Tag
A company shares the practical difficulties of purchasing and maintaining an HGX B300 for AI inference, including high cost (€1.1M), power and space requirements, and limited availability, with throughput estimates for GLM 5.2.