Automotive

Horse Powertrain Builds One of Europe's First Automotive AI Factories to Cut Time to Market

Objective

Horse Powertrain is a global leader in advanced powertrain systems, employing 19,000 engineers across five R&D centers worldwide. As electrification and cost pressure reshaped the automotive industry, the company needed to move faster—across design, inspection, and production—without sacrificing the precision its customers demand.

By deploying one of Europe's first AI factories purpose-built for the automotive industry, anchored by NVIDIA Accelerated Computing, Horse Powertrain is compressing the length of design cycles and automating quality inspection with real-time AI.

Customer

Horse Powertrain

 

Use Case

Google Cloud
Deloitte
HPE

Topic

Technology
AI Factory
Digital Twins
Industrial AI

 

Key Takeaways

  • ~40% reduction in time to market across engineering and production
  • Design changes evaluated in seconds instead of days using NVIDIA Modulus surrogate models
  • Real-time engine inspection deployed on the production line with AI-generated synthetic training data

Competing in an Era of Rapid Automotive Transformation

Horse Powertrain is an independent company, formed as a joint venture between Renault Group and Geely, built to advance three technology pillars:

  1. Ultra-low-emissions combustion engines
  2. Hybrid and electrified powertrains
  3. Powertrains compatible with next-generation fuels, including eFuel, ethanol, and hydrogen

As electrification, hybridization, and cost pressures converged, the company needed to accelerate on every front simultaneously.

The barriers were structural: engineering teams siloed across multiple international sites, computational fluid dynamics (CFD) and finite element analysis (FEA) simulations that took hours or days per run, manual quality inspection on the production floor, and no real-time intelligence to optimize warehouse conditions.

What Horse Powertrain needed was a unified, scalable AI infrastructure that could turn simulation, sensor data, and operational intelligence into real competitive advantage—without sacrificing the precision and reliability demanded by the automotive industry. Early AI pilots had surfaced the hard lessons: isolated prototypes that could not scale beyond a single site, no organizational structure or governance to sustain them, and data fragmented across functions in incompatible silos. Without unified data governance, true intelligence could not emerge.

Horse Powertrain

Horse Powertrain

Launching kAIros: Building an Automotive AI Factory With NVIDIA

The answer was kAIros, a company-wide AI transformation initiative led by Horse Powertrain’s Horse Technologies division at its Valladolid, Spain, facility and designed from the outset to scale the entire enterprise.

At its core is an AI factory, a supercomputing environment powered by NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs deployed on premises with NVIDIA RTX PRO Servers via HPE, complemented by instances running on Google Cloud for additional scale and capacity. The HPE AI Factory with NVIDIA  infrastructure is unified by NVIDIA's full AI software stack, including NVIDIA CUDA-X, NVIDIA Omniverse, and NVIDIA Cosmos, enabling the company to rapidly develop and deploy AI applications across engineering and production.

Horse Powertrain also established a dedicated Center of Excellence to scale AI capabilities across the organization, bringing together cross-functional teams and developing new applications spanning agentic and industrial AI, including model training, simulation, digital twins, and algorithm optimization. Co-designed with Deloitte, the COE operates across six strategic pillars—including data governance, talent, and AI sovereignty—giving Horse defined ownership over which AI use cases are pursued, how models are validated and deployed, and who bears accountability when AI systems make consequential decisions in live production environments.

From this foundation, four flagship applications have been developed and deployed:

  • AI-Powered Final Engine Inspection: Using NVIDIA Metropolis and NVIDIA DeepStream, Horse Powertrain deployed a real-time engine inspection system on the production line. Synthetic training data allows the system to detect surface defects and quality deviations with high accuracy. A human-in-the-loop interface enables operators to continuously refine model accuracy, creating a self-improving inspection pipeline.
  • Warehouse Digital Twins: An interactive, physics-accurate warehouse digital twin was built inside NVIDIA Omniverse using NVIDIA Isaac Sim™. The twin integrates real-time data with WMS/ERP systems, IoT sensors, and cameras, surfacing live KPI dashboards and heatmaps for operations teams through an interactive 3D interface.
  • Warehouse Video Search and Reporting Agent: Built on NVIDIA VSS Blueprint with Metropolis and DeepStream, and NVIDIA Cosmos Reason VLM, this agent continuously analyzes warehouse video feeds and enables operators to query footage in natural language—connecting real-world factory operations in real time to the Omniverse digital twin to support autonomous decision-making for cobots, automated guided vehicles, and smart machinery on the production floor.
  • GPU-Accelerated Engineering Simulation: Engineers now run high-fidelity CFD and FEA simulations using multi-GPU solvers on the AI factory infrastructure. NVIDIA Modulus surrogate models allow the team to evaluate engine design changes in seconds rather than days, radically compressing the design iteration loop and enabling far more design candidates to be explored before physical prototyping.

Horse Powertrain

Accelerating Every State of Engineering and Manufacturing

The results speak to the breadth of the platform. Design changes that once required days of CFD or FEA computation can now be evaluated in seconds using NVIDIA Modulus surrogate models, enabling engineers to explore far more candidates before committing to physical builds. On the production floor, AI-powered defect detection has replaced manual QA—with synthetic training data making it possible to train high-accuracy models even where real defect imagery is scarce.

Across the organization, kAIros is targeting:

  • ~40% reduction in time to market across engineering and production
  • 40% reduction in low-value process work
  • 25% improvement in design cycle efficiency

“The complexity of the current situation in the automotive industry, coupled with increasingly demanding regulatory and market requirements, compels us to accelerate without compromising on precision. That is the purpose of kAIros: to break through the current limits of competitiveness. As a result, AI has evolved from being an isolated tool into a strategic pillar that redefines our speed, efficiency, and industrial competitiveness on a global scale.”

Patrice Haettel
CEO, Horse Technologies

Scaling AI Across 17 Countries of Advanced Manufacturing

Valladolid is the first site, not the last. Horse Powertrain is extending the kAIros AI Factory infrastructure across other countries in which the company operates, with new applications in agentic coordination, predictive maintenance, and advanced simulation already in the pipeline. The Center of Excellence gives Horse Powertrain the internal capability to deploy these use cases at pace, without external dependencies slowing each iteration.

For the automotive industry broadly, kAIros demonstrates what a purpose-built AI factory can do: serve simultaneously as the engine for engineering simulation, production quality, warehouse intelligence, and enterprise workflow automation—from a single, coherent infrastructure. The next step is already in motion: warehouse robots that will move beyond programmed routes to independently perceive, decide, and act—not just using AI as a tool, but operating as genuine AI collaborators on the factory floor. For an industry in the middle of its deepest transformation, that ambition is exactly the point.

Learn more about NVIDIA RTX PRO AI Factory.

Related Customer Stories