol. 50, pp. 108-191

Year

2026

Publication status

PUBLISHED

Peer review

STUDENT_EDITED

Method

Doctrinal and institutional policy synthesis using statutes, FERC orders and conference materials, state commission dockets and utility tariffs/settlements, technical reports, and historical context. It builds a proposal bundle rather than a systematic docket sample or causal test.

Finding recorded in corpus

The authors propose a coordinated governance package: centralized and de-duplicated large-load data, system modeling, federal and state interconnection standards, build-readiness priorities, explicit treatment of grid-connected and off-grid co-location/backup, fixed and exit fees tied to differential costs, demand-response incentives, and broader public participation.

Scope limit

United States federal, regional, state, and utility large-load governance

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How much constant new curtailable load could fit below historical seasonal system peaks across major US balancing authorities?

Internal source tag TE12

Authors

Tyler H. Norris; Tim Profeta; Dalia Patiño-Echeverri; Adam Cowie-Haskell

Publication

Nicholas Institute for Energy, Environment & Sustainability, Duke University

Year

2025

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

first-order historical load-duration/headroom calculation

Finding recorded in corpus

Across the 22 balancing authorities the first-order calculation found 76 GW of headroom at 0.25% annual energy curtailment, 98 GW at 0.5%, 126 GW at 1%, and 215 GW at 5%; at 0.25% the mean year had 85 hours with some curtailment, not merely 22 fully offline hours.

Scope limit

aggregate generation headroom proxy, not interconnection feasibility

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How much distribution hosting capacity can infrequent utility-controlled curtailment or delay unlock for large loads?

Internal source tag TE15

Authors

Nan Gu; Ge Chen; Junjie Qin

Publication

Proceedings of the 59th Hawaii International Conference on System Sciences (HICSS)

Year

2026

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

nonconvex flexibility-aware hosting-capacity optimization, analytical propositions, and test-feeder simulation

Finding recorded in corpus

In the copperplate ResStock example, intervening in 1% of annual intervals increased modeled hosting capacity from 0.88 MW to 1.56 MW (77.55%); most of 350 delay interventions required no more than 90 minutes, although one required 210 minutes, and the qualitative findings persisted on the IEEE 123-bus feeder.

Scope limit

specified feeder/load cases, not measured interconnection times

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How do temporal deferral and geographic workload shifting alter generation expansion, dispatch cost, and congestion for AI data-center interconnection?

Internal source tag TE16

Authors

Yize Chen; Xiaogui Zheng

Publication

Proceedings of the 2026 ACM Sustainability Week, 322-327

Year

2026

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

deterministic two-stage capacity-expansion optimization with hourly DC power flow

Finding recorded in corpus

In modeled cases, flexibility reduced total grid investment/operating costs by roughly 3-21%, but did not always reduce added generation: one congested case cut total cost 3.5% while investment rose 7.1%; temporal-deferral benefits largely plateaued at a three-hour window.

Scope limit

test-system scenarios, not actual utility investment outcomes

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How much additional AI data-centre IT load can be interconnected when network reconfiguration is coordinated with workload shifting and UPS constraints?

Internal source tag USLP-502

Issued / announced

2026-10-04

Evidence status

PREPRINT · model results; not verified operational deployment

Authors

Rida Fatima; Xingpeng Li

Repository

arXiv:2610.05502; submitted 2026-10-04

Method

Day-ahead mixed-integer second-order-cone optimisation; simulated 116-bus test feeder derived from IEEE 123-node model

Reported result

Coordination raised feasible IT capacity in the studied constrained bus case from ~10.5–10.6 MW to ~15.3–16.0 MW; model- and location-specific

Contribution

Separates utility-side network flexibility from data-centre workload/UPS-side flexibility

Limitation

Model scenario, not a field-validated measured effect; results cannot be extrapolated to a real grid without further validation

Source record: approved research update · 2026-10-19 · primary/official source; original 2026-09 Excel corpus preserved

Why do U.S. institutions often treat clean-energy deployment and grid reliability as conflicting, and what calibrated legal/governance reforms could integrate them?

Internal source tag USLP-009

Authors

Alexandra Klass; Joshua Macey; Shelley Welton; Hannah Wiseman

Publication

Stanford Law Review, vol. 74, pp. 969-1071

Year

2022

Publication status

PUBLISHED

Peer review

STUDENT_EDITED

Method

Doctrinal and institutional legal synthesis. The authors analyze reliability and clean-energy rules across three forms of fragmentation—substantive responsibilities, federal/regional/state/local jurisdiction, and the public-private continuum—and develop reforms for markets, transmission, reliability regulation, and regional governance.

Finding recorded in corpus

The authors argue that much of the perceived clean-energy/reliability conflict is produced by fragmented law and governance, and propose calibrated reallocations of authority and process rather than eliminating all institutional silos.

Scope limit

United States federal, regional, state, and local electricity governance

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Can the traditional utility duty to serve support a negligence-based private-law duty to adapt utility operations and planning to foreseeable climate risks, and how should that duty interact with regulation and tariff defenses?

Internal source tag USLP-013

Authors

Jim Rossi; Michael Panfil

Publication

North Carolina Law Review, vol. 100, pp. 1135-1208

Year

2022

Publication status

PUBLISHED

Peer review

STUDENT_EDITED

Method

Historical, doctrinal, evidentiary, and normative legal analysis of duty-to-serve cases and sources, tort elements and defenses, utility tariffs and regulation, and extreme-weather outage/wildfire examples. It does not estimate causal effects or litigation prevalence.

Finding recorded in corpus

The authors argue that three historically recognized duty-to-serve obligations—reasonable interruption notice, adequate and safe service, and compliance with technical operational standards—can support a negligence-based duty to take reasonable action against foreseeable climate risks; they do not propose strict liability or utility insurance against all harms.

Scope limit

United States public-utility common law and regulation

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How much capacity-accreditable reduction could AI inference fleets deliver by combining spatial migration, temporal deferral, and DVFS?

Internal source tag TE18

Authors

Chris Dunlap

Publication

Research Square preprint, version 1

Year

2026

Publication status

PREPRINT_OR_WORKING_PAPER

Peer review

WORKING_PAPER

Method

eleven-parameter engineering framework, empirical stress decorrelation, trace analysis, and conditional Monte Carlo

Finding recorded in corpus

The framework estimated central commitment depth of 24.6% for a mixed-use single facility and 40.0% for an inference-dominant facility, declining to 24.1% and 38.5% at 10 GW; lower-tail values decline more sharply as destination contention grows.

Scope limit

conditional resource-adequacy metric, not accredited market capacity

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How is the physical AI infrastructure buildout financed, and how do ownership separation, leases, guarantees, project finance, and structured claims redistribute risk?

Internal source tag USLP-010

Authors

Stijn Van Nieuwerburgh

Publication

Columbia Business School working paper

Year

2026

Publication status

PREPRINT_OR_WORKING_PAPER

Peer review

WORKING_PAPER

Method

Conceptual/descriptive finance synthesis using corporate and industry materials and an illustrative public transaction case study (Meta/Blue Owl Hyperion-Beignet); not a systematic project sample or causal design.

Finding recorded in corpus

In the Hyperion illustration, five successive four-year lease options plus a residual-value guarantee give Meta exit optionality while protecting lenders against sale-value shortfall; the author argues that this shifts and obscures risk rather than eliminating it.

Scope limit

United States AI/data-center finance with global supply-chain context

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How does inference energy and carbon vary across ML tasks, models, modalities, and task-specific versus multi-purpose architectures?

Internal source tag TE10

Authors

Alexandra Sasha Luccioni; Yacine Jernite; Emma Strubell

Publication

ACM Conference on Fairness, Accountability, and Transparency (FAccT '24), 21 pages

Year

2024

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

controlled benchmark experiment

Finding recorded in corpus

Across the tested task-specific models, mean energy per 1,000 inferences ranged from about 0.002 kWh for text classification to 2.907 kWh for image generation, a greater-than-1,450-fold task spread; multi-purpose generative models were generally more energy-intensive than task-specific models on comparable discriminative tasks.

Scope limit

selected open models, unbatched sequential inference, one hardware/cloud region

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What jurisprudential bases and institutional limits justify government provision of data, computing power and algorithms as distinct AI factors?

Internal source tag CHN-LAW-002

Year

2025

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

jurisprudential and doctrinal analysis

Finding recorded in corpus

Because data, computing power and algorithms have different public-good and market properties, government provision should be legally differentiated; for baseline computing power the paper sees a rationale for universal, low-cost supply but warns that direct public provision of high-performance capacity can create fiscal and rent-seeking risks.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What carbon-reduction effect results from moving data-center workload among the eight EDWC hubs after accounting for climate, renewable electricity and transmission?

Internal source tag CHN-ENV-009

Year

2023

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

scenario-based carbon accounting

Finding recorded in corpus

In the model, the additional emissions from data transmission are small relative to renewable-energy and climate-related reductions; one kWh of shifted workload reduces 0.053-0.344 kg CO2 across modeled routes when the first two effects are included.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How much could EDWC and additional PUE/green-electricity improvements reduce data-center emissions, and which configurations are associated with lower regional carbon intensity?

Internal source tag CHN-ENV-010

Authors

Xuemei Xie; Yuhang Han; Hao Tan

Publication

Humanities and Social Sciences Communications

Year

2024

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

LEAP scenario modeling plus fuzzy-set qualitative comparative analysis

Finding recorded in corpus

The model forecasts 2,125 Mt cumulative CO2 reduction in the project scenario during 2020-2050 and about 9,500 Mt in the advanced scenario; these are conditional forecasts, not observed project effects.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What emission-reduction and economic effects could EDWC generate by 2030, and what implementation barriers condition those effects?

Internal source tag CHN-ENV-011

Authors

Ning Zhang; Huabo Duan; Yuru Guan; Ruichang Mao; Guanghan Song; Jiakuan Yang; Yuli Shan

Publication

Engineering

Year

2025

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

scenario forecast and multi-regional input-output analysis

Finding recorded in corpus

The paper forecasts a 16-20% data-center emissions reduction by 2030, about USD 53 billion in direct electricity-cost savings, and USD 117-172 billion in indirect economic gains; these are modeled projections.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Do three representative east-west bits-migration routes reduce energy, carbon and cost after cooling, data transmission, power losses and duplicate western capital expenditure are considered?

Internal source tag CHN-ENV-012

Authors

Yingbo Zhang; Hangxin Li; Shengwei Wang

Publication

Applied Energy

Year

2025

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

route-specific techno-economic and carbon modeling

Finding recorded in corpus

Results are heterogeneous: Shanghai-Sichuan shows a large modeled carbon reduction, while Beijing-Inner Mongolia increases emissions; duplicate western capital expenditure eliminates economic benefit in the base assumptions.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Can cross-regional computing transfer reduce national data-center emissions while shifting embodied emissions and inequality from east to west?

Internal source tag CHN-ENV-013

Authors

Kaile Zhou; Ziwei Yang; Xinhui Lu; Rong Hu

Publication

Humanities and Social Sciences Communications

Year

2026

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

bottom-up scenario forecasting and attribution analysis

Finding recorded in corpus

The model forecasts large aggregate reduction by 2060 but also an east-to-west embodied carbon flow, demonstrating that national decarbonization and regional distributional burden can move in opposite directions.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How should data-center workloads and component-level energy consumption be modeled so that data centers can be coordinated with power systems?

Internal source tag CHN-ENG-004

Year

2022

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

structured technical review and model taxonomy

Finding recorded in corpus

Data-center load is not homogeneous: delay-tolerant and interactive workloads have different service constraints, and both workload processing and equipment efficiency must be represented to estimate usable flexibility.

Scope limit

China; general technical literature

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How much operating cost and energy use can a multi-data-center operator reduce by jointly scheduling workloads across time and space while preserving response-time and user-satisfaction constraints?

Internal source tag CHN-ENG-006

Year

2025

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

mathematical optimization and simulation

Finding recorded in corpus

In the authors' simulation, dynamic server configuration lowers total electricity consumption by 13.9% and purchase cost by 11.1%; adding workload transfer yields additional modeled savings, with response-specific spatiotemporal scheduling producing the largest reported cost reduction.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How can data-center energy performance be compared fairly across climates when raw PUE is climate-sensitive?

Internal source tag CHN-MEAS-014

Authors

Jian Li; Jakub Jurasz; Hailong Li; Wen-Quan Tao; Yuanyuan Duan; Jinyue Yan

Publication

Applied Energy

Year

2020

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

indicator construction, engineering simulation and case comparison

Finding recorded in corpus

COPUE is proposed to remove much of the climate effect from cross-location comparison: values above one indicate performance worse than the local benchmark and lower values indicate better relative design/operation.

Scope limit

multi-country; includes China-relevant siting measurement

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Can a combined physical and statistical framework predict location-specific PUE, total energy, CUE and electricity cost before construction?

Internal source tag CHN-MEAS-015

Authors

Yiliu Zhang; Jie Liu

Publication

Sensors

Year

2022

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

thermodynamic-statistical model with Sobol sensitivity and external validation

Finding recorded in corpus

Climate is a major determinant of predicted PUE for free-cooling designs; colder locations generally have lower annual PUE, but electricity price can reverse the cost ranking.

Scope limit

China; multi-country validation

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations. | PRE-QA CURATION NOTE: Retained as useful measurement background but moved from Core to Context because its location-energy prediction function overlaps stronger climate-normalization, field, and system studies.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Why did two similar Shanghai data centers have materially different PUE in 2017, and which operating changes can improve efficiency?

Internal source tag CHN-FIELD-017

Year

2019

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

comparative facility case study and operational test

Finding recorded in corpus

A's annual PUE was 1.310 versus B's 1.545; the paper attributes the difference to cooling equipment/control, 240 V DC/direct supply versus UPS and row-level versus room-level air delivery, and reports a one-week test where lowering chilled-water output from 9°C to 8°C increased source electricity 6%.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What is the observed energy performance of existing data-center cooling systems across seven climates, and what retrofit potential do evaporative/mechanical hybrid systems offer?

Internal source tag CHN-FIELD-016

Authors

Tiezhu Sun; Xiaojun Huang; Caihang Liang; Riming Liu; Yongcheng Yan

Publication

Sustainability

Year

2023

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

internship/field survey plus climate-based retrofit calculation

Finding recorded in corpus

Surveyed sites predominantly used mechanical cooling and mostly had PUE above 1.5; modeled hybrid cooling retrofits reduce PUE below 1.36 or 1.39, with estimated energy-saving rates from 41.23% to 73.26% and greater potential in drier climates.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations. | PRE-QA CURATION NOTE: Retained as observed/model hybrid background but moved from Core to Context because its selected seven-city cooling sample and retrofit model are narrower than the main governance questions.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How should green direct connection, PPAs, grid supply and workload time/space flexibility be jointly configured to minimize cost and emissions?

Internal source tag CHN-ENG-008

Year

2026

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

multi-scenario optimization

Finding recorded in corpus

The modeled cost response to the direct-green-power ratio is U-shaped under an 80% green-consumption constraint, and combined temporal and spatial flexibility improves the cost-emissions tradeoff relative to less flexible configurations.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How should Chinese law respond to the development, security, competition and resource-allocation problems of the AI computing-power industry?

Internal source tag CHN-LAW-001

Year

2025

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

doctrinal and conceptual legal analysis

Finding recorded in corpus

The article argues that computing-power regulation must simultaneously promote innovative capacity and safe supply while controlling AI, security and competition risks; it proposes coordinated rules for finance and energy support, supply-chain security, antitrust, standards and interoperability.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What institutional problems obstruct an integrated national computing-power system and what governance responses are required?

Internal source tag CHN-LAWPOL-003

Year

2024

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

policy and legal institutional analysis

Finding recorded in corpus

The authors identify five linked deficits: heterogeneous-resource imbalance, regional gap, lagging legal norms, an incomplete compute-electricity coordination mechanism and inadequate integration of security with development.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What opportunities and barriers arise when data-center temporal and spatial flexibility is coordinated with the electricity system?

Internal source tag CHN-ENG-005

Year

2022

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

structured review and research agenda

Finding recorded in corpus

Technical flexibility can support renewable integration and power-system operation, but the paper explicitly warns that usable flexibility depends on market incentives, business feasibility and regional coordination rather than engineering potential alone.

Scope limit

China; general technical literature

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How can adjustable resources in power and computing nodes be scheduled bidirectionally to reduce joint operating costs?

Internal source tag CHN-ENG-007

Year

2025

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

bilevel/two-stage optimization simulation

Finding recorded in corpus

For the modeled case, compute-side savings equal 4.71% of power-node operating cost and the compute node's daily operating cost falls about 0.70%.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How do Ulanqab and Qingyang differ in green-power supply, efficiency, business model, industrial ecosystem and policy innovation, and under what conditions are their models replicable?

Internal source tag CHN-POL-018

Year

2026

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

structured paired-case comparison

Finding recorded in corpus

The article characterizes Ulanqab as a resource-adaptive model centered on water-saving cooling and lifecycle efficiency, and Qingyang as an institution-innovative model centered on aggregated green-power direct supply.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Can electricity savings enabled by AI applications in China’s source-grid-load-storage system offset the electricity consumed by AI servers, and how does the balance depend on deployment efficiency?

Internal source tag CHN-TECH-019

Authors

Kaile Zhou; Ziwei Yang; Rong Hu

Publication

Communications Sustainability

Year

2026

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

meta-analysis combined with bottom-up stock-flow and 3×3 scenario modeling

Finding recorded in corpus

Only the three Efficient combinations reach net electricity saving by 2060; the best, Efficient–Technology Driven, reaches approximately 130.3% by 2040. The result is highly conditional on AI penetration, chip efficiency and PUE, and projected savings vary substantially by power-system segment and province.

Scope limit

China

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Why does technical compute flexibility fail to become a scalable power-system resource?

Internal source tag RCX-001

Authors

Aijia Luo; Nianfeng Tian; Qinglai Guo

Publication

Energy Internet 3(2), 81–85

Year

2026

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Conceptual perspective using prior studies and industry examples; no original causal estimate or representative contract sample.

Finding recorded in corpus

The authors identify bundled electricity costs and misaligned incentives, and propose connecting price signals, contractual allocation and operational scheduling.

Scope limit

China; international computing-service arrangements

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What forms of data-center flexibility exist, why had commercial DR participation remained limited, and what research/market-design barriers matter?

Internal source tag TE04

Authors

Adam Wierman; Zhenhua Liu; Iris Liu; Hamed Mohsenian-Rad

Publication

2014 International Green Computing Conference, 1-10

Year

2014

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

interdisciplinary narrative survey and conceptual synthesis

Finding recorded in corpus

Data centers possess temporal, service-quality, and geographic flexibility, but the paper found little actual demand-response participation and argued that prevailing programs were poorly matched to data-center load profiles and risk tolerance.

Scope limit

state of knowledge in 2014

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Can day-ahead carbon forecasts and virtual capacity limits shift flexible compute within Google data centers while preserving daily service capacity?

Internal source tag TE08

Authors

Ana Radovanović; Ross Koningstein; Ian Schneider; Bokan Chen; Alexandre Duarte; Binz Roy; Diyue Xiao; Maya Haridasan; Patrick Hung; Nick Care; Saurav Talukdar; Eric Mullen; Kendal Smith; MariEllen Cottman; Walfredo Cirne

Publication

IEEE Transactions on Power Systems 38(2), 1270-1280 (issue 2023; online 2022)

Year

2022

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

production system design, forecasting/optimization, and randomized cluster-day evaluation

Finding recorded in corpus

The deployed system shifted flexible batch capacity away from higher-carbon hours; in the randomized evaluation, average cluster power in the highest-carbon hours fell roughly 1-2%, while aggressive shaping could reduce completed daily compute or induce uncontrolled spatial shifts.

Scope limit

Google fleet and designated flexible batch workloads

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How much electricity did data-center IT equipment use worldwide in 2000 and 2005, and what drove the change?

Internal source tag TE01

Authors

Jonathan G. Koomey

Publication

Environmental Research Letters 3, 034008

Year

2008

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

bottom-up stock-and-power accounting

Finding recorded in corpus

Estimated aggregate data-center electricity use doubled from 2000 to 2005; direct IT equipment represented about 0.5% of world electricity in 2005 and about 1% when cooling and power-distribution overhead were included.

Scope limit

worldwide data centers, not AI-specific

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations. | PRE-QA CURATION NOTE: Retained as the foundational 2000–2005 baseline but moved from Anchor to Context during cross-corpus prioritization because it predates hyperscale cloud and accelerator-heavy AI.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Did rapid growth in computation translate into equally rapid growth in global data-center electricity between 2010 and 2018?

Internal source tag TE06

Authors

Eric R. Masanet; Arman Shehabi; Nuoa Lei; Sarah Smith; Jonathan G. Koomey

Publication

Science 367(6481), 984-986

Year

2020

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

bottom-up global stock-and-efficiency recalibration

Finding recorded in corpus

The model estimated global data-center electricity at 194 TWh in 2010 and 205 TWh in 2018, about a 6% rise, while compute instances increased about 550%; efficiency and cloud/hyperscale shifts decoupled compute growth from energy over that period.

Scope limit

2010-2018 global data centers

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What order of magnitude of electricity could newly shipped AI servers consume if contemporary AI adoption and hardware supply trends continued?

Internal source tag TE09

Authors

Alex de Vries

Publication

Joule 7(10), 2191-2194

Year

2023

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

hardware-shipment scenario calculation

Finding recorded in corpus

If NVIDIA shipped about 1.5 million AI servers annually by 2027 and that annual cohort ran at full capacity, the paper estimated 85.4-134 TWh of annual electricity; the figure is a conditional hardware scenario, not an observed sector forecast with probability bounds.

Scope limit

one vendor's projected annual shipments; excludes facility overhead in the principal calculation

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Which policy and contract mechanisms can integrate growing data-centre demand into electricity systems without shifting costs unfairly?

Internal source tag GLB-501

Issued / announced

2026-10-14

Evidence status

PUBLISHED IEA REPORT · descriptive policy scan, not binding law

Publisher

International Energy Agency (IEA), at request of France's G7 Presidency

Coverage

More than 34 policy jurisdictions covering approximately 90% of global data-centre capacity

Regulatory sample

134 identified policies and regulations; nearly 40% announced or implemented during the previous 12 months

PPA data

45 TWh worth of PPAs signed in first half 2026; almost 70% of that volume in the United States

Data interpretation

45 TWh refers to energy under SIGNED PPAs, not metered facility consumption; 90% refers to capacity coverage

Analytical contribution

Comparative principles for demand visibility, credible connection queues, non-firm access, fair tariff/risk allocation and grid flexibility

Limitation

Aggregated global policy inventory; jurisdictions' instruments differ in legal force and observability

Source record: approved research update · 2026-10-19 · primary/official source; original 2026-09 Excel corpus preserved

How could regulators condition AI data-centre interconnection on auditable public-value obligations rather than treating a facility as an unconditional passive load?

Internal source tag TE19

Authors

Shaikh Afnan Birahim

Publication

npj Environmental Social Sciences 1, 8

Year

2026

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

normative governance perspective synthesizing cited engineering, procurement, storage, heat-recovery, and environmental-governance literature

Finding recorded in corpus

The article proposes making grid access conditional on three measurable obligations: staged hourly-matched firm clean supply, certified fast-frequency/regulation services with telemetry and fail-safe guardrails, and waste-heat recovery where locally feasible. Its numerical KPIs are proposed policy targets, not findings prescribed by the cited feasibility studies.

Scope limit

proposed governance test, not an observed interconnection program or demonstrated net lifecycle benefit

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What can a reconstructed functional method legitimately explain and evaluate?

Internal source tag M01

Authors

Ralf Michaels

Publication

The Oxford Handbook of Comparative Law, pp. 339–382

Year

2006

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Conceptual reconstruction across comparative law and social-science functionalism; no effect-estimation dataset.

Finding recorded in corpus

Functional equivalence can organize context-sensitive comparison, but does not itself supply a criterion for declaring one legal order better.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How can qualitative research make disciplined descriptive and causal inferences?

Internal source tag M02

Authors

Gary King; Robert O. Keohane; Sidney Verba

Publication

Princeton University Press; chapter 1, pp. 3–33

Year

1994

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Methodological argument and illustrative examples; no original infrastructure dataset.

Finding recorded in corpus

The opening chapter links theory to observable implications and calls for explicit uncertainty and consideration of rival hypotheses.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Which research goals explain divergent quantitative and qualitative practices?

Internal source tag M03

Authors

James Mahoney; Gary Goertz

Publication

Political Analysis 14(3), 227–249

Year

2006

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

Stylized conceptual comparison across ten methodological dimensions.

Finding recorded in corpus

Case-oriented outcome explanation and population-oriented effect estimation can require different selection, measurement and generalization practices.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How should legal researchers design, code, analyze and present quantitative empirical work?

Internal source tag M04

Authors

Lee Epstein; Andrew D. Martin

Publication

The Oxford Handbook of Empirical Legal Research, chapter 38, pp. 901–925

Year

2010

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Methods tutorial illustrated with observational and experimental legal research.

Finding recorded in corpus

Coding rules and transparent documentation are integral to empirical inference; observational legal data require attention to confounding and the data-generation process.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

Under which conditions is international law produced and consequential?

Internal source tag M05

Authors

Gregory Shaffer; Tom Ginsburg

Publication

American Journal of International Law 106(1), 1–46

Year

2012

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Review and mid-level theoretical synthesis of empirical international-law scholarship.

Finding recorded in corpus

The authors advocate conditional theories of legal production and effects rather than a general yes-or-no debate over whether law matters.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How can differing review processes be described consistently?

Internal source tag M06

Authors

Maria J. Grant; Andrew Booth

Publication

Health Information and Libraries Journal 26, 91–108

Year

2009

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

Scoping searches and vocabulary analysis organized by Search, Appraisal, Synthesis and Analysis (SALSA).

Finding recorded in corpus

Review labels overlap; the actual search, appraisal, synthesis and analysis processes provide a more useful basis for understanding a review.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What should systematic reviewers report so readers can reconstruct selection and evaluate findings?

Internal source tag M07

Authors

Matthew J. Page; Joanne E. McKenzie; Patrick M. Bossuyt; Isabelle Boutron; Tammy C. Hoffmann; Cynthia D. Mulrow; Larissa Shamseer; Jennifer M. Tetzlaff; Elie A. Akl; Sue E. Brennan; Roger Chou; Julie Glanville; Jeremy M. Grimshaw; Asbjørn Hróbjartsson; Manoj M. Lalu; Tianjing Li; Elizabeth W. Loder; Evan Mayo-Wilson; Steve McDonald; Luke A. McGuinness; Lesley A. Stewart; James Thomas; Andrea C. Tricco; Vivian A. Welch; Penny Whiting; David Moher

Publication

BMJ 372, n71

Year

2021

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

Guideline development through evidence review, survey, expert meeting and iterative revisions.

Finding recorded in corpus

The reporting framework asks reviewers to disclose selection, methods, limitations and evidence supporting synthesis; transparent reporting assists appraisal without establishing validity by itself.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How should research metrics be used responsibly?

Internal source tag M08

Authors

Diana Hicks; Paul Wouters; Ludo Waltman; Sarah de Rijcke; Ismael Rafols

Publication

Nature 520, 429–431

Year

2015

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Normative manifesto; only publisher metadata/preview and author project overview examined.

Finding recorded in corpus

The author project overview warns against replacing informed judgment with metrics in research evaluation.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How can systematic coding improve empirical study of judicial opinions?

Internal source tag M09

Authors

Mark A. Hall; Ronald F. Wright

Publication

California Law Review 96(1), 63–122

Year

2008

Publication status

PUBLISHED

Peer review

STUDENT_EDITED

Method

Methodological synthesis and appraisal of legal content-analysis practices.

Finding recorded in corpus

Predetermined case selection, explicit coding and variable-level reliability assessment discipline content analysis; consensus among coders does not itself establish reproducibility.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What enables replication of quantitative empirical legal research and which reforms are needed?

Internal source tag M10

Authors

Jason M. Chin; Kathryn Zeiler

Publication

Annual Review of Law and Social Science 17, 239–260

Year

2021

Publication status

PUBLISHED

Peer review

PEER_REVIEWED

Method

Review of replication efforts, journal policies and institutional incentives.

Finding recorded in corpus

Accessible data, methods and code are necessary to attempt verification; the authors distinguish replication from extension and caution that the prevalence of legal-research problems remains incompletely measured.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How can researchers investigate within-case causal mechanisms?

Internal source tag M11

Authors

Derek Beach; Rasmus Brun Pedersen

Publication

University of Michigan Press

Year

2019

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Publisher description of a methodological monograph; book not obtained.

Finding recorded in corpus

Not established in the source record.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

What methods and substantive domains make up empirical legal research?

Internal source tag M12

Authors

Peter Cane (editor); Herbert M. Kritzer (editor)

Publication

Oxford University Press

Year

2010

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Edited handbook, not a single empirical study.

Finding recorded in corpus

Not established in the source record.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS

How can within-case evidence support policy-relevant causal inference?

Internal source tag M13

Authors

Andrew Bennett

Publication

Causality in Policy Studies, pp. 187–215

Year

2023

Publication status

PUBLISHED

Peer review

Not established in the source record.

Method

Methodological explanation with illustrative cases, Bayesian logic and discussion of rival explanations.

Finding recorded in corpus

Process tracing evaluates evidence against alternative mechanisms and can support within-case explanations, but does not estimate average treatment effects.

Scope limit

METHOD; no substantive national case

Corpus review

Preliminary source characterization only. Independent QA-00 has not certified this row; see access, method, lineage and scope limitations.

Corpus reference: CR04_Academic_Database.xlsx · WORKS