DATA CENTERS BUILT IN WEEKS

Summary from the TLDR Hardware newsletter

Runware is building its own infrastructure to make AI compute cheaper and more abundant. Sonic Inference Pods pack up to 1200 GPUs into modular 1MW data centers, deployed in weeks rather than years and purpose-built to squeeze more performance from every GPU. Run any AI workload on them through Runware Serverless [8].