CoreWeave’s Brannin McBee on the future of AI infrastructure, GPU economics, & data centers | E1925

AI software adoption is outpacing current cloud infrastructure, which was originally designed for serializable workloads rather than the parallelizable demands of modern AI. CoreWeave, represented by Chief Development Officer Brannin McBee, is addressing this by building specialized, high-performance GPU clusters at an unprecedented scale. The transition from model training to inference—where demand scales linearly with user volume—creates an immovable wall of compute requirements. Power availability and data center capacity represent the primary bottlenecks, necessitating a shift toward liquid cooling and non-blocking InfiniBand fabrics to maintain performance. While alternative hardware and open-source solutions exist, NVIDIA’s integrated software ecosystem, specifically CUDA, remains the industry standard for complex, large-scale AI workloads. This massive infrastructure build-out is expected to remain a critical, resource-constrained challenge through the end of the decade.