Case Study: PwC identifies up to 19% LLM training cost savings with PointFive

PwC identifies up to 19% LLM pipeline cost reduction with PointFive

The customer, PwC, faced the challenge of mapping and optimizing a complex, multi-region LLM training pipeline for a custom 32-billion-parameter model across eight AWS regions. Their AI research team needed to identify hidden waste across compute, storage, and data transfer components, but traditional FinOps tools were unable to provide the necessary end-to-end visibility. They turned to vendor PointFive and its DeepWaste™ Detection Engine to investigate the pipeline.

PointFive's solution mapped the entire pipeline, attributing costs and data flows across all regions and components. This analysis identified actionable inefficiencies, leading to an estimated $108K to $180K in annual savings opportunities, representing an 11-19% cost reduction. PointFive delivered a prioritized remediation plan, established monitoring for GPU usage, and provided PwC with full visibility into their mission-critical training infrastructure.

PWC