With a quick search, it looks like some ai datacenters are not designed/built with hyperscaler grade reliability in mind so the would have to be upgraded for th

With a quick search, it looks like some ai datacenters are not designed/built with hyperscaler grade reliability in mind so the would have to be upgraded for that purpose. https://blogs.microsoft.com/blog/2025/11/12/infinite-scale-the-architecture-behind-the-azure-ai-superfactory/?hl=en-CA#:~:text=We%20are%20pushing%20the%20envelope,at%203%C3%979%20cost. https://blogs.microsoft.com/blog/2025/11/12/infinite-scale-t... We are pushing the envelope in serving this compute with cost-efficient, reliable power. The Atlanta site was selected with resilient utility power in mind and is capable of achieving 4×9 availability at 3×9 cost. By securing highly available grid power, we can also forgo traditional resiliency approaches for the GPU fleet (such as on-site generation, UPS systems and dual-corded distribution), driving cost savings for customers and faster time-to-market for Microsoft.