Kinaxis achieved up to a 12X reduction in end-to-end calculation time for supply chain optimization using NVIDIA cuOpt and AI infrastructure, cutting a semiconductor planning model’s runtime from >3 hours to ~17 minutes. The model handled nearly 50 million decision variables across 40,000 SKUs and a six-quarter daily horizon. Core solve time improved 23X, enabling interactive scenario iteration and strengthening agent-driven orchestration. Kinaxis serves 400+ global enterprises managing $200B+ in inventory and generating 250,000 scenarios monthly.
According to www.semiconductor-digest.com, Kinaxis Inc. has achieved up to a 12X reduction in total end-to-end calculation time for large-scale supply chain optimization using NVIDIA cuOpt and NVIDIA AI infrastructure within its Maestro platform.
Performance Gains on Semiconductor Planning Model
In benchmark testing on a semiconductor wafer planning use case, the optimized model handled nearly 50 million decision variables, spanning more than 40,000 SKUs across a six-quarter daily planning horizon. Planning cycles dropped from more than three hours to approximately 17 minutes. Core optimization solve time improved 23X, reducing compute time by more than 95% — shifting runtimes from hours to minutes while maintaining comparable solution quality.
Operational Impact for Global Planners
This acceleration enables a strategic shift from static, long-running batch processes to interactive scenario iteration — a core tenet of Kinaxis’ concurrent supply chain orchestration approach. With over 400 global enterprises relying on Maestro, and the platform orchestrating decisions across more than $200 billion in inventory and generating 250,000 scenarios monthly, reduced latency directly enhances responsiveness to demand volatility and production constraints.
Enabling Agent-Driven Orchestration
GPU-accelerated optimization strengthens Kinaxis’ agent-driven strategy: agent-based workflows trigger multiple optimization runs as they evaluate alternatives. Faster solve times expand the number of feasible scenarios agents can assess within tight operational decision windows — supporting more adaptive, real-time decision-making across extended supply networks.
Technical Integration & Industry Context
The advancement integrates:
- Up to 12X faster solve times for semiconductor wafer planning
- GPU-accelerated optimization using NVIDIA cuOpt
- Execution on NVIDIA AI infrastructure for industrial-scale models
- Native integration within the Maestro platform, unifying data, people, and decisions
This development follows broader industry momentum toward hardware-accelerated supply chain analytics. SAP has integrated GPU-accelerated forecasting into its Integrated Business Planning suite; Blue Yonder reported similar speedups using NVIDIA GPUs for demand sensing in 2023. Meanwhile, Gartner’s 2024 Hype Cycle for Supply Chain Technology identified “AI-augmented supply chain planning” as entering the Slope of Enlightenment — with performance at scale cited as a key adoption barrier now being addressed through infrastructure innovation.
“This milestone demonstrates how accelerated computing can change the way large-scale planning problems are solved. When optimization shifts from hours to minutes, organizations gain the ability to iterate more frequently and evaluate more alternatives. That iterative speed is essential to enabling concurrent supply chain orchestration and advancing our agent-driven strategy.” — Gelu Ticala, Chief Technology Officer at Kinaxis
“The increasing complexity of global supply chains demands a fundamental shift to accelerated decision-making. By integrating NVIDIA cuOpt into its Maestro platform, Kinaxis is empowering customers to achieve planning agility and scenario iteration to help navigate rapid change.” — Alex Fender, Director of Decision Intelligence at NVIDIA
For supply chain professionals, these gains translate into tangible workflow improvements: planners can now test sensitivity to raw material shortages, port delays, or demand spikes within a single shift — not across days. The reduction in computational latency also lowers the barrier to embedding prescriptive logic into frontline tools, supporting decentralized decision rights without sacrificing model fidelity. As Kinaxis co-presents the results at NVIDIA GTC 2026, practitioners should anticipate tighter integration between optimization engines and execution systems — particularly in capital-intensive, high-variability sectors like semiconductors, automotive, and pharmaceuticals.
Source: www.semiconductor-digest.com
This article is compiled from international media sources by the SCI.AI editorial team.
