{"684611":{"#nid":"684611","#data":{"type":"news","title":"AI\u2019s Ballooning Energy Consumption Puts Spotlight On Data Center\u00a0Efficiency","body":[{"value":"\u003Cdiv\u003E\u003Cp\u003EArtificial intelligence is growing fast, and so are the number of computers that power it. Behind the scenes, this rapid growth is putting a huge strain on the data centers that run AI models. These facilities are using \u003Ca href=\u0022https:\/\/theconversation.com\/ai-supercharges-data-center-energy-use-straining-the-grid-and-slowing-sustainability-efforts-232697\u0022\u003Emore energy than ever\u003C\/a\u003E.\u003C\/p\u003E\u003Cp\u003EAI models are getting larger and more complex. Today\u2019s most advanced systems have billions of parameters, the numerical values derived from training data, and run across thousands of computer chips. To keep up, companies have responded by adding more hardware, more chips, more memory and more powerful networks. This brute force approach has helped AI make big leaps, but it\u2019s also created a new challenge: Data centers are becoming energy-hungry giants.\u003C\/p\u003E\u003Cp\u003ESome tech companies are responding by looking to \u003Ca href=\u0022https:\/\/blog.ucs.org\/mike-jacobs\/power-hungry-why-data-centers-are-developing-their-own-energy-sources-to-fuel-ai\/\u0022\u003Epower data centers on their own\u003C\/a\u003E with \u003Ca href=\u0022https:\/\/techcrunch.com\/2025\/08\/21\/gas-power-plants-approved-for-metas-10b-data-center-and-not-everyone-is-happy\/\u0022\u003Efossil fuel\u003C\/a\u003E and \u003Ca href=\u0022https:\/\/www.techradar.com\/pro\/google-is-building-a-small-nuclear-reactor-in-tennessee-to-power-its-data-centers\u0022\u003Enuclear power plants\u003C\/a\u003E. AI energy demand has also spurred efforts to make \u003Ca href=\u0022https:\/\/www.wsj.com\/tech\/ai\/the-new-chips-designed-to-solve-ais-energy-problem-1ba9cac1\u0022\u003Emore efficient computer chips\u003C\/a\u003E.\u003C\/p\u003E\u003Cp\u003EI\u2019m a \u003Ca href=\u0022https:\/\/scholar.google.com\/citations?hl=en\u0026amp;user=HmBa_6gAAAAJ\u0026amp;view_op=list_works\u0026amp;sortby=pubdate\u0022\u003Ecomputer engineer\u003C\/a\u003E and a professor \u003Ca href=\u0022https:\/\/ece.gatech.edu\/directory\/divya-mahajan\u0022\u003Eat Georgia Tech\u003C\/a\u003E who specializes in high-performance computing. I see another path to curbing AI\u2019s energy appetite: Make data centers more resource aware and efficient.\u003C\/p\u003E\u003Ch2\u003EEnergy and Heat\u003C\/h2\u003E\u003Cp\u003EModern AI data centers can use as much electricity as a \u003Ca href=\u0022https:\/\/www.cnbc.com\/2024\/11\/23\/data-centers-powering-ai-could-use-more-electricity-than-entire-cities.html\u0022\u003Esmall city\u003C\/a\u003E. And it\u2019s not just the computing that eats up power. Memory and cooling systems are major contributors, too. As AI models grow, they need more storage and faster access to data, which generates more heat. Also, as the chips become more powerful, removing heat becomes a central challenge.\u003C\/p\u003E\u003Cp\u003E\u003Ca href=\u0022https:\/\/images.theconversation.com\/files\/688784\/original\/file-20250902-56-s2j1vb.jpg?ixlib=rb-4.1.0\u0026amp;q=45\u0026amp;auto=format\u0026amp;w=1000\u0026amp;fit=clip\u0022\u003E\u003Cimg src=\u0022https:\/\/images.theconversation.com\/files\/688784\/original\/file-20250902-56-s2j1vb.jpg?ixlib=rb-4.1.0\u0026amp;q=45\u0026amp;auto=format\u0026amp;w=754\u0026amp;fit=clip\u0022 alt=\u0022Small blue and green lights arranged in columns glow behind black mesh screens\u0022\u003E\u003C\/a\u003E\u003C\/p\u003E\u003Cp\u003EData centers house thousands of interconnected computers. \u003Ca href=\u0022https:\/\/www.gettyimages.com\/detail\/news-photo\/data-servers-during-an-open-day-at-the-digital-realty-data-news-photo\/1475272476\u0022\u003EAlberto Ortega\/Europa Press via Getty Images\u003C\/a\u003E\u003C\/p\u003E\u003Cp\u003ECooling isn\u2019t just a technical detail; it\u2019s a major part of the energy bill. Traditional cooling is done with specialized air conditioning systems that remove heat from server racks. New methods like \u003Ca href=\u0022https:\/\/blogs.nvidia.com\/blog\/blackwell-platform-water-efficiency-liquid-cooling-data-centers-ai-factories\/\u0022\u003Eliquid cooling\u003C\/a\u003E are helping, but they also require careful planning and water management. Without smarter solutions, the energy requirements and costs of AI could become unsustainable.\u003C\/p\u003E\u003Cp\u003EEven with all this advanced equipment, many data centers aren\u2019t running efficiently. That\u2019s because \u003Ca href=\u0022https:\/\/proceedings.mlr.press\/v235\/wang24bp.html\u0022\u003Edifferent parts of the system don\u2019t always talk\u003C\/a\u003E to each other. For example, scheduling software might not know that a chip is overheating or that a network connection is clogged. As a result, some servers sit idle while others struggle to keep up. This lack of coordination can lead to wasted energy and underused resources.\u003C\/p\u003E\u003Ch2\u003EA Smarter Way Forward\u003C\/h2\u003E\u003Cp\u003EAddressing this challenge requires rethinking how to design and manage the systems that support AI. That means moving away from brute-force scaling and toward smarter, more specialized infrastructure.\u003C\/p\u003E\u003Cp\u003EHere are three key ideas:\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EAddress variability in hardware.\u003C\/strong\u003E Not all chips are the same. Even within the same generation, chips vary in how fast they operate and how much heat they can tolerate, leading to heterogeneity in both performance and energy efficiency. Computer systems in data centers should recognize differences among chips in performance, heat tolerance and energy use, and adjust accordingly.\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EAdapt to changing conditions.\u003C\/strong\u003E AI workloads vary over time. For instance, thermal hotspots on chips can trigger the chips to slow down, fluctuating grid supply can cap the peak power that centers can draw, and bursts of data between chips can create congestion in the network that connects them. Systems should be designed to respond in real time to things like temperature, \u003Ca href=\u0022https:\/\/www.canarymedia.com\/articles\/utilities\/google-ai-data-center-flexibility-help-grid\u0022\u003Epower availability\u003C\/a\u003E and data traffic.\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EBreak down silos.\u003C\/strong\u003E Engineers who design chips, software and data centers should work together. When these teams collaborate, they can find new ways to save energy and improve performance. To that end, my colleagues, students and I at Georgia Tech\u2019s \u003Ca href=\u0022https:\/\/coe.gatech.edu\/academics\/ai-for-engineering\/ai-makerspace\u0022\u003EAI Makerspace\u003C\/a\u003E, a high-performance AI data center, are exploring these challenges hands-on. We\u2019re working across disciplines, from hardware to software to energy systems, to build and test AI systems that are efficient, scalable and sustainable.\u003C\/p\u003E\u003Ch2\u003EScaling With Intelligence\u003C\/h2\u003E\u003Cp\u003EAI has the potential to transform science, medicine, education and more, but risks hitting limits on performance, energy and cost. The future of AI depends not only on better models, but also on better infrastructure.\u003C\/p\u003E\u003Cp\u003ETo keep AI growing in a way that benefits society, I believe it\u2019s important to shift from scaling by force to scaling with intelligence.\u003C!-- Below is The Conversation\u0027s page counter tag. Please DO NOT REMOVE. --\u003E\u003Cimg src=\u0022https:\/\/counter.theconversation.com\/content\/254192\/count.gif?distributor=republish-lightbox-basic\u0022 alt=\u0022The Conversation\u0022 width=\u00221\u0022 height=\u00221\u0022\u003E\u003C!-- End of code. If you don\u0027t see any code above, please get new code from the Advanced tab after you click the republish button. The page counter does not collect any personal data. More info: https:\/\/theconversation.com\/republishing-guidelines --\u003E\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u003Cem\u003EThis article is republished from \u003C\/em\u003E\u003Ca href=\u0022https:\/\/theconversation.com\u0022\u003E\u003Cem\u003EThe Conversation\u003C\/em\u003E\u003C\/a\u003E\u003Cem\u003E under a Creative Commons license. Read the \u003C\/em\u003E\u003Ca href=\u0022https:\/\/theconversation.com\/ais-ballooning-energy-consumption-puts-spotlight-on-data-center-efficiency-254192\u0022\u003E\u003Cem\u003Eoriginal article\u003C\/em\u003E\u003C\/a\u003E\u003Cem\u003E.\u003C\/em\u003E\u003C\/p\u003E\u003C\/div\u003E","summary":"","format":"full_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003EArtificial intelligence is growing fast, and so are the number of computers that power it. Behind the scenes, this rapid growth is putting a huge strain on the data centers that run AI models.\u0026nbsp;\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Artificial intelligence is growing fast, and so are the number of computers that power it. Behind the scenes, this rapid growth is putting a huge strain on the data centers that run AI models. "}],"uid":"27469","created_gmt":"2025-09-08 13:34:41","changed_gmt":"2026-03-19 13:11:42","author":"Kristen Bailey","boilerplate_text":"","field_publication":"","field_article_url":"","location":"Atlanta, GA","dateline":{"date":"2025-09-03T00:00:00-04:00","iso_date":"2025-09-03T00:00:00-04:00","tz":"America\/New_York"},"extras":[],"hg_media":{"677940":{"id":"677940","type":"image","title":"These \u2018chillers\u2019 on the roof of a data center in Germany, seen from above, work to cool the equipment inside the building. ","body":"\u003Cp\u003EThese \u2018chillers\u2019 on the roof of a data center in Germany, seen from above, work to cool the equipment inside the building. \u003Ca href=\u0022https:\/\/newsroom.ap.org\/detail\/GermanyDataCenter\/ff354b47c6a34682b8a76f9ca89613ed\/photo\u0022\u003EAP Photo\/Michael Probst\u003C\/a\u003E\u003C\/p\u003E","created":"1757338556","gmt_created":"2025-09-08 13:35:56","changed":"1757338556","gmt_changed":"2025-09-08 13:35:56","alt":"These \u2018chillers\u2019 on the roof of a data center in Germany, seen from above, work to cool the equipment inside the building. ","file":{"fid":"261898","name":"file-20250902-56-fgh9og.jpg","image_path":"\/sites\/default\/files\/2025\/09\/08\/file-20250902-56-fgh9og.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2025\/09\/08\/file-20250902-56-fgh9og.jpg","mime":"image\/jpeg","size":547761,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2025\/09\/08\/file-20250902-56-fgh9og.jpg?itok=LJ6j5HUe"}}},"media_ids":["677940"],"related_links":[{"url":"https:\/\/theconversation.com\/ais-ballooning-energy-consumption-puts-spotlight-on-data-center-efficiency-254192","title":"Read This Article on The Conversation"}],"groups":[{"id":"658168","name":"Experts"},{"id":"1214","name":"News Room"},{"id":"1188","name":"Research Horizons"}],"categories":[{"id":"194606","name":"Artificial Intelligence"}],"keywords":[{"id":"187915","name":"go-researchnews"},{"id":"192863","name":"go-ai"}],"core_research_areas":[{"id":"193655","name":"Artificial Intelligence at Georgia Tech"}],"news_room_topics":[{"id":"71881","name":"Science and Technology"}],"event_categories":[],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Ch5\u003EAuthor:\u003C\/h5\u003E\u003Cp\u003E\u003Ca href=\u0022https:\/\/theconversation.com\/profiles\/divya-mahajan-2366440\u0022 rel=\u0022author\u0022\u003EDivya Mahajan\u003C\/a\u003E, assistant professor of Computer Engineering, Georgia Institute of Technology\u003C\/p\u003E\u003Ch5\u003EMedia Contact:\u003C\/h5\u003E\u003Cp\u003EShelley Wunder-Smith\u003Cbr\u003E\u003Ca href=\u0022mailto:shelley.wunder-smith@research.gatech.edu\u0022\u003Eshelley.wunder-smith@research.gatech.edu\u003C\/a\u003E\u003C\/p\u003E","format":"limited_html"}],"email":[],"slides":[],"orientation":[],"userdata":""}}}