Traditional, centralized hyperscale facilities are increasingly straining power grids, prompting a move toward distributed edge networks. Lektra is addressing this bottleneck by building an infrastructure layer that deploys AI capacity wherever energy is available. By integrating Thunder Compute’s orchestration platform, Lektra can now virtualize GPUs across this decentralized network, significantly boosting hardware utilization and deployment speed.

Karl Andersen, CEO of Lektra, argues that the next phase of artificial intelligence will not be sustained by a handful of massive data centers. Instead, he envisions an energy-aware architecture that scales through geographic distribution. The combined technology stack allows cloud providers to manage high-performance workloads across multiple sites, offering a resilient alternative to the grid-locked models currently dominating the market. This shift toward what the companies describe as Physical AI reflects the growing need to align computational growth with the practical constraints of modern energy infrastructure.