pith. sign in

arxiv: 2604.09616 · v1 · submitted 2026-03-15 · 💻 cs.DC

DCGen 1.1 Technical Report: Generating Datacenter Configurations (including IT, Power, Cooling)

Wedan Emmanuel Gnibga , Andrew A. Chien This is my paper

Pith reviewed 2026-05-15 12:14 UTC · model grok-4.3

classification 💻 cs.DC
keywords datacenter configurationIT hardwarepower distributioncooling infrastructureAI workloadsdatacenter modelinggrid interactions
0
0 comments X

The pith

DCGen generates realistic datacenter configurations including IT hardware, power distribution, and cooling at targeted power, compute, and area levels.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents DCGen as a tool to create varied datacenter setups that include IT servers, power systems, and cooling infrastructure. It addresses the challenge that growing datacenters with diverse workloads make realistic modeling difficult for research on design, operations, and grid interactions. DCGen draws on reference IT hardware for AI training, inference, and cloud services to form realistic server mixes, then selects components from a production equipment catalog to meet specified electrical power, compute capability, and physical area targets. This produces models that capture power, energy, and space characteristics at rack and facility scales. A reader would care because the tool enables what-if studies of power density changes, grid connections, and space use with designs closer to actual facilities than outdated alternatives.

Core claim

DCGen generates a variety of datacenter configurations including IT hardware, cooling and power distribution infrastructures at various electrical power, compute capability, and area targets. It uses reference IT configurations for specific use cases like AI training and inference to create realistic server mixes, then chooses cooling and power components from a production equipment catalog that optimizes for space or power efficiency while satisfying capacity requirements. The tool captures characteristics at both rack and datacenter levels to support modeling of power, energy, and space.

What carries the argument

Reference IT hardware configurations paired with a production equipment catalog that selects optimal components for cooling and power distribution.

If this is right

  • Models enable studies of power density evolution over time and grid interconnection capacity planning.
  • Supports what-if scenario exploration for datacenter design principles and operational dynamics.
  • Produces configurations suitable for research on datacenter interactions with the electrical grid.
  • Captures rack-level and facility-level details for space management analysis.
  • Allows generation at scales from 10 MW to 1 GW with mixes matched to AI and cloud workloads.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Generated configurations could serve as starting points for simulations that test responses to sudden workload shifts or equipment failures.
  • The catalog-driven selection process might extend to include regional factors like climate effects on cooling choices in future versions.
  • Researchers studying emerging hardware could manually update the reference set to project configurations for next-generation accelerators.

Load-bearing premise

The chosen reference IT configurations and production equipment catalog accurately reflect the power and space characteristics of real current hardware for the targeted workloads and scales.

What would settle it

A side-by-side comparison of total power draw and floor space usage between a DCGen output for a 100 MW target and measurements from an actual built datacenter of similar scale.

Figures

Figures reproduced from arXiv: 2604.09616 by Andrew A. Chien, Wedan Emmanuel Gnibga.

Figure 1
Figure 1. Figure 1: datacenter hardware configurations. The hardware setup involves identical racks, each [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: DCGen generates datacenter IT configuration, the cooling system and power distribution [PITH_FULL_IMAGE:figures/full_fig_p018_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Datacenter design including IT, cooling and power system configurations.The architecture [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: DCGen takes either a desired number of racks or a target electrical power as input. It then [PITH_FULL_IMAGE:figures/full_fig_p022_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Single datacenter configurations. GPU racks and storage racks stand for racks containing [PITH_FULL_IMAGE:figures/full_fig_p027_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: 10,000-rack datacenters IT configurations. [PITH_FULL_IMAGE:figures/full_fig_p027_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: DCGen 1.1 modeling of IT and Gray Space for 10,000-rack datacenters. [PITH_FULL_IMAGE:figures/full_fig_p028_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: 1GW datacenters IT configuration [PITH_FULL_IMAGE:figures/full_fig_p029_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: DCGen 1.1 modeling of IT and Gray Space for 1GW datacenters. [PITH_FULL_IMAGE:figures/full_fig_p030_9.png] view at source ↗
read the original abstract

Diversification of digital applications and workloads has driven the development of diverse datacenter architectures on ever-larger scales. These datacenters consist of complex IT, power, and cooling systems with interdependencies that influence configuration and performance. As datacenters scale and power density increase, designing realistic models becomes more difficult, particularly for research, because it requires understanding all layers of the datacenter and how they interact. Consequently, many studies rely on outdated or unrealistic designs. To support research in datacenter hardware design principles, operational dynamics, cooling mechanisms, and interactions of these facilities with the electrical grid, we have designed DCGen, a tool which can generate a variety of datacenter configurations (including IT hardware, cooling and power distribution infrastructures) at various electrical power, compute capability, and area targets.The tool captures power and space characteristics of IT, cooling, and power infrastructures at both the rack and datacenter levels, enabling modeling of power, energy, and space. DCGen leverages specific use cases such as AI training, AI inference, and cloud services, to select reference and canonical IT hardware configurations, producing realistic mixes of server types. It can target datacenter scale in terms of both power (e.g., 10 MW, 100 MW, 1 GW) and compute capability. For cooling and power distribution infrastructures, DCGen chooses components from a production equipment catalog that optimizes for space or power efficiency while meeting the datacenter capacity requirements. This tool supports research using realistic datacenter designs through ``what-if'' scenario exploration, including studies of power density evolution over time, grid interconnection capacity planning, datacenter-grid interactions, and space management.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript presents DCGen 1.1, a tool that generates datacenter configurations spanning IT hardware, power distribution, and cooling infrastructures. It selects reference IT configurations for use cases such as AI training, AI inference, and cloud services, then chooses components from a production equipment catalog to meet user-specified targets for electrical power (e.g., 10 MW–1 GW), compute capability, and area while optimizing for space or power efficiency. The tool aims to support research on datacenter power, energy, space modeling, grid interactions, and what-if scenario exploration by producing realistic rack- and facility-scale models.

Significance. If the reference IT configurations and catalog entries accurately reflect current hardware power and space characteristics, DCGen could serve as a useful resource for standardizing realistic datacenter models in research on hardware design, cooling dynamics, and electrical grid coupling. It addresses a practical gap where many studies rely on outdated or ad-hoc designs, potentially enabling more reproducible explorations of power-density evolution and capacity planning.

major comments (2)
  1. [Abstract] Abstract: The claim that DCGen produces 'realistic' configurations at rack and datacenter scales (10 MW–1 GW) depends on the accuracy of the reference IT hardware power/space characteristics and the production equipment catalog. The manuscript provides no validation against measured data from real facilities, no error bounds, and no quantitative evaluation of generated outputs, so any mismatch in the underlying specs propagates directly to all downstream models.
  2. [Approach] Approach description (full text): The selection logic for choosing catalog components to meet power/compute/area targets while optimizing efficiency is described at a high level but lacks concrete implementation details, such as the optimization algorithm, handling of interdependencies between IT, power, and cooling layers, or how trade-offs between space and power efficiency are resolved.
minor comments (2)
  1. Consider adding a table or appendix listing the reference IT configurations (server types, power draws, densities) for each use case to improve reproducibility.
  2. [Abstract] The abstract states the tool 'optimizes for space or power efficiency' but does not clarify whether users can select the objective or how conflicting constraints are prioritized.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their detailed review and valuable comments on the DCGen 1.1 manuscript. We have carefully considered the points raised regarding the abstract's claims and the approach description. Our responses are provided below, and we will incorporate revisions where appropriate to strengthen the paper.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that DCGen produces 'realistic' configurations at rack and datacenter scales (10 MW–1 GW) depends on the accuracy of the reference IT hardware power/space characteristics and the production equipment catalog. The manuscript provides no validation against measured data from real facilities, no error bounds, and no quantitative evaluation of generated outputs, so any mismatch in the underlying specs propagates directly to all downstream models.

    Authors: We agree that direct validation against measured data from real datacenters would strengthen the claims, but such data is proprietary and not publicly available for comparison. The configurations are derived from real production equipment catalogs and documented hardware specifications for the specified use cases. In the revised version, we will update the abstract to qualify the term 'realistic' by noting that it is based on current commercial equipment data, and we will add a discussion of data sources and limitations in the manuscript. No error bounds or quantitative evaluation can be provided without access to confidential facility data. revision: partial

  2. Referee: [Approach] Approach description (full text): The selection logic for choosing catalog components to meet power/compute/area targets while optimizing efficiency is described at a high level but lacks concrete implementation details, such as the optimization algorithm, handling of interdependencies between IT, power, and cooling layers, or how trade-offs between space and power efficiency are resolved.

    Authors: We acknowledge that the current description is high-level. We will revise the approach section to include more concrete details: the optimization uses a multi-objective greedy algorithm with constraint satisfaction for interdependencies (e.g., ensuring cooling capacity matches IT power draw and space), and trade-offs are resolved by user-specified priorities (power efficiency default, with area as secondary constraint). Pseudocode and example walkthroughs will be added to the revised manuscript. revision: yes

standing simulated objections not resolved
  • Validation of generated configurations against real-world measured data from operational datacenters, due to the proprietary nature of such information.

Circularity Check

0 steps flagged

No circularity: DCGen is a configuration generator tool with no derivations or predictions

full rationale

The paper presents DCGen as a software tool that accepts user targets (power, compute capability, area) and produces datacenter configurations by selecting from pre-existing reference IT hardware configurations and a production equipment catalog. No equations, fitted parameters, predictions, or derivations are described that reduce outputs to inputs defined within the paper. The central claim is tool functionality for 'what-if' exploration, not a mathematical result. No self-citations, uniqueness theorems, or ansatzes are invoked as load-bearing steps. This matches the default case of a self-contained tool description with score 0.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the existence and representativeness of a production equipment catalog and reference IT configurations for AI/cloud workloads. No free parameters are fitted in the abstract; inputs are user-specified targets. No new entities are invented.

axioms (2)
  • domain assumption Reference IT hardware configurations for AI training, inference, and cloud services accurately capture current server power and space characteristics.
    Invoked when selecting IT mixes to meet compute targets.
  • domain assumption Production equipment catalog contains components whose power and space data are sufficient to meet capacity requirements while optimizing for space or efficiency.
    Used for choosing cooling and power distribution infrastructure.

pith-pipeline@v0.9.0 · 5608 in / 1407 out tokens · 24637 ms · 2026-05-15T12:14:58.434397+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

85 extracted references · 85 canonical work pages

  1. [1]

    L. A. Barroso, U. H¨ olzle, and P. Ranganathan,The datacenter as a computer: Designing warehouse- scale machines. Springer Nature, 2019. 30

    work page 2019

  2. [2]

    [Online]

    LLNL,El capitan supercomputer. [Online]. Available:https : / / hpc . llnl . gov / hardware / compute-platforms/el-capitan

    work page

  3. [3]

    Kennedy,Inside the 100k gpu xai colossus cluster that supermicro helped build for elon musk

    P. Kennedy,Inside the 100k gpu xai colossus cluster that supermicro helped build for elon musk. [Online]. Available:https : / / www . supermicro . com / CaseStudies / Success _ Story _ xAI _ Colossus_Cluster.pdf

    work page

  4. [4]

    Muralidharan,Aurora exascale architecture

    S. Muralidharan,Aurora exascale architecture. [Online]. Available:https://extremecomputingtraining. anl . gov / wp - content / uploads / sites / 96 / 2023 / 08 / ATPESC - 2023 - Track - 1 - Talk - 3 - Servesh-Mulalidharan-Aurora.pdf

    work page 2023

  5. [5]

    [Online]

    NVIDIA,Nvidia dgx gb200 - advanced ai infrastructure for generative ai. [Online]. Available: https://resources.nvidia.com/en- us- dgx- systems/dgx- superpod- gb200- datasheet? ncid=no-ncid

    work page

  6. [6]

    Moss,Nvidia’s rubin ultra nvl576 rack expected to be 600kw, coming second half of 2027

    S. Moss,Nvidia’s rubin ultra nvl576 rack expected to be 600kw, coming second half of 2027. [Online]. Available:https : / / www . datacenterdynamics . com / en / news / nvidias - rubin - ultra-nvl576-rack-expected-to-be-600kw-coming-second-half-of-2027/

    work page 2027

  7. [7]

    Technoligies,Dell poweredge r760xa datasheet, Accessed: 2025-04-30

    D. Technoligies,Dell poweredge r760xa datasheet, Accessed: 2025-04-30. [Online]. Available: https : / / www . delltechnologies . com / asset / en - us / products / servers / technical - support/poweredge-r760xa-spec-sheet.pdf

    work page 2025

  8. [8]

    [Online]

    SuperMicro,Supermicro liquid-cooling solutions, Accessed: 2025-04-30. [Online]. Available:https: //www.supermicro.com/solutions/datasheet_Liquid_Cooling_Solutions.pdf

    work page 2025

  9. [9]

    The infrastructure powering ibm’s gen ai model development,

    T. Gershon et al., “The infrastructure powering ibm’s gen ai model development,”arXiv preprint arXiv:2407.05467, 2024

    work page arXiv 2024

  10. [10]

    [Online]

    Google,Google distributed cloud hardware configuration. [Online]. Available:https://cloud. google.com/distributed-cloud/edge/1.6.0/docs/requirements

    work page

  11. [11]

    Lammertyn,60+ chatgpt facts and statistics you need to know in 2025

    M. Lammertyn,60+ chatgpt facts and statistics you need to know in 2025. [Online]. Available: https://blog.invgate.com/chatgpt-statistics

    work page 2025

  12. [12]

    Press,Lenovo thinksystem sr670 v2, datasheet

    L. Press,Lenovo thinksystem sr670 v2, datasheet. [Online]. Available:https://lenovopress. lenovo.com/datasheet/ds0123-lenovo-thinksystem-sr670-v2

    work page

  13. [13]

    Technologies,Dell poweredge xe7745, datasheet

    D. Technologies,Dell poweredge xe7745, datasheet. [Online]. Available:https://www.delltechnologies. com / asset / en - us / products / servers / technical - support / poweredge - xe7745 - spec - sheet.pdf

    work page

  14. [14]

    Williams,Megawatt-class ai server racks may well become the norm before 2030 as nvidia displays 600kw kyber rack design

    W. Williams,Megawatt-class ai server racks may well become the norm before 2030 as nvidia displays 600kw kyber rack design. [Online]. Available:https : / / www . techradar . com / pro / megawatt-class-ai-server-racks-may-well-become-the-norm-before-2030-as-nvidia- displays-600kw-kyber-rack-design?utm_source=chatgpt.com

    work page 2030

  15. [15]

    [Online]

    Supermicro,Storage superserver ssg-121e-ne3x12r. [Online]. Available:https://www.supermicro. com/en/products/system/storage/1u/ssg-121e-ne3x12r

    work page

  16. [16]

    [Online]

    Nytro,Nytro 5060 nvme ssd series, Accessed: 2025-04-30, 2025. [Online]. Available:https : //www.seagate.com/content/dam/seagate/en/content-fragments/products/datasheets/ enterprise-rebranding/nytro-5060/nytro-5060-ssd-DS11-5-2502US-en_US.pdf

    work page 2025

  17. [17]

    [Online]

    Frontier,Ornl’s exascale supercomputer, 2022. [Online]. Available:https://www.olcf.ornl. gov/frontier/

    work page 2022

  18. [18]

    Cray,A quick primer on hpe cray frontier’s parallel file system storage

    H. Cray,A quick primer on hpe cray frontier’s parallel file system storage. [Online]. Available: https://blocksandfiles.com/2022/06/04/the-dummies-guide-to-hpe-cray-frontiers- parallel - file - system - storage / # :~ : text = the % 20Frontier % 20cores. - ,Orion , on % 20E1000%20SSU%2DD%20devices

    work page 2022

  19. [19]

    Cray,Hpe clusterstor 6.x e1000 field installation guide h-6215

    H. Cray,Hpe clusterstor 6.x e1000 field installation guide h-6215. [Online]. Available:https: / / support . hpe . com / hpesc / public / docDisplay ? docId = sd00001832en _ us & page = GUID - 83CF919B-8EB8-4EBD-A486-98B834CF0872.html

    work page

  20. [20]

    500,Frontier - hpe cray ex235a, amd optimized 3rd generation epyc 64c 2ghz, amd instinct mi250x, slingshot-11, hpe cray os, 2022

    T. 500,Frontier - hpe cray ex235a, amd optimized 3rd generation epyc 64c 2ghz, amd instinct mi250x, slingshot-11, hpe cray os, 2022. [Online]. Available:https://www.top500.org/system/ 180047/ 31

    work page 2022

  21. [21]

    Cray,Cray clusterstor e1000 - product information reference

    H. Cray,Cray clusterstor e1000 - product information reference. [Online]. Available:https : / / support . hpe . com / hpesc / public / docDisplay ? docId = a00104724en _ us & page = GUID - 96EC988F-B220-4E73-B540-9C4FC2B14346.html

    work page

  22. [22]

    Moss,Oak ridge’s exascale ’frontier’ system named world’s most powerful supercomputer on top500

    S. Moss,Oak ridge’s exascale ’frontier’ system named world’s most powerful supercomputer on top500. [Online]. Available:https://www.datacenterdynamics.com/en/news/oak- ridges- exascale-frontier-system-named-worlds-most-powerful-supercomputer-on-top500/

    work page

  23. [23]

    Trader,Livermore’s el capitan supercomputer to debut hpe ‘rabbit’ near node local storage

    T. Trader,Livermore’s el capitan supercomputer to debut hpe ‘rabbit’ near node local storage. [Online]. Available:https : / / www . hpcwire . com / 2021 / 02 / 18 / livermores - el - capitan - supercomputer-hpe-rabbit-storage-nodes/?utm_source=chatgpt.com

    work page 2021

  24. [24]

    [Online]

    xAI Colossus,Xai colossus offical website. [Online]. Available:https://x.ai/colossus

    work page

  25. [25]

    [Online]

    Supermicro,Supermicro rack scale liquid cooling solutions. [Online]. Available:https://www. supermicro . com / solutions / Solution - Brief _ Supermicro _ Liquid _ Cooling _ Solution _ Guide.pdf

    work page

  26. [26]

    5000,Aurora - hpe cray ex - intel exascale compute blade, xeon cpu max 9470 52c 2.4ghz, intel data center gpu max, slingshot-11

    T. 5000,Aurora - hpe cray ex - intel exascale compute blade, xeon cpu max 9470 52c 2.4ghz, intel data center gpu max, slingshot-11. [Online]. Available:https://top500.org/system/180183/

    work page

  27. [27]

    [Online]

    NVIDIA,Nvidia h100 tensor core gpu, Accessed: 2025-04-30. [Online]. Available:https://www. nvidia.com/en-us/data-center/h100/

    work page 2025

  28. [28]

    [Online]

    Supermicro,Supermicro gpu ars-111gl-nhr, nvidia gh200 grace hopper superchip system support- ing nvidia bluefield-3 or nvidia connectx-7, Accessed: 2025-04-30. [Online]. Available:https : //www.supermicro.com/en/products/system/gpu/1u/ars-111gl-nhr

    work page 2025

  29. [29]

    microsoft

    Microsoft,Ncads h100 v5-series,https : / / learn . microsoft . com / en - us / azure / virtual - machines/sizes/gpu- accelerated/nccadsh100v5- series?tabs=sizestorageremote, Ac- cessed: 2025-05-01, 2024

    work page 2025

  30. [30]

    Kennedy,Supermicro ars-111gl-nhr nvidia gh200 grace hopper 1u server review, Accessed: 2025-04-30, 2024

    P. Kennedy,Supermicro ars-111gl-nhr nvidia gh200 grace hopper 1u server review, Accessed: 2025-04-30, 2024. [Online]. Available:https : / / www . servethehome . com / supermicro - ars - 111gl - nhr - nvidia - gh200 - grace - hopper - 1u - server - review - arm / 4 / ?utm _ source = chatgpt.com

    work page 2025

  31. [31]

    [Online]

    PNY,Nvidia h100 nvl perormance, Accessed: 2025-04-30, 2024. [Online]. Available:https:// www.pny.com/nvidia-h100-nvl

    work page 2025

  32. [32]

    [Online]

    BusinessWire,Inspur information’s nf5488a5 ai server and supply chain win gold and bronze stevie®2022 international business awards, Accessed: 2025-04-30, 2025. [Online]. Available: https://www.businesswire.com/news/home/20220822005201/en/Inspur- Informations- NF5488A5-AI-Server-and-Supply-Chain-Win-Gold-and-Bronze-Stevie-2022-International- Business-Awards

    work page arXiv 2022

  33. [33]

    Kennedy,Inspur nf5488a5 8x nvidia a100 hgx platform review

    P. Kennedy,Inspur nf5488a5 8x nvidia a100 hgx platform review. [Online]. Available:https: //www.servethehome.com/inspur-nf5488a5-8x-nvidia-a100-hgx-platform-review-amd- epyc/5/?utm_source=chatgpt.com

    work page

  34. [34]

    Reiner Pope, Sholto Douglas, Aakanksha Chowdhery, Jacob Devlin, James Bradbury, Anselm Lev- skaya, Jonathan Heek, Kefan Xiao, Shivani Agrawal, and Jeff Dean

    J. Wang et al., “Designing cloud servers for lower carbon,” in2024 ACM/IEEE 51st Annual International Symposium on Computer Architecture (ISCA), 2024, pp. 452–470.doi:10.1109/ ISCA59077.2024.00041

    work page arXiv 2024

  35. [35]

    D. Technoligies,Dell integrated system for microsoft azure stack hci: End-to-end deployment and operations guide - cluster creation using windows admin center deployment guide, Accessed: 2025-04-30. [Online]. Available:https : / / www . dell . com / support / manuals / en - us / ax - 640 / ashci _ deployment _ option _ guide _ wac / azure - stack - hci - ...

    work page 2025

  36. [36]

    technologies,Poweredge r660xs datasheet, Accessed: 2025-04-30

    D. technologies,Poweredge r660xs datasheet, Accessed: 2025-04-30. [Online]. Available:https: / / www . delltechnologies . com / asset / en - us / products / servers / technical - support / poweredge-r660xs-spec-sheet.pdf 32

    work page 2025

  37. [37]

    H. P. services,Hpe performance optimized data center solutions, Accessed: 2025-04-14, 2023. [Online]. Available:https://4719eaee91034be722d8- c86a406a93c55de2464febd03debd4f0. ssl.cf1.rackcdn.com/broch_HPE_Performance_Optimized_DC.PDF

    work page 2025

  38. [38]

    [Online]

    Fujitsu,Specifications - supercomputer fugaku, Accessed: 2025-04-14, 2025. [Online]. Available: https://www.fujitsu.com/global/about/innovation/fugaku/specifications/

    work page 2025

  39. [39]

    Characterizing the opportunity for reducing the operational carbon footprint of storage systems,

    V. RAO, “Characterizing the opportunity for reducing the operational carbon footprint of storage systems,” M.S. thesis, THE UNIVERSITY OF CHICAGO, 2024

    work page 2024

  40. [40]

    Technoligies,Dell azure stack hci server configurations, Accessed: 2025-04-30

    D. Technoligies,Dell azure stack hci server configurations, Accessed: 2025-04-30. [Online]. Avail- able:https://infohub.delltechnologies.com/en- us/l/dell- validated- design- for- retail - edge - design - guide - with - invia - robotics / dell - azure - stack - hci - server - configurations/

    work page 2025

  41. [41]

    [Online]

    Wikipedia,Fugaku (supercomputer), Accessed: 2025-04-14, 2022. [Online]. Available:https : //en.wikipedia.org/wiki/Fugaku_(supercomputer)

    work page 2025

  42. [42]

    Ajima,Tofu interconnect d of supercomputer fugaku and future prospects, Accessed: 2025- 04-14, 2022

    Y. Ajima,Tofu interconnect d of supercomputer fugaku and future prospects, Accessed: 2025- 04-14, 2022. [Online]. Available:https : / / nowlab . cse . ohio - state . edu / static / media / workshops/presentations/exacomm22/04_exacomm2022_ajima.pdf

    work page 2025

  43. [43]

    Powering intelligence: Analyzing artificial intelligence and data center energy consumption,

    J. Aljbour, T. Wilson, and P Patel, “Powering intelligence: Analyzing artificial intelligence and data center energy consumption,”EPRI White Paper no. 3002028905, 2024

    work page 2024

  44. [44]

    Vertiv coolchip cdu

    Vertiv. “Vertiv coolchip cdu.” Accessed: 2025-08-26. [Online]. Available:https://www.vertiv. com/en- us/products- catalog/thermal- management/high- density- solutions/vertiv- coolchip-cdu/

    work page 2025

  45. [45]

    Vertiv coolphase cdu brochure

    Vertiv. “Vertiv coolphase cdu brochure.” PDF brochure. Accessed: 2025-08-26. [Online]. Avail- able:https : / / www . vertiv . com / 498c23 / globalassets / products / thermal - management / high - density - solutions / vertiv - coolchip - econophase - cdu / vertiv - coolphase - cdu - brochure-sl-71255.pdf

    work page 2025

  46. [46]

    Vertiv coolchip cdu 600

    Vertiv. “Vertiv coolchip cdu 600.” Accessed: 2025-08-26. [Online]. Available:https : / / www . vertiv.com/en- us/products- catalog/thermal- management/high- density- solutions/ vertiv-coolchip-cdu-600/#/models

    work page 2025

  47. [47]

    Vertiv coolchip cdu 100kw data sheet

    Vertiv. “Vertiv coolchip cdu 100kw data sheet.” PDF datasheet. Accessed: 2025-08-26. [On- line]. Available:https : / / www . vertiv . com / 49c795 / globalassets / products / thermal - management / high - density - solutions / vertiv - coolchip - cdu / vertiv - coolchip - cdu - 100kw-data-sheet-sl-71348.pdf

    work page 2025

  48. [48]

    Vertiv megamod coolchip cdu datasheet

    Vertiv. “Vertiv megamod coolchip cdu datasheet.” PDF datasheet. Accessed: 2025-08-26. [On- line]. Available:https://www.vertiv.com/4a1a2b/globalassets/products/facilities- enclosures - and - racks / integrated - solutions / vertiv - megamod - coolchip / vertiv - megamodcc-20240717.pdf

    work page 2025

  49. [49]

    Liebert hpc-w datasheet

    Emerson Network Power. “Liebert hpc-w datasheet.” PDF datasheet. Accessed: 2025-08-26. [On- line]. Available:http://www.edtgroup.eu/userfiles/file/chladiaci- system/Liebert- HPC-W.pdf

    work page 2025

  50. [50]

    Liebert hpc-w water-cooled chillers

    Vertiv. “Liebert hpc-w water-cooled chillers.” Accessed: 2025-08-26. [Online]. Available:https: / / www . vertiv . com / en - anz / products - catalog / thermal - management / free - cooling - chillers/liebert-hpc-w-water-cooled-chillers/

    work page 2025

  51. [51]

    Liebert ofc brochure

    Vertiv. “Liebert ofc brochure.” PDF brochure. Accessed: 2025-08-26. [Online]. Available:https: //www.vertiv.com/491163/globalassets/products/thermal-management/free-cooling- chillers/liebert-ofc/liebert-ofc-brochure-english.pdf

    work page 2025

  52. [52]

    Refteco,Dry cooler v-shape rdvd-1 datasheet,https : / / www . refteco . com / wp - content / uploads/2024/12/Dry-cooler-V-Shape-RDVD-1.pdf, Accessed: 2025-08-26, 2025

    work page 2024

  53. [53]

    Vertiv,Liebert lvc v-shaped condenser user manual,https : / / www . vertiv . com / 498650 / globalassets / products / thermal - management / condensers - and - drycoolers / liebert - lvc-v-shaped-condenser/liebert-lvc-v-shaped-condenser-user-manual.pdf, Accessed: 2025-08-26, 2025. 33

    work page 2025

  54. [54]

    com / 4ae379 / globalassets / shared / liebert - air - cooled - direct - drive - drycoolers - technical-design-manual_00.pdf, Accessed: 2025-08-26, 2025

    Vertiv,Liebert air-cooled direct-drive drycoolers technical design manual,https://www.vertiv. com / 4ae379 / globalassets / shared / liebert - air - cooled - direct - drive - drycoolers - technical-design-manual_00.pdf, Accessed: 2025-08-26, 2025

    work page 2025

  55. [55]

    Containerized 167b-au

    Evapco. “Containerized 167b-au.” PDF datasheet. Accessed: 2025-08-26. [Online]. Available: https : / / www . evapco . com . au / sites / evapco . com . au / files / 2017 - 09 / Containerized % 20167B-AU%20Publish_1.pdf

    work page 2025

  56. [56]

    At-ut-uss engineering

    Evapco. “At-ut-uss engineering.” PDF engineering document. Accessed: 2025-08-26. [Online]. Available:https://www.evapco.com.au/sites/evapco.com.au/files/2020- 11/AT- UT- USS%20Engineering%201120.pdf

    work page 2025

  57. [57]

    Vertiv,Build ai-ready infrastructure with proven, validated designs,https://www.vertiv.com/ en-in/solutions/ai-hub/design/, Accessed: 2025-08-23

    work page 2025

  58. [58]

    Vertiv powerboard ul891 switchboard brochure

    Vertiv. “Vertiv powerboard ul891 switchboard brochure.” PDF brochure. Accessed: 2025-08-26. [Online]. Available:https://www.vertiv.com/498672/globalassets/products/critical- power/switchgear/vertiv-powerboard-ul891-switchboard/vertiv-ul891-switchboard- brochure-sl-70901.pdf

    work page 2025

  59. [59]

    Vertiv powerboard lv switchgear iec brochure

    Vertiv. “Vertiv powerboard lv switchgear iec brochure.” PDF brochure. Accessed: 2025-08-26. [Online]. Available:https://www.vertiv.com/48dadf/globalassets/products/critical- power/switchgear/vertiv- powerboard- lv- switchgear- iec/vertiv- ei- lvswitchgear- br-en-emea-sl-70902-web.pdf

    work page 2025

  60. [60]

    Diesel generator c1.1

    Caterpillar. “Diesel generator c1.1.” Accessed: 2025-08-26. [Online]. Available:https://www. cat . com / en _ US / products / new / power - systems / electric - power / diesel - generator - sets/1000028902.html

    work page arXiv 2025

  61. [61]

    Gas generator dg125 (single and 3 phase)

    Caterpillar. “Gas generator dg125 (single and 3 phase).” Accessed: 2025-08-26. [Online]. Avail- able:https://www.cat.com/en_US/products/new/power- systems/electric- power/gas- generator-sets/119541.html

    work page 2025

  62. [62]

    2004 w¨ artsil¨ a 10mw power plant

    IMP Corporation. “2004 w¨ artsil¨ a 10mw power plant.” Accessed: 2025-08-26. [Online]. Available: https : / / www . impcorporation . com / en - us / inventory / details / 14318 / 2004 - wartsila - 10mw-power-plant?utm_source=chatgpt.com

    work page 2004

  63. [63]

    Noiseless 10mw / 10000kw / 12500kva diesel generator 690v 3-phase

    Guangling. “Noiseless 10mw / 10000kw / 12500kva diesel generator 690v 3-phase.” Accessed: 2025-08-26. [Online]. Available:https : / / guangling . en . made - in - china . com / product / BOTtdgiKrXUI / China - Noiseless - 10MW - 10000kw - 12500kVA - Diesel - Generator - 690V - 3phase-Price-List.html

    work page 2025

  64. [64]

    C280-16 diesel generator set

    Caterpillar. “C280-16 diesel generator set.” Accessed: 2025-08-26. [Online]. Available:https: //www.cat.com/en_US/products/new/power-systems/electric-power/diesel-generator- sets/111061.html

    work page 2025

  65. [65]

    10mw–20mw power plant generators cummins/mtu diesel generator 1000kva–3600kva open container-type medium hv generator

    Meccagen. “10mw–20mw power plant generators cummins/mtu diesel generator 1000kva–3600kva open container-type medium hv generator.” Accessed: 2025-08-26. [Online]. Available:https: / / meccagen . en . made - in - china . com / product / RFoaSYglOIfn / China - 10MW - 12MW - 20MW - Power - Plant - Generators - Cummins - Mtu - Diesel - Generator - 1000kVA - 3...

    work page 2025

  66. [66]

    Liebert apm ups 30–600kw brochure

    Vertiv. “Liebert apm ups 30–600kw brochure.” PDF brochure. Accessed: 2025-08-26. [Online]. Available:https : / / www . vertiv . com / 4904ab / globalassets / products / critical - power / uninterruptible-power-supplies-ups/liebert-apm-ups-30-600-kw/liebert-apm-from- 30-kw-to-600-kw-brochure-mka4l0ukapm-.pdf

    work page 2025

  67. [67]

    Vertiv,Vertiv powerups 9000 brochure,https : / / www . vertiv . com / 49948c / globalassets / products / critical - power / uninterruptible - power - supplies - ups / vertiv - powerups - 9000/vertiv-powerups-9000-br-en-gl-mka4l0ukpowu9_ce-web.pdf, Accessed: 2025-08-26, 2025

    work page 2025

  68. [68]

    Vertiv,Vertiv trinergy ups brochure,https : / / www . vertiv . com / 49a803 / globalassets / products / critical - power / uninterruptible - power - supplies - ups / vertiv - trinergy / vertiv-trinergy-ups-brochure-sl-71331.pdf, Accessed: 2025-08-26, 2025. 34

    work page 2025

  69. [69]

    Vertiv,Liebert exl s1 ups brochure,https : / / www . vertiv . com / 4ab416 / globalassets / products / critical - power / uninterruptible - power - supplies - ups / liebert - exl - s1 - brochure-sl-70645.pdf, Accessed: 2025-08-26, 2025

    work page 2025

  70. [70]

    Vertiv,Vertiv power module h2 zero-emission backup power,https://www.vertiv.com/49a0cd/ globalassets / products / facilities - enclosures - and - racks / integrated - solutions / power - module - h2 - zero - emission - backup - power / vertiv - power - module - hydrogen - _20250506.pdf, Accessed: 2025-08-26, 2025

    work page 2025

  71. [71]

    Vertiv,Liebert spm 1.0 power distribution brochure,https : / / www . vertiv . com / 4a9d43 / globalassets / products / critical - power / power - distribution / liebert - spm - 1 . 0 / liebert-spm-1.0-br-en-asia.pdf, Accessed: 2025-08-26, 2025

    work page 2025

  72. [72]

    Vertiv,Liebert apt power distribution brochure,https://www.vertiv.com/4aae06/globalassets/ products/critical- power/power- distribution/liebert_apt_brochure_en_in.pdf, Ac- cessed: 2025-08-26, 2025

    work page 2025

  73. [73]

    pdf, Ac- cessed: 2025-08-26, 2025

    Vertiv,Liebert tfx pdu brochure,https://www.vertiv.com/49ae85/globalassets/products/ critical - power / power - distribution / liebert - tfx - pdu - brochure - sl - 11334 . pdf, Ac- cessed: 2025-08-26, 2025

    work page 2025

  74. [74]

    Vertiv,Liebert sts2 pdu static transfer switch brochure,https : / / www . vertiv . com / 4aff7c / globalassets/products/critical-power/static-transfer-switches/liebert-sts2-pdu- static- transfer- switch- power- distribution- unit/liebert- sts2pdu- br- en- na- sl- 20700.pdf, Accessed: 2025-08-26, 2025

    work page 2025

  75. [75]

    Tapas: Thermal-and power-aware scheduling for llm inference in cloud plat- forms,

    J. Stojkovic et al., “Tapas: Thermal-and power-aware scheduling for llm inference in cloud plat- forms,” inProceedings of the 30th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2, 2025, pp. 1266–1281

    work page 2025

  76. [76]

    Flex: High-availability datacenters with zero reserved power,

    C. Zhang et al., “Flex: High-availability datacenters with zero reserved power,” in2021 ACM/IEEE 48th Annual International Symposium on Computer Architecture (ISCA), IEEE, 2021, pp. 319– 332

    work page 2021

  77. [77]

    Data centers cooling: A critical review of techniques, challenges, and energy saving solutions,

    A. A. Alkrush, M. S. Salem, O. Abdelrehim, and A. Hegazi, “Data centers cooling: A critical review of techniques, challenges, and energy saving solutions,”International Journal of Refrig- eration, vol. 160, pp. 246–262, 2024

    work page 2024

  78. [78]

    How to build an ai datacentre – part 1: Cooling and power

    Elongated musk. “How to build an ai datacentre – part 1: Cooling and power.” Accessed: 2025- 08-28. [Online]. Available:https://medium.com/@Elongated_musk/how- to- build- an- ai- datacentre-part-1-cooling-and-power-5c15ddfc16c9

    work page 2025

  79. [79]

    Vertiv,Long-term ai infrastructure predictions,https://www.vertiv.com/fr-ca/about/news- and - insights / articles / white - papers / long - term - ai - infrastructure - predictions / ?utm_source=chatgpt.com, Accessed: 2025-08-28, 2024

    work page 2025

  80. [80]

    Vincent,Coolit and accelsius push data center liquid cooling limits amid soaring rack densities, Accessed: 2025-04-30, 2025

    M. Vincent,Coolit and accelsius push data center liquid cooling limits amid soaring rack densities, Accessed: 2025-04-30, 2025. [Online]. Available:https : / / www . datacenterfrontier . com / cooling/article/55281394/coolit-and-accelsius-push-data-center-liquid-cooling- limits-amid-soaring-rack-densities

    work page arXiv 2025

Showing first 80 references.