AI’s power crunch is sending data-centre builders to the banks
THE SO WHAT
The scarce resource has flipped from GPUs to megawatts—data-center builders are now raising billions in project finance just to secure power. If you’re planning AI capacity, treat long-term power purchase agreements and grid interconnects as first-class roadmap items, not back-office procurement.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIAnthropic signs a $10bn compute deal with a week-old cloud startup
A $10B, six-year compute deal with a week-old cloud startup shows how desperate top labs are to lock in power and capacity—even if it means betting on new providers. If you’re a heavy AI user, start mapping alternative compute paths now instead of assuming the big three clouds will always have room for you.
Applied AIWorld Bank Urges Emerging Nations to Embrace AI Amid Weak Growth
The World Bank telling emerging economies to “move fast” on AI is a signal that multilateral pressure will now push AI adoption as a development lever. If you operate in these markets, expect new incentives, pilot programs, and regulatory frameworks that favor AI-enabled vendors.
Applied AIIs the future of data centers portable? Runware builds a pod to find out
Modular “Sonic Inference Pod” data centers are a bet that AI workloads need to be dropped where power and cooling are available, not just where hyperscalers already sit. If you’re capacity-constrained, portable pods could become a real option in your infra mix within a planning cycle or two.
Applied AIExclusive: Nscale to provide compute to Oxford, UCL and Imperial AI researchers
Dedicated Nscale capacity for Oxford, UCL, and Imperial means UK frontier research is getting quasi-hyperscaler treatment—academic labs will prototype workloads that look like tomorrow’s commercial demand. If you build AI infra or tools, treat these labs as upstream customers and early signal on what next-gen clusters need to support.