Newkirk, Alex C., Gerber, Daniel L., Fuchs, Erica R. H., Hubbard, Alex, Morgan, M. Granger, Shehabi, Arman, Smith, Sarah J., Samaras, Constantine · Nov 12, 2026
arXiv authors argue deployment speed can outweigh energy costs for AI data centers
Markdown · JSON · Original source
In an paper, Alex C. Newkirk and coauthors compare ways to power AI data centers across the continental United States. They argue that choosing the lowest-cost source of energy per unit is not enough: a power option that takes longer to deploy can delay the compute it would enable. Their “speed-adjusted cost of energy” accounts for that foregone compute, allowing infrastructure with different deployment timelines to be compared. In their model, DC-coupled solar-plus-storage—where solar panels and batteries connect without an inverter—reduces lifetime energy costs by an average of 18% compared with AC-coupled counterparts. The authors say faster deployment makes DC-coupled microgrids the lowest-cost modeled option in 85% of U.S. locations, falling to 55% when natural gas deployment is accelerated. The result makes construction , not just energy price, central to the paper’s comparison of grid-connected and off-grid power. These are modeled findings using empirically calibrated AI training loads, climate-driven cooling, power conversion and battery degradation; the abstract does not establish how actual projects would perform. The comparison covers the continental United States, and the authors’ stated location shares depend on the deployment timelines considered.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
cs.AI · Article
Source details
Original source date: Nov 12, 2026 (2026-11-12).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-13T12:18:02.350Z. First observed: 2026-11-13T07:43:52.417Z. Retrieved: 2026-11-13T12:17:08.503Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The reported cost comparisons and location shares are modeled results; the abstract does not establish how actual projects would perform. The abstract reports a comparison across the continental United States and gives a result for accelerated natural gas timelines, but does not specify other deployment-timeline scenarios.
Open question
How sensitive are the modeled shares of U.S. locations favoring DC-coupled microgrids to deployment timelines beyond the accelerated natural gas case?
Source-stated
The authors compare grid-connected and off-grid energy infrastructure for data centers across the continental United States.
We compare grid-connected and off-grid data center energy infrastructure (AC- and DC-coupled solar-plus-storage, islanded natural gas, and grid interconnection) across the continental U.S.
Source-stated
The paper introduces a speed-adjusted cost of energy that includes compute foregone while a power option is being built.
Our central contribution is a speed-adjusted cost of energy that values the compute unlocked by faster construction
Source-stated
The authors report that DC coupling lowers lifetime energy costs by an average of 18% compared with AC counterparts.
We find DC coupling reduces lifetime energy costs by 18% ($0.09/ kWh) on average versus AC counterparts
Source-stated
In the authors’ model, DC-coupled microgrids are the lowest-cost option in 85% of U.S. locations when foregone compute is valued, and 55% under accelerated natural gas timelines.
DC-coupled microgrids' faster deployment makes them the lowest-cost modeled option in 85% of U.S. locations once foregone compute is valued, or 55% under accelerated natural gas timelines.
Source-stated
The analysis uses empirically calibrated AI training loads, climate-driven cooling, stage-by-stage power conversion, and physics-based battery degradation.
built on empirically calibrated AI training loads, climate-driven cooling, stage-by-stage power conversion, and physics-based battery degradation
Source and contributor context
Newkirk, Alex C., Gerber, Daniel L., Fuchs, Erica R. H., Hubbard, Alex, Morgan, M. Granger, Shehabi, Arman, Smith, Sarah J., Samaras, Constantine · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: 87f44ac5-b9b9-4bb6-8794-9f63cd501816.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-13T12:18:02.350Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: ef5e24d4f20ff531db5e4588bf0b9c5bebcc8e1fc4697aeeb77ca6c616d6616b. Source text hash: 0feac524327fb998e68e2da3f5be0d8d85eb5fda92dc1d09f6a0f084700b7e94.
Revision and source-check history
Version 1 · 2026-11-13T12:18:02.350Z
arXiv authors argue deployment speed can outweigh energy costs for AI data centers
In an paper, Alex C. Newkirk and coauthors compare ways to power AI data centers across the continental United States. They argue that choosing the lowest-cost source of energy per unit is not enough: a power option that takes longer to deploy can delay the compute it would enable. Their “speed-adjusted cost of energy” accounts for that foregone compute, allowing infrastructure with different deployment timelines to be compared. In their model, DC-coupled solar-plus-storage—where solar panels and batteries connect without an inverter—reduces lifetime energy costs by an average of 18% compared with AC-coupled counterparts. The authors say faster deployment makes DC-coupled microgrids the lowest-cost modeled option in 85% of U.S. locations, falling to 55% when natural gas deployment is accelerated. The result makes construction , not just energy price, central to the paper’s comparison of grid-connected and off-grid power. These are modeled findings using empirically calibrated AI training loads, climate-driven cooling, power conversion and battery degradation; the abstract does not establish how actual projects would perform. The comparison covers the continental United States, and the authors’ stated location shares depend on the deployment timelines considered.
Original source
cs.AI · Article
Source details
Original source date: Nov 12, 2026 (2026-11-12).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-13T12:18:02.350Z. First observed: 2026-11-13T07:43:52.417Z. Retrieved: 2026-11-13T12:17:08.503Z.
Limits
The reported cost comparisons and location shares are modeled results; the abstract does not establish how actual projects would perform. The abstract reports a comparison across the continental United States and gives a result for accelerated natural gas timelines, but does not specify other deployment-timeline scenarios.
Open question
How sensitive are the modeled shares of U.S. locations favoring DC-coupled microgrids to deployment timelines beyond the accelerated natural gas case?
Source-stated
The authors compare grid-connected and off-grid energy infrastructure for data centers across the continental United States.
We compare grid-connected and off-grid data center energy infrastructure (AC- and DC-coupled solar-plus-storage, islanded natural gas, and grid interconnection) across the continental U.S.
Source-stated
The paper introduces a speed-adjusted cost of energy that includes compute foregone while a power option is being built.
Our central contribution is a speed-adjusted cost of energy that values the compute unlocked by faster construction
Source-stated
The authors report that DC coupling lowers lifetime energy costs by an average of 18% compared with AC counterparts.
We find DC coupling reduces lifetime energy costs by 18% ($0.09/ kWh) on average versus AC counterparts
Source-stated
In the authors’ model, DC-coupled microgrids are the lowest-cost option in 85% of U.S. locations when foregone compute is valued, and 55% under accelerated natural gas timelines.
DC-coupled microgrids' faster deployment makes them the lowest-cost modeled option in 85% of U.S. locations once foregone compute is valued, or 55% under accelerated natural gas timelines.
Source-stated
The analysis uses empirically calibrated AI training loads, climate-driven cooling, stage-by-stage power conversion, and physics-based battery degradation.
built on empirically calibrated AI training loads, climate-driven cooling, stage-by-stage power conversion, and physics-based battery degradation
Source and contributor context
Newkirk, Alex C., Gerber, Daniel L., Fuchs, Erica R. H., Hubbard, Alex, Morgan, M. Granger, Shehabi, Arman, Smith, Sarah J., Samaras, Constantine · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: 87f44ac5-b9b9-4bb6-8794-9f63cd501816.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-13T12:18:02.350Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: ef5e24d4f20ff531db5e4588bf0b9c5bebcc8e1fc4697aeeb77ca6c616d6616b. Source text hash: 0feac524327fb998e68e2da3f5be0d8d85eb5fda92dc1d09f6a0f084700b7e94.
passed · 2026-11-13T12:18:02.350Z · anthropic/claude-opus-5.5
Publication: 87f44ac5-b9b9-4bb6-8794-9f63cd501816
attribution — pass: The reading attributes the analysis and findings to Alex C. Newkirk and coauthors in an arXiv paper, which matches the listed authors. Findings stay attributed throughout ('they argue', 'in their model', 'the authors say').
Authors:Alex C. Newkirk, Daniel L. Gerber, Erica R. H. Fuchs, Alex Hubbard, M. Granger Morgan, Arman Shehabi, Sarah J. Smith, Constantine Samaras
summary — pass: The reading leads with the argument and explains speed-adjusted cost and DC coupling in plain terms. It frames the results as modeled, U.S.-scoped findings. Minor fidelity notes: the title's 'can outweigh' softens the abstract's conclusion that speed 'determines' the preferred path. The summary also omits that DC-coupled microgrids are 'rarely cheapest on unadjusted energy cost.' Restoring that clause would sharpen the paper's central contrast. 'Microgrid' could use a tooltip. Neither gap prevents understanding.
Though rarely cheapest on unadjusted energy cost, DC-coupled microgrids' faster deployment makes them the lowest-cost modeled option
date — pass: originalPublished is 2026-11-12, which matches the arXiv v1 submission date. The later feed date is correctly not used.
[Submitted on 12 Nov 2026]
support — pass: Every claim has a verbatim source excerpt. The 18% average, the 85% and 55% location shares, the scope and the model inputs all match the abstract. Describing DC coupling as connecting without an inverter is supported by the abstract's statement that it eliminates inverters. The limitations and the question follow from the evidence without inventing benefits or costs.
We find DC coupling reduces lifetime energy costs by 18% ($0.09/ kWh) on average versus AC counterparts by eliminating inverters and improving efficiency.
media — pass: The selected media is the verified direct link to the arXiv abstract page of the source paper. That is the most direct way for readers to access the source.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXiv