"The next leap forward" - Nvidia is investing $100bn in OpenAI, and will start by deploying as much power for 10 nuclear reactors (www.techradar.com)

Nvidia announced a massive infrastructure and investment deal with OpenAI: it will deploy at least 10 gigawatts of Nvidia-powered systems — equating to millions of GPUs and roughly the electrical output of ten nuclear reactors — and will progressively invest up to $100 billion as each gigawatt comes online. The first gigawatt, using Nvidia’s Vera Rubin platform, could be operational as early as H2 2026. OpenAI named Nvidia its “preferred strategic compute and networking partner,” committing both companies to co‑optimize hardware, software and networking roadmaps; the collaboration is also linked to OpenAI’s larger Project Stargate, a $500 billion push to build new AI infrastructure with Nvidia and Oracle. This is significant because compute remains the main limiter of model scale and capability: provisioning multi‑gigawatt systems accelerates training and inference of next‑generation foundation models and could materially shorten timelines to far more capable AI systems. Technical implications include massive data‑center power, cooling and networking demands, tighter hardware‑software co‑design, and potential vendor lock‑in from concentrated compute supply. The deal marks arguably the largest AI infrastructure project ever and will reshape where and how cutting‑edge models are developed, with both economic upside for AI deployment and open questions about concentration, energy use, and geopolitical risk.