NVIDIA: European Retailers Cut AI Costs and Simulate Warehouses

Open models deliver frontier-level results at lower cost

NVIDIA says retail leaders and technology companies across Europe are using NVIDIA-accelerated open and frontier models for agentic commerce, digital product discovery and retail research and development.

Shopify post-trains open models using the NVIDIA Megatron Bridge library for faster fine-tuning and NVIDIA NeMo RL for reinforcement learning. It post-trained NVIDIA Nemotron 3.5 Lightning for the GraphQL subagent inside Sidekick, its AI assistant for merchants. The subagent retrieves store data for several features, including Sidekick Pulse, which researches a merchant's store and recommends improvements. On Shopify's internal benchmark, the model surpasses a leading frontier model in accuracy, cuts serving costs by 90% and responds 7x faster, according to NVIDIA.

NVIDIA CCluster vLLM inference optimization software also helped Shopify reach 4.5x faster performance and up to 12x lower inference cost across five other use cases: catalog enrichment, merchant assistant, e-commerce shopping simulation, product taxonomy classification, and merchant and buyer recommendations.

Other examples include:

• Zooplus: with Infosys and NVIDIA technologies, the online pet retailer built a production-ready prototype in its AWS environment that reached 86% search accuracy, using the NVIDIA Nemotron Retriever library and the NVIDIA Blueprint for retail catalog enrichment.
• Photoroom: its Image Editing API standardizes about 40,000 primary images a day across 70 countries for Vestiaire Collective. Listings it processed sold at a 5% higher rate of liquidity, with 3% to 7% more unique views per product page.
• Mirakl: it runs open models on NVIDIA GPUs on its own infrastructure and through European cloud providers such as Scaleway. Its Catalog Transformer has onboarded more than 70 million products.

Digital twins for brands, research and stores

Omi launched ProductDrop, a workflow that anchors every generated image to an approved digital twin, so brands can scale creative output across catalogs without introducing product errors. L'Oréal uses the approach within its CreAItech program. Omi runs TRELLIS 3D asset generation on NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, which speeds up inference by 10x and reduces pipeline latency by 4x, NVIDIA says.

L'Oréal is also integrating NVIDIA ALCHEMI into its research workflow, moving from months of experimentation to hours of computational screening of ingredient combinations. It is testing OrbMol-v2, an open source molecular simulation model from Orbital Industries, inside CurieOS. According to NVIDIA, OrbMol-v2 is five times faster than comparable models and has 21% lower error than the next-best model of its kind. Researchers can use it to identify promising combinations for solubilizing UV filters before lab testing.

ALLSIDES launched its 3D Digital Twin Factory, a pipeline that turns physical objects into SimReady 3D digital twins in minutes. The platform serves brands including adidas, Amazon, Nike, Zalando and Zara.

Simulation reaches the warehouse and the fitting room

Otto Group extended its autonomous warehouse under human supervision initiative, built with Reply on NVIDIA Omniverse libraries and running on Google Cloud, to its largest fulfillment center in Haldensleben, Germany. New capabilities let logistics teams design, test and optimize operations entirely in simulation before any physical implementation.

CATCHES reported production results from its RealFit deployment with AMIRI: conversion up 55%, add-to-cart up 80% and gross sales up 56%, after seasonal adjustment against like-for-like styles. The company also launched PredictiveFit, which extends physics-based virtual try-on to retailers, and CASI, a personal fashion agent that shops, styles and sizes for customers over WhatsApp. CASI is powered by open weight models including NVIDIA Nemotron 3.

New retail blueprints in early access

At GTC Berlin, Nebius is showcasing three NVIDIA Blueprints for retail, all in early access and running on Nebius infrastructure:

• Agent Harness for Employee Productivity: an agentic assistant combining scheduling, task management, inventory alerts and operational insights for store managers.
• Inventory Simulation and Optimization: models demand scenarios and tests replenishment strategies to help reduce overstock and stockouts.
• Agentic Finetuning Flywheel: a self-improving loop that generates training data, fine-tunes open source models and feeds real-world performance into the next training cycle.

Retailers interested in the blueprints can request early access. The results come from NVIDIA's own account and the companies' internal measurements, but they indicate where European retail is putting effort: cheaper inference with open models, and testing operations in simulation before committing in the physical world.