How EPRI's Flex MOSAIC Framework Is Rewriting the Rules of Data Center Interconnection The race to power AI isn't just about gigawatts. It's about time. And right now, time is the one resource the data center industry can't afford to waste. As artificial intelligence and advanced manufacturing push electricity demand to levels the grid was never designed to handle, a single bottleneck has emerged at the center of nearly every major project: Time-to-Power. Interconnection queu

For the past two years, data center site selection has been almost entirely driven by one variable: substation proximity. Developers have pushed into secondary and tertiary markets they would have passed over five years ago simply to secure MW commitments ahead of their competitors. That calculus is now insufficient. The shift toward generative AI workloads has exposed a second constraint that power-first site selection often overlooks: fiber. A site with 100 MW of available

For years, Canada was viewed by US developers as a reliable, if quiet, secondary market for disaster recovery. That changed in 2025. Today, as power queues in Northern Virginia and Columbus stretch into the next decade, Canada has emerged as a promising relief valve for the AI infrastructure data center supercycle. With a projected market value of $13 billion by 2030, the question isn’t if you should be in Canada (you probably should), it’s where the remaining power is hidden

The data center industry is no stranger to public pushback, but the scale of the newly approved Stratos Data Center Project in northern Utah represents a new frontier of data center development risk. As hyperscalers and AI developers hunt for massive tracts of land with power potential, the Stratos project, a staggering 40,000-acre proposal, serves as a cautionary tale of how quickly booming narratives can turn into PR and regulatory nightmares. The Stratos Data Center site,

Unlike generation assets driven primarily by resource availability, BESS site selection is fundamentally a grid- and market-constrained optimization problem. The best sites are not just “available land near transmission,” but locations where congestion, pricing volatility, interconnection capacity, and infrastructure adjacency align to support revenue stacking and scalable deployment.