[{"data":1,"prerenderedAt":129},["ShallowReactive",2],{"9XKH4Afqp1x2AI6ZZXwyAZsiysaDeLGj58Yl2RF1sUo":3,"_apollo:default":127,"_apollo:identified":128},{"allSettings":4,"seo":10,"posts":21},{"customSettings":5,"__typename":9},{"urlFront":6,"urlPreview":7,"__typename":8},"https:\u002F\u002Fblog.3ds.com","https:\u002F\u002Fblog-build-prd.itvpc.3ds.com","CustomSettings","Settings",{"social":11,"openGraph":17,"__typename":20},{"twitter":12,"__typename":16},{"cardType":13,"username":14,"__typename":15},"summary_large_image","dassault3ds","SEOSocialTwitter","SEOSocial",{"defaultImage":18,"__typename":19},null,"SEOOpenGraph","SEOConfig",{"nodes":22,"__typename":126},[23],{"id":24,"slug":25,"title":26,"uri":27,"excerpt":28,"locale":29,"featuredImage":32,"tableOfContents":40,"content":45,"date":46,"translations":47,"author":48,"tags":61,"globalTags":69,"brands":77,"industries":80,"keywords":108,"seo":114,"__typename":125},"cG9zdDozMTcyODI=","virtual-twins-improve-ai-factories","How Can Virtual Twins Improve AI Factories?","\u002Findustries\u002Fhigh-tech\u002Fvirtual-twins-improve-ai-factories","\u003Cp>Dassault Systèmes and Vertiv collaborate to advance the model-based design, engineering and validation of next-generation AI infrastructure. \u003C\u002Fp>\n",{"locale":30,"__typename":31},"en_US","Locale",{"node":33,"__typename":39},{"large":34,"__typename":35,"medium_large":34,"thumbnail":36,"srcSet":37,"sizes":38},"https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002Fvertiv-hero_withbanner.png","MediaItem","https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002Fvertiv-hero_withbanner-150x150.png","https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002Fvertiv-hero_withbanner-300x169.png 300w, https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002Fvertiv-hero_withbanner.png 580w","(max-width: 300px) 100vw, 300px","NodeWithFeaturedImageToMediaItemConnectionEdge",[41,42,43,44],"Why Do AI Factories Need to be More Efficient?|why-do-ai-factories-need-to-be-more-efficient-0","How Does the Dassault Systèmes-Vertiv-NVIDIA Collaboration Work?|how-does-the-dassault-systemes-vertiv-nvidia-collaboration-work-1","The MBSE Difference|the-mbse-difference-2","Related Content on Industrial AI|related-content-on-industrial-ai-3","\n\u003Cp>AI factories are among the most complex systems ever built. Unlike traditional factories that build discrete physical objects, AI factories continuously manufacture intelligence, in the form of tokens. And unlike traditional data centers, AI factories don’t just store and process data for general computing, they’re specialized to train and operate artificial intelligence at scale. \u003C\u002Fp>\n\n\n\n\u003Cp>They’re dedicated infrastructure combining data center facility, power, cooling, networking, compute, software and controls optimized for AI workloads such as training, inference, agents and generative experiences. Within these highly coupled systems, changes in one domain can have significant downstream effects on others.\u003C\u002Fp>\n\n\n\n\u003Cp>With enterprises racing to gain the most value possible from Industrial AI deployments, AI factories have become essential infrastructure for scaling. As such, engineers must be able to validate how a complete infrastructure system will perform, before it is built and deployed. To accelerate systems engineering of the next generation of AI factories, Dassault Systèmes and digital infrastructure leader \u003Ca href=\"https:\u002F\u002Fwww.vertiv.com\u002Fen-us\u002F\" target=\"_blank\" rel=\"noreferrer noopener\">Vertiv\u003C\u002Fa> have announced plans to expand our collaboration showcased at Computex 2026, leveraging the \u003Ca href=\"https:\u002F\u002Fwww.3ds.com\u002Fcloud\u002F3dexperience-evaluation-offer?utm_medium=cpc&utm_source=google&utm_campaign=202601_glo_sea_en_op51508_labl_brand-topV2_nam_all&utm_term=3dexperience-phrase&utm_content=search&gad_source=1&gad_campaignid=23907835644&gbraid=0AAAAADqOeJRLFo5bFaSV2tECz4hzg-6El&gclid=CjwKCAjwotzWBhBFEiwA1JhPoNm_j9jEcCXjPP3LFQg3BZrgGttn2jZ9q6KLd84E0Fese9Ahe3ZG-BoCIIwQAvD_BwE\">\u003Cstrong>3D\u003C\u002Fstrong>EXPERIENCE platform\u003C\u002Fa> with Vertiv’s converged physical infrastructure expertise to enable AI factory infrastructure to be configured, simulated and validated before construction begins.\u003C\u002Fp>\n\n\n\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fwww.vertiv.com\u002Fen-us\u002Fproducts-catalog\u002Ffacilities-enclosures-and-racks\u002Fintegrated-solutions\u002Fvertiv-onecore-prefabricated-hybrid-built-data-center\u002F\">Vertiv OneCore \u003C\u002Fa>is a hybrid-built data center solution that integrates critical power, thermal management and supporting infrastructure. Its manufactured design simplifies shipping, installation and project execution, while improving reliability, cost efficiency and predictability. The \u003Cstrong>3D\u003C\u002Fstrong>EXPERIENCE platform’s model-based systems engineering, simulation and virtual twin capabilities can represent Vertiv infrastructure as an integrated digital system, providing traceability and visibility into the impacts of design changes.\u003C\u002Fp>\n\n\n\n\u003Cp>This collaboration, demonstrated at \u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fgtc\u002F\">NVIDIA GTC\u003C\u002Fa> Berlin, builds upon the companies’ respective partnerships with NVIDIA and integration with the \u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fdata-center\u002Fproducts\u002Fdsx\u002F\">NVIDIA DSX\u003C\u002Fa> ecosystem. It will advance the model-based design, engineering and validation of AI factory infrastructure. The goal: Improve AI factory performance by accelerating time to first token, increasing productive compute from available power and reducing energy use.\u003C\u002Fp>\n\n\n\n\u003Cp>As stated in the announcement:\u003C\u002Fp>\n\n\n\n\u003Cp>Targeted customer outcomes include accelerating the path from requirements to a configured and validated AI factory solution by up to 10x and cutting the defined infrastructure deployment period by up to 50%. Under the assumptions in Vertiv’s model, the combined approach can bring capacity online approximately 10 weeks sooner than a modeled conventional baseline, producing up to 30% more cumulative token output within 12 months of project kickoff at the same installed capacity. The gain results solely from earlier facility readiness and does not reflect improved operating efficiency or a measured customer result.\u003C\u002Fp>\n\n\n\n\u003Ch2 class=\"wp-block-heading\" id=\"why-do-ai-factories-need-to-be-more-efficient-0\">\u003Cstrong>Why Do AI Factories Need to be More Efficient?\u003C\u002Fstrong>\u003C\u002Fh2>\n\n\n\n\u003Cp>Increasing demand for \u003Ca href=\"https:\u002F\u002Fwww.3ds.com\u002Findustrial-ai\">Industrial AI\u003C\u002Fa> and autonomous agents increases demand on—and for—AI factories, but turning energy and data into AI models and tokens is no small feat. AI factories are hugely complex and a huge investment: energy and cooling are major infrastructure and operating costs, not to mention the expense of the hardware itself. That was true for the first generation of one-off AI factories, and with the landscape evolving to neoclouds, efficient use of power and time-to-token is essential for the economics to work.\u003C\u002Fp>\n\n\n\n\u003Cp>Operating efficiently directly determines how much AI you can produce, which products are economically viable at scale and how cheaply they can be offered. But an issue with the infrastructure means you’re essentially paying for compute power you aren’t using. When you’re creating value out of data by industrializing AI, idle hardware is wasted capital.\u003C\u002Fp>\n\n\n\n\u003Cp>The \u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fdata-center\u002Fproducts\u002Fdsx\u002F\">NVIDIA\u003C\u002Fa>\u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fdata-center\u002Fproducts\u002Fdsx\u002F\"> \u003C\u002Fa>\u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fdata-center\u002Fproducts\u002Fdsx\u002F\">DSX \u003C\u002Fa>\u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fdata-center\u002Fproducts\u002Fdsx\u002F\">AI \u003C\u002Fa>\u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fdata-center\u002Fproducts\u002Fdsx\u002F\">F\u003C\u002Fa>\u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fdata-center\u002Fproducts\u002Fdsx\u002F\">actory \u003C\u002Fa>\u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fdata-center\u002Fproducts\u002Fdsx\u002F\">Platform\u003C\u002Fa> brings together reference designs, software and partner technologies to help design, build and operate AI factories as integrated systems spanning compute, networking, storage, power and cooling.\u003C\u002Fp>\n\n\n\n\u003Cp>Configurable AI factory virtual twins make it possible to optimize designs through systems engineering and simulation. This bridges the gap between concept and deployment, maximizing token performance per megawatt. And with Vertiv infrastructure building blocks designed to align with the \u003Ca href=\"https:\u002F\u002Fdocs.nvidia.com\u002Fdsx?_gl=1*1vt5pt6*_gcl_aw*R0NMLjE3ODU3Nzg2ODEuQ2owS0NRanctTURUQmhDZ0FSSXNBS0FrZGxTQUNmSmp6b1NLRVFmdXhmLUJOUFhJN0wwbkFkaXNyLTVPQ1p6cEh1Z09nYXdRT0JReEdjQWFBdTFwRUFMd193Y0I.*_gcl_au*OTgwMTk4NzI4LjE3ODc3MDE5NjQuLS4tLjE3ODc3MDE5NjQuOTkwNjE2NjA0LjE3OTI0MjI4NTMuMTc5MjQ1Mjc3NQ..#dsx-reference-designs?ncid=no-ncid\">NVIDIA DSX Reference Design\u003C\u002Fa>, part of \u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fdata-center\u002Fproducts\u002Fdsx\u002Fsim\u002F\">NVIDIA DSX Sim\u003C\u002Fa>, on the \u003Cstrong>3D\u003C\u002Fstrong>EXPERIENCE platform, engineers will be able to validate infrastructure systems before they are built, for greater confidence in performance that accelerates time to first token, increases productive compute from available power and reduces energy use.\u003C\u002Fp>\n\n\n\n\u003Cfigure class=\"wp-block-image size-full\">\u003Cimg loading=\"lazy\" decoding=\"async\" width=\"580\" height=\"326\" src=\"https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002Fvertiv_01_hd_withbanner.jpg\" alt=\"virtual twin of an AI factory - AI factories - Dassault Systemes blog \" class=\"wp-image-317373\" srcset=\"https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002Fvertiv_01_hd_withbanner.jpg 580w, https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002Fvertiv_01_hd_withbanner-300x169.jpg 300w\" sizes=\"auto, (max-width: 580px) 100vw, 580px\" \u002F>\u003Cfigcaption class=\"wp-element-caption\">A look at an AI factory virtual twin inside the \u003Cstrong>3D\u003C\u002Fstrong>EXPERIENCE agentic platform.\u003C\u002Ffigcaption>\u003C\u002Ffigure>\n\n\n\n\u003Ch2 class=\"wp-block-heading\" id=\"how-does-the-dassault-systemes-vertiv-nvidia-collaboration-work-1\">\u003Cstrong>How Does the Dassault Systèmes-Vertiv-NVIDIA \u003C\u002Fstrong>C\u003Cstrong>ollaboration Work?\u003C\u002Fstrong>\u003C\u002Fh2>\n\n\n\n\u003Cp>Earlier this year, \u003Ca href=\"https:\u002F\u002Fblog.3ds.com\u002Ftopics\u002Fcompany-news\u002Fai-factory-virtual-twins\u002F\">\u003C\u002Fa>\u003Ca href=\"https:\u002F\u002Fblog.3ds.com\u002Ftopics\u002Fcompany-news\u002Fai-factory-virtual-twins\u002F\">Dassault Systèmes and Vertiv showcased at Computex in Taipei a configurable virtual twin of Vertiv’s SmartRun infrastructure\u003C\u002Fa>, developed using Dassault Systèmes’ \u003Ca href=\"https:\u002F\u002Fwww.3ds.com\u002Ftechnologies\u002Fmbse-software\">model-based systems engineering (MBSE) capabilities\u003C\u002Fa> on the \u003Cstrong>3D\u003C\u002Fstrong>EXPERIENCE platform and integrated with \u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fdata-center\u002Fproducts\u002Fdsx\u002F\">NVIDIA DSX\u003C\u002Fa> and \u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fomniverse\u002F\">NVIDIA Omniverse\u003C\u002Fa> workflows.\u003C\u002Fp>\n\n\n\n\u003Cp>The newest phase of the collaboration extends this work toward the broader Vertiv OneCore infrastructure architecture for AI factories built on NVIDIA DSX:\u003C\u002Fp>\n\n\n\n\u003Cul class=\"wp-block-list\">\n\u003Cli>The NVIDIA DSX Reference Design provides a common architectural foundation and engineering guidance for designing and deploying AI factories, meaning developers can build on the reference infrastructure and adapt it to their needs.\u003C\u002Fli>\n\n\n\n\u003Cli>Vertiv provides the standardized, configurable infrastructure building blocks for AI factories\u003C\u002Fli>\n\n\n\n\u003Cli>The \u003Cstrong>3D\u003C\u002Fstrong>EXPERIENCE platform’s MBSE, simulation and virtual twin capabilities transform these blocks into a unified digital system that offers traceability and visibility into the impacts of design changes.\u003C\u002Fli>\n\u003C\u002Ful>\n\n\n\n\u003Cp>“With Vertiv and NVIDIA, we’re accelerating AI factory design and deployment by engineering the factory as one system, with compute, networking, power and cooling co-designed and co-simulated in a virtual twin,” said Stéphane Sireau, Vice President, \u003Ca href=\"https:\u002F\u002Fwww.3ds.com\u002Findustries\u002Fhigh-tech\">High-Tech Industry\u003C\u002Fa> at Dassault Systèmes. “Integration issues get resolved during design, not on site, shortening the time to token-producing capacity.”\u003C\u002Fp>\n\n\n\n\u003Ch2 class=\"wp-block-heading\" id=\"the-mbse-difference-2\">\u003Cstrong>The MBSE Difference\u003C\u002Fstrong>\u003C\u002Fh2>\n\n\n\n\u003Cp>Using the \u003Cstrong>3D\u003C\u002Fstrong>EXPERIENCE platform, Vertiv building blocks designed to align with the NVIDIA DSX Reference Design can be turned into configurable system models that serve as the starting point for customer-specific AI factory designs. Instead of locking teams into a fixed design from day one, this model-based approach gives engineers the flexibility to test different configurations, carry design changes across disciplines, validate performance at both the system and multi-physics levels and generate engineering deliverables from a single shared environment.\u003C\u002Fp>\n\n\n\n\u003Cp>Applying MBSE moves engineering closer to a more flexible, continuously co-designed and validated approach to AI factory infrastructure — one that maintains design intent from the earliest configuration work through construction, commissioning and future technology upgrades. Design decisions can be assessed within the context of an entire AI factory.\u003C\u002Fp>\n\n\n\n\u003Cp>For customers, that can mean a shorter design-to-commissioning cycle thanks to adaptable, pre-validated designs; an incremental improvement in token efficiency through pre-validated building blocks and better use of available power capacity; lower energy consumption; lower deployment risk; more predictable outcomes; and stronger digital continuity across the full infrastructure lifecycle.\u003C\u002Fp>\n\n\n\n\u003Cp>“The infrastructure system design must not be an afterthought,” said Mahesh Deshpande, senior director and global business consultant for the High-Tech Industry at Dassault Systèmes. “You can’t just think about the best compute, the best site, but think of it all as a singular, integrated system. The pre-validation ensures that, even before the components are sourced—before the first brick is laid—you have a functional, integrated system in the virtual space that’s giving you answers about what you’re trying to achieve.” \u003C\u002Fp>\n\n\n\n\u003Cp>\u003Cem>At \u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fgtc\u002F\">NVIDIA GTC\u003C\u002Fa> in Berlin, Dassault Systèmes and Vertiv showcased elements of the design and deployment approach, demonstrating how facility requirements inform infrastructure configuration, how teams can evaluate power and cooling constraints virtually and how Vertiv applies factory-built infrastructure to accelerate physical deployment.\u003C\u002Fem>\u003C\u002Fp>\n\n\n\n\u003Ch2 class=\"wp-block-heading\" id=\"related-content-on-industrial-ai-3\">Related Content on Industrial AI\u003C\u002Fh2>\n\n\n\n\u003Cul class=\"wp-block-list\">\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fblog.3ds.com\u002Ftopics\u002Fcompany-news\u002Fai-factory-virtual-twins\u002F\" data-type=\"link\" data-id=\"https:\u002F\u002Fblog.3ds.com\u002Ftopics\u002Fcompany-news\u002Fai-factory-virtual-twins\u002F\">Accelerating AI factory innovation with virtual twins\u003C\u002Fa> \u003C\u002Fli>\n\n\n\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fblog.3ds.com\u002Ftopics\u002Fcompany-news\u002Fwhat-makes-dassault-systemes-industrial-ai-unique\u002F\">What makes Dassault Systèmes’ approach to Industrial AI unique? \u003C\u002Fa>\u003C\u002Fli>\n\n\n\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fblog.3ds.com\u002Fbrands\u002Fsimulia\u002Fhow-simulation-ai-combine-power-virtual-twins\u002F\">How simulation and AI combine to power virtual twins\u003C\u002Fa> \u003C\u002Fli>\n\n\n\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.3ds.com\u002Fpartners\u002Fstrategic-partnerships\u002F\">Learn more about strategic alliances at Dassault Systèmes \u003C\u002Fa>\u003C\u002Fli>\n\n\n\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fblog.3ds.com\u002Ftopics\u002Fcompany-news\u002Fthe-advantage-of-purpose-built-ai-systems\u002F\">The advantage of purpose-built AI systems\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\n\n\n\u003Cp>\u003C\u002Fp>\n","2026-10-22T05:00:00",[],{"node":49,"__typename":60},{"nicename":50,"description":51,"slug":50,"name":52,"firstName":53,"lastName":54,"avatar":55,"__typename":59},"patrickball","Patrick is a Senior Communications Manager on the Corporate Publishing team here at Dassault Systèmes. An experienced journalist, marketer, speechwriter and storyteller, Patrick's words have appeared on pages and stages around the world.","Patrick Ball","Patrick","Ball",{"default":56,"url":57,"__typename":58},"mm","https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2022\u002F03\u002Fpatrick-ball-scaled-1-96x96.jpg","Avatar","User","NodeWithAuthorToUserConnectionEdge",{"nodes":62,"__typename":68},[63],{"id":64,"name":65,"uri":66,"__typename":67},"dGVybTo4NTUz","Company News","\u002Ftopics\u002Fcompany-news\u002F","Taxonomy_topic","PostToTaxonomy_topicConnection",{"nodes":70,"__typename":76},[71],{"id":72,"name":73,"uri":74,"__typename":75},"dGVybTo4ODMy","AI","\u002Ftags\u002Fai\u002F","Taxonomy_tag","PostToTaxonomy_tagConnection",{"nodes":78,"__typename":79},[],"PostToTaxonomy_brandConnection",{"edges":81,"nodes":105,"__typename":107},[82],{"isPrimary":83,"node":84,"__typename":104},true,{"slug":85,"name":86,"industries":87,"__typename":103},"high-tech","High-Tech",{"nodes":88,"__typename":102},[89],{"translations":90,"__typename":101},[91,95,98],{"post_title":92,"locale":93,"__typename":94},"Hautes technologies","fr_FR","Translation",{"post_title":96,"locale":97,"__typename":94},"ハイテク","ja",{"post_title":99,"locale":100,"__typename":94},"하이테크","ko_KR","Industry","Taxonomy_industryToIndustryConnection","Taxonomy_industry","PostToTaxonomy_industryConnectionEdge",[106],{"name":86,"slug":85,"__typename":103},"PostToTaxonomy_industryConnection",{"nodes":109,"__typename":113},[110],{"name":111,"__typename":112},"SPARK","Taxonomy_keyword","PostToTaxonomy_keywordConnection",{"title":115,"metaDesc":116,"opengraphAuthor":117,"opengraphDescription":116,"opengraphTitle":26,"opengraphUrl":118,"opengraphSiteName":119,"opengraphPublishedTime":120,"opengraphModifiedTime":121,"twitterTitle":117,"twitterDescription":117,"readingTime":122,"metaRobotsNoindex":123,"__typename":124},"How Can Virtual Twins Improve AI Factories? - Dassault Systèmes blog","Dassault Systèmes and Vertiv collaborate to advance the model-based design, engineering and validation of next-generation AI infrastructure.","","https:\u002F\u002Fblog-frontoffice-contrib-prd.itvpc.3ds.com\u002Findustries\u002Fhigh-tech\u002Fvirtual-twins-improve-ai-factories\u002F","Dassault Systèmes blog","2026-10-22T05:00:00+00:00","2026-10-22T07:11:01+00:00",6,"index","PostTypeSEO","Post","RootQueryToPostConnection",{},{},1792653127197]