From Concrete to Compute: Why Clichmont Is Building AI Infrastructure Instead of Renting It
Clichmont is making a different bet from GPU-rental neoclouds like CoreWeave, Crusoe, and Lambda. Instead of focusing mainly on leasing chips, it wants to own the data centers, power, cooling, connectivity, and supply chain that make AI compute possible. The core thesis is that GPUs matter, but infrastructure constraints such as electricity and site capacity may become the deeper bottlenecks. That model gives more control over successive generations of AI hardware, but it also demands far more capital, longer planning cycles, and tight coordination across land, power, construction, and customer demand. Site selection depends on more than cheap energy: climate, land availability, grid access, and operating efficiency all matter. The interview centers on why ownership might create a more durable advantage, what capital-light rivals may be overlooking, what risks come with building physical infrastructure, and why a token exists in an infrastructure-first company.
