o9 Solutions Demonstrates Over 10x Faster Optimization for Enterprise Supply Chain Planning with NVIDIA cuOpt
o9 Solutions announced that NVIDIA cuOpt — NVIDIA’s GPU-accelerated optimization solver — delivers over 10x faster solve times for large-scale supply chain challenges with near-parity solution quality when integrated into the o9 platform and run on NVIDIA Blackwell accelerated computing infrastructure.
o9’s benchmark results from its ongoing collaboration with NVIDIA was conducted on a production-representative supply chain planning dataset comprising approximately 30 million variables, 15.7 million constraints, and 94 million nonzeros — the scale of complexity typical of large-enterprise inventory optimization and production planning workloads. NVIDIA cuOpt v26.06 on NVIDIA B200 GPUs completed the optimization solve in 57.4 seconds with optimal solution status. The same problem solved by a CPU-based solver required 661.7 seconds, resulting in over a 10x reduction in solve time and cutting a near-11-minute compute time to under one minute.
Critically, solution quality was maintained. The cuOpt objective value was within 0.008% of the CPU solver’s optimal solution with both runs reporting optimal status, demonstrating that the speed gains do not come at the expense of solution quality.
The cuOpt integration was enabled by o9’s solver-agnostic platform architecture, which allows enterprise clients across a wide range of industries — including consumer goods, retail, high tech, and industrial manufacturing, and more — to deploy and combine best-in-class solvers based on problem type, scale, and performance requirements without re-platforming. o9 continues to support multiple optimization engines for different workload types, ensuring customers have access to the right technology for each planning problem automatically.
“The world of optimization is evolving rapidly, and our Digital Brain platform for planning and decision-making is built to evolve with it. Demonstrating a 10x-plus speedup on a real, 30-million variable supply chain — with best-in-class solution quality — is a meaningful proof point for what GPU-accelerated planning can deliver. Our architecture is uniquely positioned to allow o9 enterprise clients to immediately benefit from our collaboration with NVIDIA,” said Chakri Gottemukkala, Co-Founder, CEO, and Chairman, o9.
For large enterprises, optimization solve time directly constrains planning agility. A production planning or inventory optimization workload that previously required an overnight batch compute window can now be solved on demand in under a minute, enabling supply chain teams to run more scenarios, respond to disruptions faster, and embed optimization more continuously into their planning cycles.
The 10x speedup demonstrated on o9’s production-scale supply chain dataset represents a shift in the economics of enterprise optimization — from a scarce, scheduled compute resource to one that can be run iteratively against real-time data within active planning workflows.
Ashwin Rao, Chief Technology Officer, o9, said, “These benchmark results reflect what our collaboration with NVIDIA has been building toward. For large-scale Linear Programming (LP) workloads, including inventory optimization, production planning, and network flow, cuOpt on the NVIDIA B200 delivers both the speed and the solution quality that enterprises require. And because the o9 platform is solver-agnostic, our clients can access this today without changing their planning environment.”
The collaboration between o9 — a member of the NVIDIA Inception program for startups — and NVIDIA extends beyond LP optimization. For Mixed-Integer Linear Programming (MILP), which governs some of the most complex planning decisions in production scheduling, facility location, and network design, NVIDIA’s cuOpt MILP solver is under active development and showing meaningful early progress. As this capability matures, o9 customers will be positioned to benefit from GPU-accelerated MILP performance through the same solver-agnostic platform framework. Additionally, the o9 collaboration with NVIDIA is exploring the frontier of Nonlinear Programming (NLP) — optimization problems that involve nonlinear objective functions or constraints, which arise in advanced demand sensing, pricing optimization, and AI-driven planning scenarios.
