{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,5]],"date-time":"2026-09-05T16:02:30Z","timestamp":1788624150255,"version":"build-2803163510"},"reference-count":128,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,6]]},"DOI":"10.1109\/isca66397.2026.00029","type":"proceedings-article","created":{"date-parts":[[2026,8,4]],"date-time":"2026-08-04T19:14:39Z","timestamp":1785870879000},"page":"194-210","source":"Crossref","is-referenced-by-count":1,"title":["Rearchitecting the Datacenter Lifecycle for AI"],"prefix":"10.1109","author":[{"given":"Jovan","family":"Stojkovic","sequence":"first","affiliation":[{"name":"The University of Texas at Austin"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chaojie","family":"Zhang","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"\u00cd\u00f1igo","family":"Goiri","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ricardo","family":"Bianchini","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2025.3532853"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.678"},{"key":"ref3","article-title":"Amazon AWS","volume-title":"Power usage effectiveness","year":"2025"},{"key":"ref4","article-title":"AMD","volume-title":"AMD Instinct\u2122 GPUs","year":"2025"},{"key":"ref5","article-title":"I. T. Association","volume-title":"Infiniband architecture specification","year":"2001"},{"key":"ref6","volume-title":"Microsoft and NVIDIA accelerate AI development and performance","author":"Azure","year":"2025"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/MM.2024.3426478"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1002\/cpe.2963"},{"key":"ref9","volume-title":"Using Large Language Models in Learning and Teaching","author":"Bowne","year":"2024"},{"key":"ref10","volume-title":"Language Models are Few-Shot Learners","author":"Brown","year":"2020"},{"key":"ref11","volume-title":"Why 2024 will be the year of inference","author":"Capital","year":"2025"},{"key":"ref12","article-title":"Murakkab: Resource-efficient agentic workflow orchestration in cloud platforms","author":"Chaudhry","year":"2025","journal-title":"arXiv preprint"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1007\/s11432-025-4665-8"},{"key":"ref14","article-title":"Power stabilization for ai training datacenters","author":"Choukse","year":"2025","journal-title":"arXiv preprint"},{"key":"ref15","volume-title":"Introducing A4X VMs powered by NVIDIA GB200","author":"Cloud","year":"2025"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1145\/3695053.3731409"},{"key":"ref17","article-title":"CoreSite","volume-title":"AI and the Data Center: Driving Greater Power Density","year":"2025"},{"key":"ref18","article-title":"Cyfuture Cloud","volume-title":"What is the Price of an NVIDIA DGX H100 System?","year":"2025"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1189"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1016\/j.applthermaleng.2016.10.093"},{"key":"ref21","article-title":"DeepSeek","volume-title":"Introducing DeepSeek-V3","year":"2025"},{"key":"ref22","article-title":"Hybrid llm: Cost-efficient and quality-aware query routing","author":"Ding","year":"2024","journal-title":"arXiv preprint"},{"key":"ref23","article-title":"Energy.gov","volume-title":"Evaporative Coolers","year":"2024"},{"key":"ref24","article-title":"Epoch AI","volume-title":"Data on notable ai models","year":"2024"},{"key":"ref25","volume-title":"Frontier language models have become much smaller","author":"Erdil","year":"2024"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1145\/1250662.1250665"},{"key":"ref27","first-page":"111","article-title":"Cost-Efficient Large Language Model serving for multiturn conversations with CachedAttention","volume-title":"2024 USENIX Annual Technical Conference (USENIX ATC 24)","author":"Gao","year":"2024"},{"key":"ref28","article-title":"GitHub","volume-title":"The world\u2019s most widely adopted AI developer tool","year":"2024"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/ICDCS.2011.19"},{"key":"ref30","volume-title":"Tokenweave: Efficient compute-communication overlap for distributed llm inference","author":"Gond","year":"2025"},{"key":"ref31","article-title":"Google","volume-title":"Growing the internet while reducing energy consumption","year":"2025"},{"key":"ref32","article-title":"Google Cloud","volume-title":"Enabling 1 MW IT racks and liquid cooling at OCP EMEA Summit","year":"2025"},{"key":"ref33","article-title":"Grand View Research","volume-title":"U.S. Generative AI Market Size & Trends","year":"2025"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1145\/1496091.1496103"},{"key":"ref35","article-title":"GridStatus.io Team","year":"2025","journal-title":"GridStatus: Real-Time Power Grid Status Monitoring"},{"key":"ref36","article-title":"Managing the Server Migration Process: The HP Approach to Reducing Operational Costs","author":"Grieser","year":"2010","journal-title":"HP, Tech. Rep."},{"key":"ref37","article-title":"Groq","volume-title":"What is a Language Processing Unit?","year":"2025"},{"key":"ref38","volume-title":"Efficiently Modeling Long Sequences with Structured State Spaces","author":"Gu","year":"2022"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-025-09422-z"},{"key":"ref40","volume-title":"System-performance and cost modeling of Large Language Model training and inference","author":"Guo","year":"2025"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1145\/3470496.3527408"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1145\/3617995"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1115\/IPACK2022-97447"},{"key":"ref44","volume-title":"Trends in Machine Learning Hardware","author":"Hobbhahn","year":"2023"},{"key":"ref45","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2023.findings-acl.507","volume-title":"Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes","author":"Hsieh","year":"2023"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1145\/3173162.3173190"},{"key":"ref47","author":"Hu","journal-title":"LoRA: Low-Rank Adaptation of Large Language Models"},{"key":"ref48","volume-title":"Our commitment to climate-conscious data center cooling","author":"H\u00f6lzle","year":"2022"},{"key":"ref49","article-title":"IBM","volume-title":"What is total cost of ownership (TCO)?","year":"2025"},{"key":"ref50","article-title":"Intel","volume-title":"Intel\u00ae Xeon\u00ae 6980P Processor","year":"2025"},{"key":"ref51","volume-title":"Intel\u00ae Xeon\u00ae Platinum 8593Q Processor","year":"2025"},{"key":"ref52","volume-title":"Intel\u00ae Xeon\u00ae Processor E7\u20138890 v3","year":"2025"},{"key":"ref53","article-title":"Intelligent router for llm workloads: Improving performance through workload-aware load balancing","author":"Jain","year":"2024","journal-title":"arXiv preprint"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA52012.2021.00055"},{"key":"ref55","article-title":"HexGen: Generative Inference of Large Language Model over Heterogeneous Environment","author":"Jiang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1145\/3579371.3589350"},{"key":"ref57","volume-title":"Microsoft Q3 FY 2025: Azure and AI Services Drive Strong Growth","author":"Kirkpatrick","year":"2025"},{"key":"ref58","volume":"01","author":"Koomey","year":"2007","journal-title":"A Simple Model for Determining True Total Cost of Ownership for Data Centers"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1145\/3600006.3613165"},{"key":"ref60","volume-title":"60+ ChatGPT Statistics And Facts You Need to Know in 2024","author":"Lammertyn","year":"2024"},{"key":"ref61","volume-title":"Tesla A100 Server Total Cost of Ownership Analysis","author":"Li","year":"2025"},{"key":"ref62","article-title":"Large language model inference acceleration: A comprehensive hardware perspective","volume-title":"arXiv preprint","author":"Li","year":"2025"},{"key":"ref63","article-title":"Thunderbolt: Throughput-Optimized, Qualityof-Service-Aware Power Capping at Scale","volume-title":"Proceedings of the 14th USENIX Symposium on Operating Systems Design and Implementation (OSDI)","author":"Li","year":"2020"},{"key":"ref64","author":"Li","year":"2025","journal-title":"EcoServe: Designing carbon-aware ai inference systems"},{"key":"ref65","volume-title":"The unseen ai disruptions for power grids: Llm-induced transients","author":"Li","year":"2024"},{"key":"ref66","article-title":"Qserve: W4a8kv4 quantization and system co-design for efficient 1lm serving","author":"Lin","year":"2024","journal-title":"arXiv preprint"},{"key":"ref67","author":"Liu","year":"2024","journal-title":"DeepSeek-V3 Technical Report"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.1145\/3669940.3707215"},{"key":"ref69","article-title":"Meta","volume-title":"2024 Sustainability Report","year":"2024"},{"key":"ref70","article-title":"Meta","volume-title":"Llama2\u201370B"},{"key":"ref71","volume-title":"Llama3\u201370B","year":"2024"},{"key":"ref72","article-title":"Meta","volume-title":"The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation","year":"2025"},{"key":"ref73","doi-asserted-by":"publisher","DOI":"10.1080\/01621459.1949.10483310"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.1145\/3620665.3640411"},{"key":"ref75","article-title":"Microsoft","volume-title":"Measuring datacenter energy and water use to improve Microsoft Cloud sustainability","year":"2025"},{"key":"ref76","article-title":"Microsoft Azure","volume-title":"AI Lifecycle Compass","year":"2026"},{"key":"ref77","doi-asserted-by":"publisher","DOI":"10.1145\/3695053.3731051"},{"key":"ref78","article-title":"How Much Do GPU Clusters Really Cost? Calculating Cluster TCO and the Real Impact of Downtime","volume-title":"SemiAnalysis Newsletter","author":"Nanos","year":"2026"},{"key":"ref79","volume-title":"The cost of compute: A $7 trillion race to scale data centers","author":"Noffsinger","year":"2025"},{"key":"ref80","article-title":"NVIDIA","volume-title":"Cooling and Airflow Optimization","year":"2024"},{"key":"ref81","article-title":"NVIDIA","volume-title":"DGX H100: AI for Enterprise","year":"2024"},{"key":"ref82","volume-title":"Introduction to the NVIDIA DGX A100 System","year":"2024"},{"key":"ref83","article-title":"NVIDIA","volume-title":"NVIDIA Blackwell Platform Arrives to Power a New Era of Computing","year":"2024"},{"key":"ref84","article-title":"NVIDIA","volume-title":"Chill Factor: NVIDIA Blackwell Platform Boosts Water Efficiency by Over 300x","year":"2025"},{"key":"ref85","volume-title":"LLM Model Pruning and Knowledge Distillation with NVIDIA NeMo Framework","year":"2025"},{"key":"ref86","volume-title":"NVIDIA Collective Communications Library (NCCL)","year":"2025"},{"key":"ref87","volume-title":"NVIDIA ConnectX-7 Adapter Cards User Manual","year":"2025"},{"key":"ref88","volume-title":"NVIDIA Dynamo Platform","year":"2025"},{"key":"ref89","volume-title":"NVIDIA Quantum-2 InfiniBand Platform","year":"2025"},{"key":"ref90","volume-title":"NVLink and NVLink Switch","year":"2025"},{"key":"ref91","volume-title":"RDMA over Converged Ethernet (RoCE)","year":"2025"},{"key":"ref92","article-title":"OpenAI","volume-title":"Models","year":"2025"},{"key":"ref93","doi-asserted-by":"publisher","DOI":"10.1145\/3620666.3651329"},{"key":"ref94","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA59077.2024.00019"},{"key":"ref95","doi-asserted-by":"publisher","DOI":"10.1038\/s41746-023-00958-w"},{"key":"ref96","volume-title":"Q2 earnings call: CEO\u2019s remarks","author":"Pichai","year":"2025"},{"key":"ref97","doi-asserted-by":"publisher","DOI":"10.1145\/3373376.3378533"},{"key":"ref98","doi-asserted-by":"publisher","DOI":"10.1109\/HPEC58863.2023.10363447"},{"key":"ref99","volume-title":"Amazon EC2 P6e-GB200 UltraServers and P6-B200 instances","author":"Services","year":"2025"},{"key":"ref100","volume-title":"Outrageously large neural networks: The sparsely-gated mixture-of-experts layer","author":"Shazeer","year":"2017"},{"key":"ref101","author":"Shoeybi","year":"2019","journal-title":"Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism"},{"key":"ref102","article-title":"Towards Greener LLMs: Bringing Energy-Efficiency to the Forefront of LLM Inference","author":"Stojkovic","year":"2024","journal-title":"arXiv preprint"},{"key":"ref103","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA59077.2024.00040"},{"key":"ref104","doi-asserted-by":"publisher","DOI":"10.1145\/3676641.3716025"},{"key":"ref105","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA61900.2025.00102"},{"key":"ref106","doi-asserted-by":"publisher","DOI":"10.1145\/3695053.3731411"},{"key":"ref107","volume-title":"Breaking Down the AI server data center cost","author":"Uvation","year":"2025"},{"key":"ref108","volume-title":"Liquid vs. Air Cooling. Which is the Capex winner?","author":"Torell","year":"2025"},{"key":"ref109","article-title":"Attention is All You Need","author":"Vaswani","year":"2017","journal-title":"Advances in Neural Information Processing Systems (NeurIPS)"},{"key":"ref110","article-title":"Step-3 is large yet affordable: Model-system co-design for cost-effective decoding","author":"Wang","year":"2025","journal-title":"arXiv preprint"},{"key":"ref111","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA59077.2024.00041"},{"key":"ref112","doi-asserted-by":"publisher","DOI":"10.1109\/PACT.2009.34"},{"key":"ref113","volume-title":"How Amazon Achieved Rocketing Sales and Growth From AI","author":"Wheeler","year":"2025"},{"key":"ref114","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA.2016.48"},{"key":"ref115","article-title":"Google\u2019s Neural Machine Translation System: Bridging the Gap between Human and Machine Translation","volume":"abs\/1609.08144","author":"Wu","year":"2016","journal-title":"CoRR"},{"key":"ref116","doi-asserted-by":"publisher","DOI":"10.1145\/3669940.3707239"},{"key":"ref117","article-title":"A survey on knowledge distillation of large language models","author":"Xu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref118","article-title":"Orca: A Distributed Serving System for Transformer-Based Generative Models","volume-title":"Proceedings of the 16th USENIX Symposium on Operating Systems Design and Implementation (OSDI)","author":"Yu","year":"2022"},{"key":"ref119","volume-title":"LLM Inference Unveiled: Survey and Roofline Model Insights","author":"Yuan","year":"2024"},{"key":"ref120","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA52012.2021.00033"},{"key":"ref121","volume-title":"How Much Does it Cost to Build a Data Center","author":"Zhang","year":"2025"},{"key":"ref122","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2024.emnlp-main.498","volume-title":"A Comprehensive Survey of Scientific Large Language Models and Their Applications in Scientific Discovery","author":"Zhang","year":"2024"},{"key":"ref123","first-page":"196","article-title":"Atom: Low-bit quantization for efficient and accurate llm serving","volume-title":"Proceedings of Machine Learning and Systems","volume":"6","author":"Zhao","year":"2024"},{"key":"ref124","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA59077.2024.00077"},{"key":"ref125","article-title":"DistServe: Disaggregating Prefill and Decoding for Goodput-optimized Large Language Model Serving","volume-title":"18th USENIX Symposium on Operating Systems Design and Implementation (OSDI)","author":"Zhong","year":"2024"},{"key":"ref126","article-title":"PetS: A Unified Framework for Parameter-Efficient Transformers Serving","volume-title":"Proceedings of the USENIX Annual Technical Conference (USENIX ATC)","author":"Zhou","year":"2022"},{"key":"ref127","article-title":"Megascale-infer: Serving mixtureof-experts at scale with disaggregated expert parallelism","author":"Zhu","year":"2025","journal-title":"arXiv preprint"},{"key":"ref128","volume-title":"Artificial Intelligence (AI) Market Size and Growth 2025 to 2034","author":"Zoting","year":"2025"}],"event":{"name":"2026 ACM\/IEEE 53rd Annual International Symposium on Computer Architecture (ISCA)","location":"Raleigh, NC, USA","start":{"date-parts":[[2026,6,27]]},"end":{"date-parts":[[2026,7,1]]}},"container-title":["2026 ACM\/IEEE 53rd Annual International Symposium on Computer Architecture (ISCA)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11617393\/11617394\/11617794.pdf?arnumber=11617794","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T05:52:35Z","timestamp":1785909155000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11617794\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6]]},"references-count":128,"URL":"https:\/\/doi.org\/10.1109\/isca66397.2026.00029","relation":{},"subject":[],"published":{"date-parts":[[2026,6]]}}}