Enterprise Optimization Algorithms Scale to Massive ComplexityNVIDIA cuOpt mPDLP scales linear programming to 100M+ variablesTechnology Services › Scientific Research and Development Services › Research and Development in the Physical, Engineering, and Life SciencesDataNVIDIA introduced mPDLP, a multi-GPU solver for linear programming problems that distributes optimization across NVLink-connected GPUs, achieving nearly 10x faster solve times on benchmark problems and up to 6x lower peak memory usage per GPU. For supply chain and energy-grid planning teams, this breakthrough removes the single-GPU bottleneck that historically forced large optimization problems to take hours to converge, enabling faster decision cycles in real-time planning workflows.Oct 8, 2026Read full article →
AI Agents Transform Supply Chain and LogisticsNVIDIA and Palantir codify supply chain expertise with NemotronDataLLMNVIDIA and Palantir built a system combining Palantir Foundry's data layer, NVIDIA's cuOpt optimization, and a post-trained Nemotron 3.5 Lightning model to automate weekly material allocation decisions across NVIDIA's complex global supply chain. For commerce practitioners, this demonstrates how to embed human expert judgment into AI workflows, compressing time-to-rack while maintaining human oversight of critical operational decisions.Sep 11, 2026Read full article →
