{"id":50864,"date":"2025-11-24T21:30:07","date_gmt":"2025-11-24T21:30:07","guid":{"rendered":"https:\/\/www.newsbeep.com\/us-ny\/50864\/"},"modified":"2025-11-24T21:30:07","modified_gmt":"2025-11-24T21:30:07","slug":"meta-is-building-a-data-center-the-size-of-manhattan-how-cool-is-that","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/us-ny\/50864\/","title":{"rendered":"Meta is building a data center the size of Manhattan. How cool is that?"},"content":{"rendered":"<p>Meta is\u00a0spending hundreds of\u00a0billions of dollars on\u00a0a string of\u00a0<a href=\"https:\/\/datacentrereview.com\/2025\/07\/meta-earmarks-hundreds-of-billions-for-multi-gigawatt-ai-data-centres\/\" target=\"_blank\" rel=\"noopener nofollow\">multi gigawatt AI\u00a0data\u00a0centers<\/a>\u00a0that will be\u00a0less like traditional business\u00a0facilities, but\u00a0cities in their own right.\u00a0\u00a0<\/p>\n<p>That\u2019s\u00a0not\u00a0hyperbole.\u00a0<a href=\"https:\/\/d.docs.live.net\/033b01ce9bef3050\/freelance%20work\/spa\/edge%20dcs%20background.docx\" target=\"_blank\" rel=\"noopener nofollow\">Mark Zuckerberg<\/a>\u00a0in July\u00a0said that its Hyperion data\u00a0center\u00a0would \u201cbe able to scale up to 5GW over several years, and \u201cmultiple more\u201d of these titan clusters\u00a0would be built.\u00a0<\/p>\n<p>\u201cJust one of these covers a significant part of the footprint of Manhattan,\u201d the Facebook founder\u00a0said. To prove his point, he\u00a0posted\u00a0a gif that\u00a0illustrated precisely how one\u00a0of these\u00a0massive\u00a0facilities would fit across the\u00a0sliver of land that is home to 1.66 million people.\u00a0<\/p>\n<p>But\u00a0if\u00a0running\u00a0a city like\u00a0Manhattan\u00a0is a challenge,\u00a0keeping\u00a0it\u00a0cool is even harder.\u00a0\u00a0<\/p>\n<p>Manhattan is a classic example of the phenomenon where cities are several degrees hotter than surrounding areas, thanks to the city\u2019s \u201cbuilt\u00a0environment.\u201d\u00a0[<a href=\"https:\/\/journalism.columbia.edu\/news\/urban-heat-island-effect-making-new-yorkers-hotter-study-finds\" target=\"_blank\" rel=\"noopener nofollow\">urban heat island effect<\/a>]\u00a0<\/p>\n<p>And\u00a0that\u2019s\u00a0without running millions of GPUs\u00a0and associated equipment.\u00a0<\/p>\n<p>We know that data\u00a0center\u00a0operators\u00a0already face a massive challenge when it comes to managing the heat their facilities produce \u2013 and\u00a0calming\u00a0the public\u2019s concern over their\u00a0potential\u00a0environmental impact.\u00a0<\/p>\n<p>So, how should Meta approach the cooling challenge\u00a0for these truly titanic data\u00a0centers?\u00a0Here are some thoughts based\u00a0on our experience\u00a0of cooling at scale.\u00a0<\/p>\n<p>Calculating the options\u00a0<\/p>\n<p>Current\u00a0cutting-edge\u00a0data\u00a0center\u00a0designs stretch into the 100s of\u00a0kWs, and\u00a0we can safely assume that Meta is laying its plans with\u00a0<a href=\"https:\/\/developer.nvidia.com\/blog\/nvidia-800-v-hvdc-architecture-will-power-the-next-generation-of-ai-factories\/%5d\" target=\"_blank\" rel=\"noopener nofollow\">NVIDIA\u2019s<\/a>\u00a0roadmap in mind, and that envisions 1MW racks.\u00a0\u00a0<\/p>\n<p>Traditional evaporative cooling is\u00a0considered\u00a0a non-starter for data\u00a0centers\u00a0of this scale\u00a0except for peripheral activities.\u00a0Just the amount of water needed would be astronomical, tens of millions of Olympic sized swimming pools.\u00a0<\/p>\n<p>So liquid cooling is the most\u00a0practical\u00a0option\u00a0for\u00a0the majority of\u00a0cooling in these\u00a0next\u00a0generation\u00a0\u00a0data\u00a0centers.\u00a0\u00a0<\/p>\n<p>But what type of liquid cooling?\u00a0Immersion\u00a0does have an edge when it comes to\u00a0pure\u00a0cooling potential.\u00a0But is complex and\u00a0relatively inflexible.\u00a0Direct to chip is the\u00a0most likely option\u00a0here. Indeed,\u00a0NVIDIA\u2019s\u00a0own reference designs lean towards\u00a0direct to chip\u00a0liquid cooling for its\u00a0cutting-edge\u00a0designs.\u00a0\u00a0\u00a0<\/p>\n<p>Opting for\u00a0direct to chip\u00a0liquid cooling is the easy part though.\u00a0Meta\u2019s engineers\u00a0then\u00a0face the challenge of implementing it at titanic scale.\u00a0Taking the right approach from the outset could make\u00a0their\u00a0lives\u00a0easier, for years into the future.\u00a0<\/p>\n<p>Just consider the amount of kit involved. Whether you do some\u00a0sums on the back of a napkin or\u00a0drop\u00a0a few prompts into ChatGPT,\u00a0Meta\u2019s planned data\u00a0centers\u00a0are likely to house\u00a0between one and two million GPUs spread across ten to twenty thousand racks,\u00a0along\u00a0with associated networking and storage.\u00a0<\/p>\n<p>That is an incredible amount of\u00a0equipment\u00a0to\u00a0deploy and\u00a0manage.\u00a0And a correspondingly massive amount of cooling\u00a0infrastructure\u00a0to deploy and\u00a0maintain\u00a0too.\u00a0And there will\u00a0likely be\u00a0failures or upgrades that require components to be replaced\u00a0over time.\u00a0<\/p>\n<p>So,\u00a0when designing this cooling infrastructure,\u00a0Meta will need to consider not just the amount\u00a0of heat\u00a0it must shift today, or next year, but generations of silicon into the future.\u00a0<\/p>\n<p>This means\u00a0Meta will\u00a0need\u00a0a cooling solution, or solutions,\u00a0which\u00a0doesn\u2019t\u00a0just offer raw cooling capability.\u00a0It needs a liquid cooling approach that\u00a0can\u00a0evolve alongside its plans.\u00a0\u00a0<\/p>\n<p>Even someone with the vast resources of Meta\u00a0won\u2019t\u00a0want to tie-up capital or delay deploying\u00a0racks while it for\u00a0cooling components\u00a0to be delivered. But neither will it want\u00a0cooling equipment\u00a0sat\u00a0in storage, waiting to be deployed, or\u00a0in place, but\u00a0underutilized.\u00a0\u00a0<\/p>\n<p>It needs\u00a0partners\u00a0who can supply at scale, continuously, and which have\u00a0an engineering ecosystem capable of deploying and install cooling at the same pace it rolls off the production line.\u00a0<\/p>\n<p>Even better,\u00a0Meta\u00a0should\u00a0choose\u00a0a\u00a0liquid cooling architecture\u00a0that can be deployed incrementally. Why wait until the whole of Manhattan is covered in\u00a0racks before flicking the on\u00a0switch?\u00a0Why not start serving customers\u00a0or refining models\u00a0when\u00a0the equivalent of Central Park is ready to go online.\u00a0<\/p>\n<p>A liquid vision?\u00a0<\/p>\n<p>That\u2019s\u00a0just the start of the data\u00a0center\u00a0lifecycle though.\u00a0From that point on manageability is key. The racks of compute and storage\u00a0in a data\u00a0center\u00a0are\u00a0designed with redundancy\u00a0in mind. Components are standardized and\u00a0hot-swappable\u00a0as far as possible.\u00a0so,\u00a0Zuckerberg\u2019s engineers should ensure the same\u00a0applies\u00a0across the infrastructure keeping it all cool.\u00a0\u00a0<\/p>\n<p>A decoupled system \u2013 where the control console is a in a separate unit for example \u2013 will give Meta more options when it comes to scaling up over time.\u00a0If there are faults in one\u00a0component, it\u00a0won\u2019t\u00a0need to replace the entire unit.\u00a0So, cooling units and\u00a0individual\u00a0components, such as pumps or sensors, should be\u00a0specced\u00a0in the same way.\u00a0\u00a0<\/p>\n<p>And of course, this will be all for nothing, if those components are not easily accessible \u2013 and spares readily available.\u00a0Front access will be\u00a0critical and\u00a0will also allow more flexibility when it\u00a0come\u00a0to positioning units and designing\u00a0the aisles.\u00a0<\/p>\n<p>Presumably,\u00a0Meta\u00a0will have the brain and compute power to develop\u00a0cutting edge\u00a0predictive\u00a0analytics\u00a0to manage maintenance.\u00a0But it will help immensely if the cooling architecture it chooses can serve up the right, real time information\u00a0to inform those systems.\u00a0<\/p>\n<p>At a broader level,\u00a0the\u00a0heat\u00a0the cooling system extracts\u00a0must\u00a0still\u00a0go somewhere,\u00a0Municipal\u00a0authorities may\u00a0be grateful for the jobs a data\u00a0center\u00a0will\u00a0bring but\u00a0may not\u00a0he\u00a0happy if it\u00a0creates\u00a0microclimates in the immediate area. Environmental groups\u00a0and\u00a0regulators\u00a0will be\u00a0on the watch for any adverse environmental impact.\u00a0<\/p>\n<p>Meta should be thinking\u00a0from the outset about how this\u00a0energy can be repurposed, whether for district heat systems\u00a0or other industrial uses\u00a0such as factories or heating greenhouses.\u00a0A cooling partner with\u00a0a broader vision of how heat can be reused,\u00a0and\u00a0the engineering skills to\u00a0make\u00a0this\u00a0happen, will be\u00a0a great help\u00a0here too.\u00a0<\/p>\n<p>Meta\u2019s datacentre ambitions\u00a0are more than just\u00a0a massive test of its resolve to dominate AI.\u00a0They\u00a0represent\u00a0an engineering bet of truly titanic proportions. And that means they\u00a0also present the biggest test yet for the liquid cooling sector.\u00a0It\u2019s\u00a0in all our\u00a0interests that\u00a0it passes.\u00a0<\/p>\n","protected":false},"excerpt":{"rendered":"Meta is\u00a0spending hundreds of\u00a0billions of dollars on\u00a0a string of\u00a0multi gigawatt AI\u00a0data\u00a0centers\u00a0that will be\u00a0less like traditional business\u00a0facilities, but\u00a0cities in&hellip;\n","protected":false},"author":2,"featured_media":50865,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[35],"tags":[75,84,83,9,24,63],"class_list":["post-50864","post","type-post","status-publish","format-standard","has-post-thumbnail","category-manhattan","tag-manhattan","tag-manhattan-headlines","tag-manhattan-news","tag-new-york","tag-new-york-city","tag-nyc"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/us-ny\/wp-json\/wp\/v2\/posts\/50864","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/us-ny\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/us-ny\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-ny\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-ny\/wp-json\/wp\/v2\/comments?post=50864"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/us-ny\/wp-json\/wp\/v2\/posts\/50864\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-ny\/wp-json\/wp\/v2\/media\/50865"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/us-ny\/wp-json\/wp\/v2\/media?parent=50864"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-ny\/wp-json\/wp\/v2\/categories?post=50864"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-ny\/wp-json\/wp\/v2\/tags?post=50864"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}