$100B Nuclear AI Data Center: Brookfield & NextEra Build Massive Energy Complex
Canada's Brookfield and US utility NextEra Energy are investing $100 billion to transform a former nuclear enrichment site in Paducah, Kentucky into the world's largest AI data center campus.
I've been monitoring the AI compute bottleneck closely over the past two years, but today's announcement represents a shift in scale that fundamentally alters global tech infrastructure. Canadian asset manager Brookfield is partnering with NextEra Energy — America's largest electric utility — to build a mega-scale $100 billion nuclear AI data center complex in Paducah, Kentucky.
This isn't just another server warehouse. It is a full-scale energy and compute ecosystem built directly on top of a decommissioned nuclear enrichment site.
Why Nuclear Power is the Only Answer for Frontier AI
The math behind 2026 frontier AI training is brutal. Big Tech is projected to spend over $700 billion on AI infrastructure this year alone. But the primary bottleneck isn't GPU supply anymore — it's gigawatts of clean, continuous baseload power.
Traditional grid infrastructure cannot support multi-gigawatt clusters without causing regional blackouts. By redeveloping the former Paducah Gaseous Diffusion Plant, Brookfield and NextEra are tapping into preexisting high-voltage transmission infrastructure capable of routing gigawatts of dedicated nuclear power directly into AI server halls.
The Hardware and Compute Architecture
The Paducah complex will feature liquid-cooled cluster architecture optimized for next-generation hardware like Nvidia's Vera Rubin platform and AMD's Strix Halo clusters. Combining zero-carbon nuclear energy with ultra-dense rack designs solves both the carbon footprint problem and the thermal dissipation challenge simultaneously.
For developers and researchers building on-device AI or managing hybrid cloud pipelines, having clean, continuous cloud compute behind your APIs ensures long-term price stability. If you're testing AI workflows locally before scaling to cloud infrastructure, running local models on a compact edge server like the ASUS NUC 14 Pro AI (around $899) is an ideal way to prototype before deploying to multi-gigawatt clusters.
The race for AI dominance has officially entered its infrastructure era. And the winner won't just be who has the best algorithm, but who controls the energy grid.
David tests AI tools, gadgets, and developer platforms hands-on before writing about them. His work focuses on making complex tech approachable — without the hype. He has covered 100+ products across AI, gadgets, and software for TechPixelly.
