[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-post-\u002Ffr\u002Fnews\u002Fnvidia-gtc-2026-keynote-jensen-huang":3,"news-related-fr-\u002Ffr\u002Fnews\u002Fnvidia-gtc-2026-keynote-jensen-huang":592,"hreflang-nvidia-gtc-2026":612},{"doc":4,"redirect":8},{"id":5,"title":6,"author":7,"author_avatar":8,"body":9,"category":567,"date":568,"description":569,"extension":570,"image":571,"meta":572,"navigation":573,"path":574,"seo":575,"seoDescription":576,"seoTitle":577,"stem":578,"tags":579,"translationKey":590,"updatedAt":8,"__hash__":591},"news_fr\u002Ffr\u002Fnews\u002Fnvidia-gtc-2026-keynote-jensen-huang.md","Nvidia GTC 2026 : toutes les annonces de Jensen Huang — Vera Rubin, Feynman, robotaxis et DLSS 5","Paul Lefizelier",null,{"type":10,"value":11,"toc":547},"minimark",[12,53,58,73,77,90,95,110,128,163,170,181,185,194,205,216,227,241,245,256,260,269,276,280,297,304,308,327,345,356,366,384,387,391,400,411,414,418,432,436,447,461,464,487,499,502,506,540,542],[13,14,15,16,20,21,24,25,28,29,32,33,36,37,40,41,44,45,48,49],"p",{},"Ce 16 mars 2026, ",[17,18,19],"strong",{},"Jensen Huang"," monte sur scène pour la keynote principale de la ",[17,22,23],{},"Nvidia GTC 2026",". Le CEO fondateur de Nvidia dévoile ",[17,26,27],{},"Vera Rubin",", la nouvelle architecture GPU qui succède à Blackwell, aux côtés des accélérateurs ",[17,30,31],{},"H300"," et de la plateforme robotique ",[17,34,35],{},"Isaac",". Il crée la surprise en annonçant ",[17,38,39],{},"Feynman",", l'architecture prévue pour 2028, les ",[17,42,43],{},"robotaxis"," avec BYD, Hyundai et Nissan, et ",[17,46,47],{},"DLSS 5",", le rendu neuronal pour le gaming. Voici le résumé complet des annonces officielles de la keynote. ",[50,51,52],"em",{},"Article mis à jour le 17 mars 2026 avec toutes les annonces confirmées.",[54,55,57],"h2",{"id":56},"quest-ce-que-la-gtc-et-pourquoi-ça-compte","Qu'est-ce que la GTC et pourquoi ça compte",[13,59,60,61,64,65,68,69,72],{},"La ",[17,62,63],{},"GTC"," (GPU Technology Conference) est l'événement annuel phare de Nvidia depuis 2009. Chaque édition donne le tempo de l'industrie du calcul accéléré. En 2026, Nvidia pèse environ ",[17,66,67],{},"3 500 milliards de dollars"," de capitalisation boursière. La Big Tech prévoit de dépenser ",[17,70,71],{},"650 milliards de dollars"," en infrastructure IA cette année. Ce que Jensen Huang annonce ce soir à la GTC influence directement les feuilles de route d'OpenAI, Anthropic, Mistral et de tous les labs qui dépendent des puces Nvidia.",[54,74,76],{"id":75},"vera-rubin-le-successeur-de-blackwell","Vera Rubin : le successeur de Blackwell",[13,78,79,81,82,85,86,89],{},[17,80,27],{}," est la nouvelle architecture GPU de Nvidia, conçue pour entraîner et faire tourner des modèles à ",[17,83,84],{},"1 000 milliards de paramètres"," et au-delà. Elle succède à l'architecture ",[17,87,88],{},"Blackwell",", qui avait déjà repoussé les limites du calcul IA en 2024-2025.",[91,92,94],"h3",{"id":93},"les-chiffres-officiels-confirmés-ce-soir","Les chiffres officiels confirmés ce soir",[13,96,97,98,101,102,105,106,109],{},"Le rack ",[17,99,100],{},"NVL72"," (un châssis serveur contenant 72 GPU Vera Rubin et 36 CPU Vera) totalise ",[17,103,104],{},"220 000 milliards de transistors",". C'est un monstre de calcul. Jensen Huang a déclaré sur scène que la bande passante interne de ce seul rack ",[17,107,108],{},"dépasse la bande passante totale de l'internet mondial",".",[13,111,112,113,116,117,120,121,124,125,109],{},"Les performances parlent d'elles-mêmes. Vera Rubin NVL72 offre ",[17,114,115],{},"3,3 fois les performances d'inférence"," du Blackwell GB300 NVL72. Le ",[17,118,119],{},"coût par token est divisé par 10"," par rapport à Blackwell. La bande passante mémoire ",[17,122,123],{},"HBM4"," dépasse ",[17,126,127],{},"3 TB\u002Fs",[13,129,130,131,134,135,140,141,144,145,148,149,148,152,148,155,158,159,162],{},"Nvidia positionne Vera Rubin comme la puce de référence pour les ",[17,132,133],{},"modèles frontières"," — ceux qui alimentent ChatGPT, Claude, Gemini et les futurs ",[136,137,139],"a",{"href":138},"\u002Fnews\u002Fnyne-ai-agent-human-context-uc-berkeley","agents IA autonomes",". Les GPU sont déjà ",[17,142,143],{},"en production chez TSMC"," et seront disponibles sur ",[17,146,147],{},"AWS",", ",[17,150,151],{},"Google Cloud",[17,153,154],{},"Azure",[17,156,157],{},"Oracle"," et ",[17,160,161],{},"CoreWeave"," dès le second semestre 2026.",[13,164,165,166,169],{},"Jensen Huang a également dévoilé ",[17,167,168],{},"Vera Rubin Space One"," — une version de l'architecture spécifiquement conçue pour les satellites et l'espace. Les premiers clients sont des agences spatiales et des opérateurs de constellations de satellites. C'est la première fois que Nvidia cible directement l'orbite basse comme marché pour ses GPU IA.",[13,171,172,173,176,177,180],{},"Les ",[17,174,175],{},"accélérateurs H300",", également annoncés ce soir, ciblent un cas d'usage complémentaire : l'",[17,178,179],{},"inférence ultra-rapide",". Quand Vera Rubin entraîne, le H300 fait tourner. Ensemble, ils couvrent toute la chaîne de valeur de l'IA générative.",[54,182,184],{"id":183},"feynman-larchitecture-2028-que-personne-nattendait","Feynman : l'architecture 2028 que personne n'attendait",[13,186,187,188,190,191,109],{},"Jensen Huang n'a pas seulement présenté Vera Rubin. Il a dévoilé l'architecture qui lui succédera : ",[17,189,39],{},", prévue pour ",[17,192,193],{},"2028",[13,195,196,197,200,201,204],{},"Feynman sera gravée en ",[17,198,199],{},"process TSMC A16"," (1,6 nm) — le nœud de fabrication le plus avancé jamais annoncé commercialement. Nvidia serait le ",[17,202,203],{},"premier et unique client"," de TSMC à utiliser ce process en production massive.",[13,206,207,208,211,212,215],{},"L'innovation la plus marquante : la ",[17,209,210],{},"silicon photonics"," — une technologie qui utilise la lumière pour transmettre les données à l'intérieur des puces. Les données transitent par ",[17,213,214],{},"signaux optiques"," plutôt qu'électriques. La lumière remplace les fils de cuivre. Les gains en vitesse de transmission et en efficacité énergétique sont potentiellement révolutionnaires.",[13,217,218,219,222,223,226],{},"Autre annonce clé : ",[17,220,221],{},"NVLink 6ème génération"," avec ",[17,224,225],{},"optique co-packagée",". Les signaux optiques remplacent les signaux électriques à l'intérieur même des systèmes rack. Concrètement, les GPU communiquent entre eux à la vitesse de la lumière, sans les pertes liées au cuivre.",[13,228,229,230,148,233,236,237,240],{},"En annonçant Feynman deux ans à l'avance, Jensen Huang plante un drapeau. Le message est clair pour ",[17,231,232],{},"AMD",[17,234,235],{},"Intel"," et les ",[17,238,239],{},"Google TPUs"," : Nvidia a déjà conçu l'après Vera Rubin. La course aux puces IA se joue avec deux générations d'avance.",[54,242,244],{"id":243},"la-robotique-ia-au-coeur-de-la-vision-nvidia","La robotique IA au coeur de la vision Nvidia",[13,246,247,248,251,252,255],{},"Jensen Huang ne parle plus seulement de GPU pour les datacenters. Sa vision pour 2026, c'est ",[17,249,250],{},"l'usine physique IA"," : des robots humanoïdes pilotés par des ",[17,253,254],{},"world models"," entraînés sur puces Nvidia.",[91,257,259],{"id":258},"isaac-et-les-world-models","Isaac et les world models",[13,261,262,263,265,266,268],{},"La plateforme ",[17,264,35],{}," est le socle logiciel de cette ambition. Elle fournit aux développeurs les outils pour créer des robots autonomes capables de comprendre et d'interagir avec le monde physique. Les ",[17,267,254],{}," — des modèles d'IA qui simulent l'environnement réel — permettent aux robots d'apprendre sans passer par des millions d'heures d'essais physiques.",[13,270,271,272,275],{},"C'est un changement de paradigme. Nvidia ne vend plus seulement des puces. L'entreprise construit l'",[17,273,274],{},"infrastructure logicielle complète"," pour la robotique autonome.",[91,277,279],{"id":278},"robotaxis-byd-hyundai-et-nissan-rejoignent-la-plateforme-nvidia","Robotaxis : BYD, Hyundai et Nissan rejoignent la plateforme Nvidia",[13,281,282,283,148,286,148,289,292,293,296],{},"Nvidia annonce quatre nouveaux partenaires pour sa plateforme robotaxi : ",[17,284,285],{},"BYD",[17,287,288],{},"Hyundai",[17,290,291],{},"Nissan"," et un quatrième constructeur non encore nommé. Le déploiement se fait en partenariat avec ",[17,294,295],{},"Uber"," : des robotaxis Nvidia-powered seront commandables directement via l'app Uber.",[13,298,299,300,303],{},"Ces véhicules utilisent les ",[17,301,302],{},"world models Isaac"," pour naviguer en temps réel. Le signal est fort : les constructeurs automobiles traditionnels s'appuient sur l'infrastructure Nvidia plutôt que de développer leur propre stack IA. La conduite autonome passe du prototype au service grand public.",[54,305,307],{"id":306},"les-grands-gagnants-qui-a-déjà-signé-avec-nvidia","Les grands gagnants : qui a déjà signé avec Nvidia",[13,309,310,311,314,315,318,319,322,323,326],{},"Les annonces de la GTC 2026 s'accompagnent de partenariats majeurs. ",[17,312,313],{},"Thinking Machines Lab",", le laboratoire fondé par ",[17,316,317],{},"Mira Murati"," (ex-CTO d'OpenAI), a officialisé ce soir un deal « ",[17,320,321],{},"gigawatt-scale"," » — une infrastructure IA de la taille d'une centrale électrique, évaluée à ",[17,324,325],{},"50 milliards de dollars"," en puces Vera Rubin. C'est le plus gros contrat de fourniture de GPU jamais annoncé par une startup IA.",[13,328,329,332,333,336,337,340,341,344],{},[17,330,331],{},"Groq"," signe un deal à ",[17,334,335],{},"20 milliards de dollars"," avec Nvidia. Le chip ",[17,338,339],{},"LP30",", spécialisé inférence ultra-rapide, est prévu pour le ",[17,342,343],{},"Q3 2026",". Ce partenariat confirme la montée en puissance des puces d'inférence comme alternative aux GPU classiques. Groq, spécialisé dans les LPU (Language Processing Units), s'appuie sur Nvidia pour scaler son infrastructure.",[13,346,347,350,351,355],{},[17,348,349],{},"xAI",", le lab d'Elon Musk, figure également parmi les premiers clients de Vera Rubin. ",[136,352,354],{"href":353},"\u002Ffr\u002Fnews\u002Fdatabricks-ipo-134-milliards-valorisation-2026","Databricks"," renforce son partenariat avec Nvidia pour accélérer ses pipelines de données et d'entraînement.",[13,357,358,361,362,365],{},[17,359,360],{},"IBM"," officialise un partenariat élargi avec Nvidia à la GTC : GPU-native data analytics, extraction de documents non structurés et consulting IA enterprise. Cas concret présenté sur scène : l'accélération des décisions supply chain de ",[17,363,364],{},"Nestlé"," grâce aux puces Nvidia.",[13,367,368,369,222,372,148,374,148,377,158,380,383],{},"Nvidia co-fonde également l'",[17,370,371],{},"Optical Compute Interconnect Consortium",[17,373,147],{},[17,375,376],{},"Google",[17,378,379],{},"Microsoft",[17,381,382],{},"Meta",". L'objectif : définir un standard ouvert pour l'infrastructure optique IA. C'est la première fois que ces cinq géants s'alignent sur une norme commune d'interconnexion.",[13,385,386],{},"Ces accords confirment une tendance : Nvidia ne se contente plus de vendre du matériel. Le géant de Santa Clara prend des participations stratégiques dans les labs IA les plus prometteurs, verrouillant clients et influence.",[54,388,390],{"id":389},"dlss-5-le-rendu-neuronal-qui-réinvente-le-gaming","DLSS 5 : le rendu neuronal qui réinvente le gaming",[13,392,393,394,396,397,109],{},"Jensen Huang a présenté ",[17,395,47],{}," (Deep Learning Super Sampling), la cinquième génération de sa technologie de rendu assisté par IA. Cette fois, Nvidia passe au ",[17,398,399],{},"rendu neuronal guidé par la 3D",[13,401,402,403,406,407,410],{},"La différence avec DLSS 4 est fondamentale. Le moteur neuronal ne se contente plus d'upscaler des images : il ",[17,404,405],{},"génère des frames entières",". Le résultat : des performances ",[17,408,409],{},"4K photoréalistes en temps réel"," sur du hardware grand public, avec une qualité visuelle inédite.",[13,412,413],{},"L'impact dépasse le gaming. Les studios indépendants accèdent à des rendus cinématographiques sans recourir à des farms de rendu cloud. L'architecture, le cinéma et la visualisation industrielle en bénéficient aussi. Nvidia étend son empire IA bien au-delà des datacenters — jusqu'à l'écran de chaque joueur et créateur.",[54,415,417],{"id":416},"ce-que-ça-change-pour-les-devs-et-builders","Ce que ça change pour les devs et builders",[13,419,420,421,424,425,148,428,431],{},"Pour les développeurs, les annonces de la GTC 2026 ont des conséquences directes. Des puces plus puissantes signifient des ",[17,422,423],{},"coûts d'inférence en baisse",". Les API d'OpenAI, Anthropic et Mistral — celles qui alimentent ",[17,426,427],{},"Cursor",[17,429,430],{},"Replit",", GitHub Copilot et les IDE IA — tourneront sur du matériel plus performant.",[91,433,435],{"id":434},"openclaw-et-nemoclaw-lopen-source-des-agents-ia","OpenClaw et NemoClaw : l'open source des agents IA",[13,437,438,439,442,443,446],{},"Jensen Huang a annoncé ce soir ",[17,440,441],{},"OpenClaw",", la plateforme d'agents IA open source de Nvidia. Il l'a décrite comme ",[17,444,445],{},"le projet open source à la croissance la plus rapide de l'histoire",". OpenClaw fournit les briques pour construire, entraîner et déployer des agents autonomes.",[13,448,449,452,453,456,457,460],{},[17,450,451],{},"NemoClaw"," est sa version enterprise, développée en partenariat avec ",[17,454,455],{},"OpenAI",". Elle permet de déployer des agents autonomes sur infrastructure privée. Les participants à la GTC ont construit des agents en live sur les ",[17,458,459],{},"DGX Spark",", les stations de développement IA de Nvidia.",[13,462,463],{},"Concrètement, les modèles frontières deviennent plus accessibles. Le vibecoding — cette pratique qui consiste à coder en langage naturel via un agent IA — bénéficie directement de la puissance supplémentaire. Plus le hardware progresse, plus les modèles peuvent raisonner longtemps et plus les outils de développement assisté par IA gagnent en fiabilité.",[13,465,466,467,148,469,148,472,148,475,148,478,148,481,158,484,486],{},"Jensen Huang a également réuni une table ronde inédite avec ",[17,468,427],{},[17,470,471],{},"A16Z",[17,473,474],{},"AI2",[17,476,477],{},"AMP Coalition",[17,479,480],{},"Black Forest Labs",[17,482,483],{},"Reflection AI",[17,485,313],{}," autour de l'état de l'art des modèles frontier open source. Que Cursor — l'IDE IA préféré des vibecorders — soit à la même table que Mira Murati et Andreessen Horowitz est un signal fort : les outils de développement IA sont désormais des acteurs stratégiques au même titre que les labs de recherche.",[13,488,489,490,493,494,498],{},"La baisse des coûts d'inférence ouvre aussi la porte à de ",[17,491,492],{},"nouveaux cas d'usage"," : agents autonomes en production, IA embarquée dans les applications mobiles, ",[136,495,497],{"href":496},"\u002Ffr\u002Fnews\u002Fdeepseek-v4-modele-ia-open-weight-1000-milliards","modèles spécialisés fine-tunés comme DeepSeek V4"," accessibles aux PME.",[500,501],"hr",{},[54,503,505],{"id":504},"en-résumé","En résumé",[507,508,509,516,521,526,535],"ul",{},[510,511,97,512,515],"li",{},[17,513,514],{},"Vera Rubin NVL72"," (72 GPU + 36 CPU) offre 3,3× les performances d'inférence de Blackwell, avec un coût par token divisé par 10.",[510,517,518,520],{},[17,519,39],{}," (2028) utilise le process TSMC A16 1,6 nm et la silicon photonics — l'architecture GPU la plus avancée jamais annoncée.",[510,522,523,525],{},[17,524,441],{},", l'agent IA open source de Nvidia, est déjà le projet open source à la croissance la plus rapide de l'histoire.",[510,527,528,148,530,158,532,534],{},[17,529,285],{},[17,531,288],{},[17,533,291],{}," rejoignent la plateforme robotaxi Nvidia, avec déploiement via l'app Uber.",[510,536,537,539],{},[17,538,47],{}," apporte le rendu neuronal 4K photoréaliste en temps réel sur hardware grand public.",[500,541],{},[13,543,544],{},[50,545,546],{},"Avec des puces toujours plus puissantes et des modèles toujours plus capables, la question n'est plus de savoir si les agents IA autonomes remplaceront certains workflows de développement — mais quand. La GTC 2026 vient de rapprocher cette échéance.",{"title":548,"searchDepth":549,"depth":549,"links":550},"",2,[551,552,556,557,561,562,563,566],{"id":56,"depth":549,"text":57},{"id":75,"depth":549,"text":76,"children":553},[554],{"id":93,"depth":555,"text":94},3,{"id":183,"depth":549,"text":184},{"id":243,"depth":549,"text":244,"children":558},[559,560],{"id":258,"depth":555,"text":259},{"id":278,"depth":555,"text":279},{"id":306,"depth":549,"text":307},{"id":389,"depth":549,"text":390},{"id":416,"depth":549,"text":417,"children":564},[565],{"id":434,"depth":555,"text":435},{"id":504,"depth":549,"text":505},"ai","2026-03-17T08:14:00","Keynote Nvidia GTC 2026 : Jensen Huang présente Vera Rubin, Feynman, les robotaxis BYD\u002FHyundai\u002FNissan, DLSS 5, OpenClaw et la plateforme Isaac.","md","\u002Fnvidia-gtc-2026-keynote-jensen-huang-vera-rubin.jpg",{},true,"\u002Ffr\u002Fnews\u002Fnvidia-gtc-2026-keynote-jensen-huang",{"title":6,"description":569},"Nvidia GTC 2026 : Vera Rubin, Feynman, robotaxis, DLSS 5, OpenClaw — toutes les annonces de Jensen Huang. Recap complet de la keynote.","Nvidia GTC 2026 : Vera Rubin, Feynman, DLSS 5 — le recap complet","fr\u002Fnews\u002Fnvidia-gtc-2026-keynote-jensen-huang",[580,581,582,583,584,585,586,43,587,588,589],"nvidia","gtc-2026","jensen-huang","vera-rubin","feynman","gpu","robotique","dlss-5","world-models","keynote","nvidia-gtc-2026","UB-YrxbnNSyGyVBcZDzkAxv9czUvhPD1bCWrx9e2S7c",[593,600,606],{"title":594,"description":595,"path":596,"date":597,"image":598,"category":599},"Les cloud agents Cursor construisent, hébergent et publient des apps live","Les cloud agents Cursor peuvent scaffolder une app web, provisionner Postgres et le SSO Cursor, lancer des contrôles de sécurité, puis publier une URL partageable — sans compte d'hébergement séparé.","\u002Ffr\u002Fnews\u002Fcursor-live-apps-agents-cloud-deploiement-postgres-auth","2026-11-07T08:00:00","\u002Fimages\u002Fnews\u002Fcursor-live-apps-cloud-agents-deploy-postgres-auth.jpg","vibecoding",{"title":601,"description":602,"path":603,"date":604,"image":605,"category":567},"Google rend Gemini Omni 1.2 Flash généralement disponible, avec des vidéos de 15 secondes","Google DeepMind baisse le prix d'Omni Flash de 30 %, à 0,07 $ par seconde en 720p, passe de 10 à 15 secondes de clip, et ouvre le modèle sur l'API Gemini et Google Cloud.","\u002Ffr\u002Fnews\u002Fgoogle-gemini-omni-1-2-flash-video-15-secondes","2026-11-06T08:00:00","\u002Fimages\u002Fnews\u002Fgoogle-gemini-omni-1-2-flash-15-second-video.jpg",{"title":607,"description":608,"path":609,"date":610,"image":611,"category":567},"Anthropic aurait signé son premier accord de licence avec un éditeur, l'AP","Selon Adweek, Anthropic a conclu un accord de licence de contenus et de données avec l'Associated Press juste avant les élections de mi-mandat américaines, son premier avec un éditeur numérique. Aucune des deux entreprises ne l'a confirmé officiellement, et les conditions ne sont pas publiques.","\u002Ffr\u002Fnews\u002Fanthropic-premier-accord-licence-editeurs-associated-press","2026-11-05T12:35:00","\u002Fimages\u002Fnews\u002Fanthropic-ap-publisher-licensing-deal.jpg",{"en":613,"fr":1111,"es":8},{"id":614,"title":615,"author":7,"author_avatar":8,"body":616,"category":567,"date":568,"description":1101,"extension":570,"image":571,"meta":1102,"navigation":573,"path":1103,"seo":1104,"seoDescription":1105,"seoTitle":1106,"stem":1107,"tags":1108,"translationKey":590,"updatedAt":8,"__hash__":1110},"news_en\u002Fnews\u002Fnvidia-gtc-2026-keynote-jensen-huang.md","Nvidia GTC 2026: All of Jensen Huang's Announcements — Vera Rubin, Feynman, Robotaxis and DLSS 5",{"type":10,"value":617,"toc":1084},[618,648,652,665,669,681,685,698,714,741,747,757,761,769,780,790,801,811,815,825,829,838,845,849,862,869,873,889,904,913,922,936,939,943,952,963,966,970,982,986,996,1007,1010,1028,1040,1042,1046,1077,1079],[13,619,620,621,623,624,626,627,629,630,632,633,635,636,638,639,641,642,644,645],{},"On March 16, 2026, ",[17,622,19],{}," takes the stage for the main ",[17,625,23],{}," keynote. Nvidia's co-founder and CEO unveils ",[17,628,27],{},", the new GPU architecture succeeding Blackwell, alongside ",[17,631,31],{}," accelerators and the ",[17,634,35],{}," robotics platform. He also surprises the audience by revealing ",[17,637,39],{},", the architecture planned for 2028, ",[17,640,43],{}," with BYD, Hyundai and Nissan, and ",[17,643,47],{},", neural rendering for gaming. Here's the full recap of the official announcements from the keynote. ",[50,646,647],{},"Updated March 17, 2026 with all confirmed announcements.",[54,649,651],{"id":650},"what-is-gtc-and-why-it-matters","What Is GTC and Why It Matters",[13,653,654,656,657,660,661,664],{},[17,655,63],{}," (GPU Technology Conference) is Nvidia's flagship annual event, running since 2009. Each edition sets the pace for the accelerated computing industry. In 2026, Nvidia's market cap sits at roughly ",[17,658,659],{},"$3.5 trillion",". Big Tech is set to spend ",[17,662,663],{},"$650 billion"," on AI infrastructure this year. What Jensen Huang announces tonight at GTC directly shapes the roadmaps of OpenAI, Anthropic, Mistral, and every lab that depends on Nvidia chips.",[54,666,668],{"id":667},"vera-rubin-the-blackwell-successor","Vera Rubin: The Blackwell Successor",[13,670,671,673,674,677,678,680],{},[17,672,27],{}," is Nvidia's new GPU architecture, built to train and run models with ",[17,675,676],{},"1 trillion parameters"," and beyond. It succeeds ",[17,679,88],{},", which already pushed the boundaries of AI compute in 2024-2025.",[91,682,684],{"id":683},"the-official-numbers-confirmed-tonight","The Official Numbers Confirmed Tonight",[13,686,687,688,690,691,694,695,109],{},"The ",[17,689,100],{}," rack (a server chassis housing 72 Vera Rubin GPUs and 36 Vera CPUs) packs ",[17,692,693],{},"220 trillion transistors",". It's a compute behemoth. Jensen Huang stated on stage that this single rack's internal bandwidth ",[17,696,697],{},"exceeds the total bandwidth of the entire global internet",[13,699,700,701,704,705,708,709,711,712,109],{},"The performance speaks for itself. Vera Rubin NVL72 delivers ",[17,702,703],{},"3.3x the inference performance"," of the Blackwell GB300 NVL72. The ",[17,706,707],{},"cost per token is 10x lower"," than Blackwell. ",[17,710,123],{}," memory bandwidth exceeds ",[17,713,127],{},[13,715,716,717,720,721,724,725,728,729,148,731,148,733,148,735,737,738,740],{},"Nvidia positions Vera Rubin as the reference chip for ",[17,718,719],{},"frontier models"," — those powering ChatGPT, Claude, Gemini, and future ",[136,722,723],{"href":138},"autonomous AI agents",". The GPUs are already ",[17,726,727],{},"in production at TSMC"," and will be available on ",[17,730,147],{},[17,732,151],{},[17,734,154],{},[17,736,157],{}," and ",[17,739,161],{}," in the second half of 2026.",[13,742,743,744,746],{},"Jensen Huang also revealed ",[17,745,168],{}," — a version of the architecture specifically designed for satellites and space. First customers include space agencies and satellite constellation operators. It's the first time Nvidia directly targets low Earth orbit as a market for its AI GPUs.",[13,748,687,749,752,753,756],{},[17,750,751],{},"H300 accelerators",", also announced tonight, target a complementary use case: ",[17,754,755],{},"ultra-fast inference",". Where Vera Rubin trains, the H300 serves. Together, they cover the entire generative AI value chain.",[54,758,760],{"id":759},"feynman-the-2028-architecture-nobody-saw-coming","Feynman: The 2028 Architecture Nobody Saw Coming",[13,762,763,764,766,767,109],{},"Jensen Huang didn't just present Vera Rubin. He revealed the architecture that will succeed it: ",[17,765,39],{},", planned for ",[17,768,193],{},[13,770,771,772,775,776,779],{},"Feynman will be built on the ",[17,773,774],{},"TSMC A16 process"," (1.6 nm) — the most advanced manufacturing node ever announced commercially. Nvidia would be the ",[17,777,778],{},"first and only customer"," to use TSMC A16 in mass production.",[13,781,782,783,785,786,789],{},"The most striking innovation: ",[17,784,210],{}," — a technology that uses light to transmit data inside chips. Data travels via ",[17,787,788],{},"optical signals"," rather than electrical ones. Light replaces copper wires. The gains in transmission speed and energy efficiency are potentially revolutionary.",[13,791,792,793,796,797,800],{},"Another key announcement: ",[17,794,795],{},"6th-generation NVLink"," with ",[17,798,799],{},"co-packaged optics",". Optical signals replace electrical signals inside rack systems themselves. In practice, GPUs communicate at the speed of light, without the losses inherent to copper.",[13,802,803,804,148,806,737,808,810],{},"By announcing Feynman two years ahead, Jensen Huang is planting a flag. The message to ",[17,805,232],{},[17,807,235],{},[17,809,239],{}," is clear: Nvidia has already designed the generation after Vera Rubin. The AI chip race is being fought two generations ahead.",[54,812,814],{"id":813},"ai-robotics-at-the-core-of-nvidias-vision","AI Robotics at the Core of Nvidia's Vision",[13,816,817,818,821,822,824],{},"Jensen Huang is no longer talking just about GPUs for data centers. His 2026 vision is the ",[17,819,820],{},"physical AI factory",": humanoid robots driven by ",[17,823,254],{}," trained on Nvidia chips.",[91,826,828],{"id":827},"isaac-and-world-models","Isaac and World Models",[13,830,687,831,833,834,837],{},[17,832,35],{}," platform provides the software foundation for this ambition. It gives developers the tools to build autonomous robots that understand and interact with the physical world. ",[17,835,836],{},"World models"," — AI models that simulate real-world environments — let robots learn without millions of hours of physical trial and error.",[13,839,840,841,844],{},"This is a paradigm shift. Nvidia is no longer just selling chips. The company is building the ",[17,842,843],{},"complete software infrastructure"," for autonomous robotics.",[91,846,848],{"id":847},"robotaxis-byd-hyundai-and-nissan-join-the-nvidia-platform","Robotaxis: BYD, Hyundai and Nissan Join the Nvidia Platform",[13,850,851,852,148,854,148,856,858,859,861],{},"Nvidia announces four new partners for its robotaxi platform: ",[17,853,285],{},[17,855,288],{},[17,857,291],{}," and a fourth unnamed manufacturer. Deployment is coupled with ",[17,860,295],{},": Nvidia-powered robotaxis will be bookable directly through the Uber app.",[13,863,864,865,868],{},"These vehicles use ",[17,866,867],{},"Isaac world models"," to navigate in real time. The signal is clear: traditional automakers are relying on Nvidia infrastructure rather than building their own AI stack. Autonomous driving is moving from prototype to mainstream service.",[54,870,872],{"id":871},"the-big-winners-who-already-signed-with-nvidia","The Big Winners: Who Already Signed With Nvidia",[13,874,875,876,878,879,881,882,884,885,888],{},"The GTC 2026 announcements come with major partnerships. ",[17,877,313],{},", the laboratory founded by ",[17,880,317],{}," (former OpenAI CTO), officially confirmed tonight a \"",[17,883,321],{},"\" deal — an AI infrastructure the size of a power plant, valued at ",[17,886,887],{},"$50 billion"," in Vera Rubin chips. It's the largest GPU supply contract ever announced by an AI startup.",[13,890,891,893,894,897,898,900,901,903],{},[17,892,331],{}," signs a ",[17,895,896],{},"$20 billion"," deal with Nvidia. The ",[17,899,339],{}," chip, specialized for ultra-fast inference, is planned for ",[17,902,343],{},". This partnership confirms the rise of inference chips as an alternative to traditional GPUs. Groq, which specializes in LPUs (Language Processing Units), is leveraging Nvidia to scale its infrastructure.",[13,905,906,908,909,912],{},[17,907,349],{},", Elon Musk's lab, is also among the first Vera Rubin customers. ",[136,910,354],{"href":911},"\u002Fnews\u002Fdatabricks-ipo-134-billion-valuation-2026"," is deepening its Nvidia partnership to accelerate its data and training pipelines.",[13,914,915,917,918,921],{},[17,916,360],{}," officially expands its Nvidia partnership at GTC: GPU-native data analytics, unstructured document extraction and enterprise AI consulting. Use case presented on stage: accelerating ",[17,919,920],{},"Nestlé's"," supply chain decisions with Nvidia chips.",[13,923,924,925,796,927,148,929,148,931,737,933,935],{},"Nvidia also co-founds the ",[17,926,371],{},[17,928,147],{},[17,930,376],{},[17,932,379],{},[17,934,382],{},". The goal: define an open standard for AI optical infrastructure. It's the first time these five giants have aligned on a common interconnect standard.",[13,937,938],{},"These deals confirm a trend: Nvidia is no longer content to just sell hardware. The Santa Clara giant is taking strategic stakes in the most promising AI labs, locking in customers and influence.",[54,940,942],{"id":941},"dlss-5-neural-rendering-reinvents-gaming","DLSS 5: Neural Rendering Reinvents Gaming",[13,944,945,946,948,949,109],{},"Jensen Huang introduced ",[17,947,47],{}," (Deep Learning Super Sampling), the fifth generation of Nvidia's AI-assisted rendering technology. This time, Nvidia moves to ",[17,950,951],{},"3D-guided neural rendering",[13,953,954,955,958,959,962],{},"The difference from DLSS 4 is fundamental. The neural engine no longer just upscales images: it ",[17,956,957],{},"generates entire frames",". The result: ",[17,960,961],{},"photorealistic 4K performance in real time"," on consumer hardware, with unprecedented visual quality.",[13,964,965],{},"The impact extends beyond gaming. Independent studios gain access to cinematic rendering without cloud render farms. Architecture, film and industrial visualization also benefit. Nvidia is extending its AI empire far beyond data centers — all the way to every gamer's and creator's screen.",[54,967,969],{"id":968},"what-this-means-for-developers-and-builders","What This Means for Developers and Builders",[13,971,972,973,976,977,148,979,981],{},"For developers, the GTC 2026 announcements have direct consequences. More powerful chips mean ",[17,974,975],{},"lower inference costs",". The APIs from OpenAI, Anthropic, and Mistral — those powering ",[17,978,427],{},[17,980,430],{},", GitHub Copilot, and AI-powered IDEs — will run on more performant hardware.",[91,983,985],{"id":984},"openclaw-and-nemoclaw-open-source-ai-agents","OpenClaw and NemoClaw: Open-Source AI Agents",[13,987,988,989,991,992,995],{},"Jensen Huang announced tonight ",[17,990,441],{},", Nvidia's open-source AI agent platform. He described it as ",[17,993,994],{},"the fastest-growing open-source project in history",". OpenClaw provides the building blocks to create, train and deploy autonomous agents.",[13,997,998,1000,1001,1003,1004,1006],{},[17,999,451],{}," is its enterprise version, developed in partnership with ",[17,1002,455],{},". It enables deploying autonomous agents on private infrastructure. GTC attendees built agents live on ",[17,1005,459],{},", Nvidia's AI development workstations.",[13,1008,1009],{},"In practical terms, frontier models become more accessible. Vibe coding — the practice of coding in natural language through an AI agent — directly benefits from the added power. The better the hardware, the longer models can reason, and the more reliable AI-assisted development tools become.",[13,1011,1012,1013,148,1015,148,1017,148,1019,148,1021,148,1023,737,1025,1027],{},"Jensen Huang also hosted an unprecedented roundtable featuring ",[17,1014,427],{},[17,1016,471],{},[17,1018,474],{},[17,1020,477],{},[17,1022,480],{},[17,1024,483],{},[17,1026,313],{}," on the state of the art in frontier open-source models. That Cursor — the vibe coders' favorite AI IDE — sat at the same table as Mira Murati and Andreessen Horowitz sends a clear signal: AI development tools are now strategic players on par with research labs.",[13,1029,1030,1031,1034,1035,1039],{},"Falling inference costs also open the door to ",[17,1032,1033],{},"new use cases",": autonomous agents in production, AI embedded in mobile apps, and ",[136,1036,1038],{"href":1037},"\u002Fnews\u002Fdeepseek-v4-open-weight-ai-model-1-trillion-parameters","fine-tuned specialized models like DeepSeek V4"," accessible to small and mid-size businesses.",[500,1041],{},[54,1043,1045],{"id":1044},"key-takeaways","Key Takeaways",[507,1047,1048,1053,1058,1063,1072],{},[510,1049,687,1050,1052],{},[17,1051,514],{}," rack (72 GPUs + 36 CPUs) delivers 3.3x the inference performance of Blackwell, with a 10x lower cost per token.",[510,1054,1055,1057],{},[17,1056,39],{}," (2028) uses the TSMC A16 1.6 nm process and silicon photonics — the most advanced GPU architecture ever announced.",[510,1059,1060,1062],{},[17,1061,441],{},", Nvidia's open-source AI agent platform, is already the fastest-growing open-source project in history.",[510,1064,1065,148,1067,737,1069,1071],{},[17,1066,285],{},[17,1068,288],{},[17,1070,291],{}," join the Nvidia robotaxi platform, with deployment through the Uber app.",[510,1073,1074,1076],{},[17,1075,47],{}," brings photorealistic 4K neural rendering in real time on consumer hardware.",[500,1078],{},[13,1080,1081],{},[50,1082,1083],{},"With ever more powerful chips and ever more capable models, the question is no longer whether autonomous AI agents will replace certain development workflows — but when. GTC 2026 just brought that timeline closer.",{"title":548,"searchDepth":549,"depth":549,"links":1085},[1086,1087,1090,1091,1095,1096,1097,1100],{"id":650,"depth":549,"text":651},{"id":667,"depth":549,"text":668,"children":1088},[1089],{"id":683,"depth":555,"text":684},{"id":759,"depth":549,"text":760},{"id":813,"depth":549,"text":814,"children":1092},[1093,1094],{"id":827,"depth":555,"text":828},{"id":847,"depth":555,"text":848},{"id":871,"depth":549,"text":872},{"id":941,"depth":549,"text":942},{"id":968,"depth":549,"text":969,"children":1098},[1099],{"id":984,"depth":555,"text":985},{"id":1044,"depth":549,"text":1045},"Nvidia GTC 2026 keynote: Jensen Huang reveals Vera Rubin, Feynman, BYD\u002FHyundai\u002FNissan robotaxis, DLSS 5, OpenClaw and the Isaac robotics platform.",{},"\u002Fnews\u002Fnvidia-gtc-2026-keynote-jensen-huang",{"title":615,"description":1101},"Nvidia GTC 2026: Vera Rubin, Feynman, robotaxis, DLSS 5, OpenClaw — all of Jensen Huang's announcements. Complete keynote recap.","Nvidia GTC 2026: Vera Rubin, Feynman, DLSS 5 — Full Recap","news\u002Fnvidia-gtc-2026-keynote-jensen-huang",[580,581,582,583,584,585,1109,43,587,588,589],"robotics","Bnup2eXuQG1az5pfF9YepEHDmBaIlb1XuisAIhZnPd8",{"id":5,"title":6,"author":7,"author_avatar":8,"body":1112,"category":567,"date":568,"description":569,"extension":570,"image":571,"meta":1439,"navigation":573,"path":574,"seo":1440,"seoDescription":576,"seoTitle":577,"stem":578,"tags":1441,"translationKey":590,"updatedAt":8,"__hash__":591},{"type":10,"value":1113,"toc":1422},[1114,1134,1136,1144,1146,1154,1156,1164,1174,1192,1196,1202,1204,1210,1216,1222,1228,1236,1238,1244,1246,1252,1256,1258,1268,1272,1274,1284,1294,1300,1306,1318,1320,1322,1328,1334,1336,1338,1346,1348,1354,1362,1364,1380,1386,1388,1390,1416,1418],[13,1115,15,1116,20,1118,24,1120,28,1122,32,1124,36,1126,40,1128,44,1130,48,1132],{},[17,1117,19],{},[17,1119,23],{},[17,1121,27],{},[17,1123,31],{},[17,1125,35],{},[17,1127,39],{},[17,1129,43],{},[17,1131,47],{},[50,1133,52],{},[54,1135,57],{"id":56},[13,1137,60,1138,64,1140,68,1142,72],{},[17,1139,63],{},[17,1141,67],{},[17,1143,71],{},[54,1145,76],{"id":75},[13,1147,1148,81,1150,85,1152,89],{},[17,1149,27],{},[17,1151,84],{},[17,1153,88],{},[91,1155,94],{"id":93},[13,1157,97,1158,101,1160,105,1162,109],{},[17,1159,100],{},[17,1161,104],{},[17,1163,108],{},[13,1165,112,1166,116,1168,120,1170,124,1172,109],{},[17,1167,115],{},[17,1169,119],{},[17,1171,123],{},[17,1173,127],{},[13,1175,130,1176,134,1178,140,1180,144,1182,148,1184,148,1186,148,1188,158,1190,162],{},[17,1177,133],{},[136,1179,139],{"href":138},[17,1181,143],{},[17,1183,147],{},[17,1185,151],{},[17,1187,154],{},[17,1189,157],{},[17,1191,161],{},[13,1193,165,1194,169],{},[17,1195,168],{},[13,1197,172,1198,176,1200,180],{},[17,1199,175],{},[17,1201,179],{},[54,1203,184],{"id":183},[13,1205,187,1206,190,1208,109],{},[17,1207,39],{},[17,1209,193],{},[13,1211,196,1212,200,1214,204],{},[17,1213,199],{},[17,1215,203],{},[13,1217,207,1218,211,1220,215],{},[17,1219,210],{},[17,1221,214],{},[13,1223,218,1224,222,1226,226],{},[17,1225,221],{},[17,1227,225],{},[13,1229,229,1230,148,1232,236,1234,240],{},[17,1231,232],{},[17,1233,235],{},[17,1235,239],{},[54,1237,244],{"id":243},[13,1239,247,1240,251,1242,255],{},[17,1241,250],{},[17,1243,254],{},[91,1245,259],{"id":258},[13,1247,262,1248,265,1250,268],{},[17,1249,35],{},[17,1251,254],{},[13,1253,271,1254,275],{},[17,1255,274],{},[91,1257,279],{"id":278},[13,1259,282,1260,148,1262,148,1264,292,1266,296],{},[17,1261,285],{},[17,1263,288],{},[17,1265,291],{},[17,1267,295],{},[13,1269,299,1270,303],{},[17,1271,302],{},[54,1273,307],{"id":306},[13,1275,310,1276,314,1278,318,1280,322,1282,326],{},[17,1277,313],{},[17,1279,317],{},[17,1281,321],{},[17,1283,325],{},[13,1285,1286,332,1288,336,1290,340,1292,344],{},[17,1287,331],{},[17,1289,335],{},[17,1291,339],{},[17,1293,343],{},[13,1295,1296,350,1298,355],{},[17,1297,349],{},[136,1299,354],{"href":353},[13,1301,1302,361,1304,365],{},[17,1303,360],{},[17,1305,364],{},[13,1307,368,1308,222,1310,148,1312,148,1314,158,1316,383],{},[17,1309,371],{},[17,1311,147],{},[17,1313,376],{},[17,1315,379],{},[17,1317,382],{},[13,1319,386],{},[54,1321,390],{"id":389},[13,1323,393,1324,396,1326,109],{},[17,1325,47],{},[17,1327,399],{},[13,1329,402,1330,406,1332,410],{},[17,1331,405],{},[17,1333,409],{},[13,1335,413],{},[54,1337,417],{"id":416},[13,1339,420,1340,424,1342,148,1344,431],{},[17,1341,423],{},[17,1343,427],{},[17,1345,430],{},[91,1347,435],{"id":434},[13,1349,438,1350,442,1352,446],{},[17,1351,441],{},[17,1353,445],{},[13,1355,1356,452,1358,456,1360,460],{},[17,1357,451],{},[17,1359,455],{},[17,1361,459],{},[13,1363,463],{},[13,1365,466,1366,148,1368,148,1370,148,1372,148,1374,148,1376,158,1378,486],{},[17,1367,427],{},[17,1369,471],{},[17,1371,474],{},[17,1373,477],{},[17,1375,480],{},[17,1377,483],{},[17,1379,313],{},[13,1381,489,1382,493,1384,498],{},[17,1383,492],{},[136,1385,497],{"href":496},[500,1387],{},[54,1389,505],{"id":504},[507,1391,1392,1396,1400,1404,1412],{},[510,1393,97,1394,515],{},[17,1395,514],{},[510,1397,1398,520],{},[17,1399,39],{},[510,1401,1402,525],{},[17,1403,441],{},[510,1405,1406,148,1408,158,1410,534],{},[17,1407,285],{},[17,1409,288],{},[17,1411,291],{},[510,1413,1414,539],{},[17,1415,47],{},[500,1417],{},[13,1419,1420],{},[50,1421,546],{},{"title":548,"searchDepth":549,"depth":549,"links":1423},[1424,1425,1428,1429,1433,1434,1435,1438],{"id":56,"depth":549,"text":57},{"id":75,"depth":549,"text":76,"children":1426},[1427],{"id":93,"depth":555,"text":94},{"id":183,"depth":549,"text":184},{"id":243,"depth":549,"text":244,"children":1430},[1431,1432],{"id":258,"depth":555,"text":259},{"id":278,"depth":555,"text":279},{"id":306,"depth":549,"text":307},{"id":389,"depth":549,"text":390},{"id":416,"depth":549,"text":417,"children":1436},[1437],{"id":434,"depth":555,"text":435},{"id":504,"depth":549,"text":505},{},{"title":6,"description":569},[580,581,582,583,584,585,586,43,587,588,589]]