OpenAI and Nvidia’s $100B AI plan will require power equal to 10 nuclear reactors (arstechnica.com)

OpenAI and Nvidia announced a strategic partnership to deploy at least 10 gigawatts of Nvidia systems for OpenAI’s AI infrastructure, with Nvidia committing up to $100 billion as the rollout proceeds. The first gigawatt of systems—based on Nvidia’s Vera Rubin platform—is slated to come online in the second half of 2026. Nvidia CEO Jensen Huang said the 10 GW target would require roughly 4–5 million GPUs (about the company’s current annual shipment volume), and Nvidia’s stock jumped nearly 4% after the announcement. OpenAI framed the investment as foundational: “Everything starts with compute,” CEO Sam Altman said, citing compute as the basis for the future AI-driven economy. Technically and operationally, this is a moonshot-scale build: 10 GW of capacity equals roughly ten nuclear reactors and would dwarf existing hyperscale data centers (most of which draw 50–100 MW). That raises immediate challenges—electricity sourcing, grid upgrades, cooling, rack-to-chip networking, and supply-chain scaling for accelerators and supporting silicon—plus environmental, siting, and regulatory implications. For the AI/ML community, the plan signals enormous demand for specialized GPUs, networking, and software stacks, and solidifies Nvidia as OpenAI’s preferred compute partner while complementing OpenAI’s other cloud relationships. The scale is unprecedented and unproven, but if realized it could materially shift economics and access to large-scale model training and inference.