PepsiCo taps Siemens and NVIDIA digital twins, sees 20% throughput lift

PepsiCo's move fits a wider pattern among consumer goods companies testing digital twins and AI on packaging and plant decisions. Nestlé unveiled an in-house AI service last summer, built with Accenture Song, that generates 3D replicas of physical products and has reached a baseline of 4,000 3D digital master products, mainly for brands like Purina, Nescafé Dolce Gusto and Nespresso. The following month, Unilever, Amcor and Asahi disclosed trials of Deepnest, Greyparrot's AI waste intelligence platform, which uses AI-camera systems in material recovery facilities to track whether packaging units are sorted, recycled or lost.

PepsiCo's collaboration goes further upstream, into facility design itself. The company says it is shifting to a digital-first planning strategy that uses physics-based digital twins and AI agents as co-designers, recreating every machine, conveyor, pallet route and operator path with physics-level accuracy. Siemens' Digital Twin Composer, built on NVIDIA Omniverse libraries, is the vehicle for that: a way to combine 2D and 3D digital twin data with real-time physical information in a photorealistic virtual scene.

What is not yet settled is how far the early U.S. results travel. The throughput, design-validation and capex figures come from initial deployment, and PepsiCo has said only that it plans to scale globally. Whether the same gains appear across other facilities and regions will depend on how well the digital models match local equipment, constraints and operating practices — and that remains the test to watch.