VCs build shared B300 pools

- Social posts report VCs and cloud builders are assembling shared B300 GPU pools with eight 275GB HBM3e GPUs per node for inference workloads. - The move shifts CPU:GPU ratios from 1:8 toward 1:1, promising better utilization but raising multi-tenant isolation and performance variability concerns for shared inference. - VCs and startups see pooled clusters as a way to avoid long GPU commitments but need sandboxing. (x.com) (x.com)