There’s a growing amount of excitement and trepidation around off-grid data centers. I’m not convinced many will actually go fully off-grid -- probably as a bridge in most cases. But it’s surfacing a more interesting tension about how we build the system. As @ceboudreau detailed, that tension played out on stage at CERAWeek between Fermi and Google. Fermi is leaning into the “power island” model that pairs data centers with dedicated generation, starting with gas and potentially nuclear. The idea is to move faster, avoid interconnection bottlenecks, and not push costs onto the broader system. Google pushed back on that framing. Their view is that once you start building isolated systems, you end up overbuilding for reliability and tie up more capital into assets that run a fraction of the time. And none of that investment actually improves the grid. There’s also a concern that it could lead to higher system costs over time if those resources aren’t shared. We will clearly see some version of both of these approaches, but I agree more with Google's @AmandaCorio on this one: “We are focusing too narrowly on this moment and the bottlenecks, and not thinking collectively as a system in 10 years."