<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
        <title>Le vrai coût de l'IA : pourquoi le prix au token est faux</title>
        <link>https://video.ut0pia.org/videos/watch/6fb6a0c8-da86-477a-ad9a-e56db7bf1e05</link>
        <description>En ce moment, une seule métrique domine les décisions autour de l'intelligence artificielle générative : le prix au token. Simple, universelle, rassurante... et pourtant profondément insuffisante. Entre le chiffre affiché sur une page tarifaire et la facture qui arrive en fin de mois, il existe un écart que peu d'organisations savent expliquer, et encore moins anticiper. Cette vidéo s'adresse à deux publics complémentaires: Aux décideurs (directions techniques, financières, ingénierie) qui voient les budgets IA gonfler sans toujours en comprendre les mécanismes., Aux  développeuses et développeurs, dont chaque décision technique a un impact direct, parfois spectaculaire, sur le coût final. Le fil conducteur : montrer que le problème du budget et celui de l'architecture ne sont pas deux sujets distincts, mais un seul et même sujet., En m'appuyant sur des données réelles et récentes, j'explore pourquoi le même travail peut coûter du simple au multiple selon la façon dont il est exécuté, pourquoi le moins cher au token n'est pas toujours le moins cher à la tâche, et pourquoi certains coûts n'apparaissent sur aucune page tarifaire. Vous découvrirez aussi l'impact des modèles OpenWeight, de la Chine, et les prochaines révolutions qui vont nous impacter. Vous en repartirez, j'espère, convaincu qu'il n'est pas possible de travailler avec un seul fournisseur, un ou deux modèles, et que cela va changer en plus chaque semaine. Bon visionnage! #📚  Sources Marché, volumes &amp; adoptionOpenRouter — Rankings : https://openrouter.ai/rankings, OpenRouter — « The Open Weight Models That Matter » (juin 2026) : https://openrouter.ai/blog/insights/the-open-weight-models-that-matter-june-2026/, OpenRouter — « DeepSeek V4 Is Earning Agentic Token Share » : https://openrouter.ai/blog/insights/deepseek-v4-adoption/, Vercel — AI Gateway Leaderboard : https://vercel.com/ai-gateway/leaderboards/labs, Artificial Analysis — AI Trends (Intelligence Index, Country Analysis, Open Weights vs Proprietary) :  https://artificialanalysis.ai/trends, Artificial Analysis : https://artificialanalysis.ai/, LM Arena : https://lmarena.ai/, Coûts réels &amp; benchmarksDatabricks — « Benchmarking Coding Agents on a Multi-Million-Line Codebase » :  https://databricks.com/blog/benchmarking-coding-agents-databricks-multi-million-line-codebase, Cursor — Insights Report : https://cursor.com/insights, The Pragmatic Engineer — « Interesting AI coding stats from Cursor » : https://blog.pragmaticengineer.com/the-pulse-interesting-ai-coding-stats-from-cursor/, Ingénierie &amp; architecture d'exécution (harness)Anthropic Engineering — « Building agents for long-running applications » (harness design) : https://www.anthropic.com/engineering/harness-design-long-running-apps, Analyses de fond (SemiAnalysis — Dylan Patel)« TokenBudgeting » : https://newsletter.semianalysis.com/p/tokenbudgeting-our-conversations, « AI Dark Output: The Visible Cost of… » :  https://newsletter.semianalysis.com/p/ai-dark-output-the-visible-cost-of, « Claude Code is the Inflection Point » : https://newsletter.semianalysis.com/p/claude-code-is-the-inflection-point, Cas entreprises &amp; presseTechCrunch — « Uber caps employee AI spending after blowing through budget in four months » : https://techcrunch.com/2026/06/02/uber-caps-employee-ai-spending-after-blowing-through-budget-in-four-months/, The Verge — « Microsoft cancels Claude Code licenses » : https://www.theverge.com/tech/930447/microsoft-claude-code-discontinued-notepad, TechCrunch — « GitHub Copilot's new token-based billing spurs consternation among devs » : https://techcrunch.com/2026/05/30/what-a-joke-github-copilots-new-token-based-billing-spurs-consternation-among-devs/, TechCrunch — « The token bill comes due » : https://techcrunch.com/2026/06/05/the-token-bill-comes-due-inside-the-industry-scramble-to-manage-ais-runaway-costs/, TechCrunch — « Tokenmaxxing is making developers less productive… » :  https://techcrunch.com/2026/04/17/tokenmaxxing-is-making-developers-less-productive-than-they-think/, The Register — « Even banks and hyperscalers are sounding the alarm about the AI bubble » :  https://www.theregister.com/ai-and-ml/2026/07/06/even-banks-and-hyperscalers-are-now-sounding-the-alarm-about-the-ai-bubble/, Contexte financier (Anthropic)- TechCrunch — « Anthropic raises $65B, nears $1T valuation ahead of IPO » :   https://techcrunch.com/2026/05/28/anthropic-raises-65-billion-nears-1t-valuation-ahead-of-ipo/TechCrunch — « Anthropic files to go public » : https://techcrunch.com/2026/06/01/anthropic-files-to-go-public/</description>
        <lastBuildDate>Sun, 12 Jul 2026 17:07:33 GMT</lastBuildDate>
        <docs>https://validator.w3.org/feed/docs/rss2.html</docs>
        <generator>PeerTube - https://video.ut0pia.org</generator>
        <image>
            <title>Le vrai coût de l'IA : pourquoi le prix au token est faux</title>
            <url>https://video.ut0pia.org/lazy-static/avatars/0287a09a-aae7-4840-9843-b416426e7046.webp</url>
            <link>https://video.ut0pia.org/videos/watch/6fb6a0c8-da86-477a-ad9a-e56db7bf1e05</link>
        </image>
        <copyright>All rights reserved, unless otherwise specified in the terms specified at https://video.ut0pia.org/about and potential licenses granted by each content's rightholder.</copyright>
        <atom:link href="https://video.ut0pia.org/feeds/video-comments.xml?videoId=6fb6a0c8-da86-477a-ad9a-e56db7bf1e05" rel="self" type="application/rss+xml"/>
    </channel>
</rss>