The primary constraint to deploying new clean energy in the US is no longer technology cost—it is grid interconnection.

CarbonForge tackles this bottleneck by optimizing energy at the MLIR compiler layer. Their optimization engine plugs into existing workflows to significantly reduce joules per token—no model changes, no hardware swaps—while also improving latency.