The briefing begins with reported financing for Crusoe and Fluidstack. Crusoe's case centers on a large cloud agreement attributed to Jane Street, while Fluidstack is described as building facilities for a major Anthropic purchase of Google tensor processing units. Together, the examples show capital flowing into data centersAI compute infrastructure is the hardware, facilities, networks, power, cooling, storage, and software used to train and run AI models. and alternatives to an exclusively Nvidia-based stack.
Gimlet Labs is presented as an inference-infrastructureInference is the use of a trained model to produce an output from supplied input. company that assigns different parts of a model workload to hardware suited to each task. Light addresses a different layer by combining cameras, sensors, chips and spatial software so robots can perceive depth, motion and position in real time.
Watch on YouTube
