IA & ÉnergieRiservato agli abbonati Jun 24, 2026 at 18:358Aggiungi ai preferiti

In the AI compute value chain, CoreWeave outsources its storage infrastructure to Backblaze for $355M. A deal that reveals how specialized GPU cloud stacks are built outside hyperscalers.
CoreWeave, the leading Nvidia GPU lessor for AI training, signed a multi-year cloud storage deal valued at $355 million with Backblaze on June 24, 2026. This partnership illustrates the strategy of specialized GPU clouds: assembling a stack through partnerships rather than vertical integration like AWS or Azure.
CoreWeave remains deliberately focused on GPU compute (Nvidia H100/GB200) and outsources storage to optimize its structural costs. Backblaze charges significantly less for storage than AWS S3 based on both providers' public pricing- a material gap at AI-scale volumes. This assembly model-network (Zayo/Telia), storage (Backblaze), compute (Nvidia)-allows specialized GPU clouds to maintain margins against hyperscalers. The deal also legitimizes Backblaze as an enterprise AI-grade provider, a segment the company has sought to penetrate since 2024.
For CRWV holders, outsourcing confirms operational scalability, but the debt structure remains the number-one risk.
CoreWeave’s dependence on Nvidia for GPU supply. Risk of repricing if hyperscalers subsidize storage to retain AI workloads. Cost pressure on CoreWeave may lead to premature renegotiation.
Backblaze’s Q2 2026 results (August 2026). Similar contract announcements from Lambda Labs or Crusoe Energy. Evolution of CoreWeave’s debt and its structural covenant in upcoming quarterly reports.
Crea un account gratuito per accedere a tutti i nostri contenuti e alla rivista settimanale.
Articolo prodotto da intelligenza artificiale, riletto sotto controllo editoriale umano.
Accedi per partecipare alla discussione.
This deal shows how AI cloud stacks are getting smarter-outsourcing storage to focus on what they do best. Smart move, but curious how costs shake out long-term.
Backblaze cashing in on AI hype-let’s see if this deal holds up when the GPU bubble pops like dot-com 2.0.
355M to store what? Another case of AI hype funneling cash into infrastructure before actual demand catches up.
Un partenariat malin qui montre que l'IA a besoin de solutions agiles, pas que des géants du cloud. L'Afrique pourrait s'inspirer de ces modèles décentralisés.
Smart move or just kicking the storage can down the road? Either way, someone’s making bank. But what do I know?
Интересный ход, но без анализа долгосрочной рентабельности сделка выглядит как очередной «хайп» вокруг ИИ-инфраструктуры.
Interesante ver cómo se segmenta el stack de IA: CoreWeave externaliza almacenamiento para optimizar costes, pero ¿a qué precio en latencia y control de datos?
355M to outsource storage? Sounds like CoreWeave’s margins are already thinner than their GPU utilization reports.
GPU cloud specializzati: mega-contratti e consolidamento del mercato del compute IA