Every day, companies around the world rely on SAP applications to run the operations that keep their businesses moving.  In fact, 84% of global commerce touches an SAP application.

Over decades, our customers have built powerful digital foundations on SAP to run end-to-end business processes across their enterprises—often extending and customizing these systems to support their unique business needs. Now, many are entering the next phase of transformation: modernizing their SAP landscapes to unlock the full potential of AI.

As companies move to cloud-based SAP environments and clean-core architectures, they are preparing to embed intelligence directly into business processes. This enables new forms of automation, with AI agents that operate across enterprise systems and execute increasingly complex tasks.

Modernizing these systems while introducing AI at scale is a significant undertaking. It requires technologies that integrate with existing applications, operate reliably within mission-critical workflows, and meet the governance standards enterprises demand.

That’s why, over the past few years, we have partnered with NVIDIA to combine advanced AI technology with deep business context. Our goal is to help organizations accelerate modernization and apply AI across the applications and processes key to their success. This collaboration will be showcased at NVIDIA GTC.

Building the foundation for enterprise-grade AI

Through our collaboration with NVIDIA, we are accelerating the entire life cycle of enterprise AI—from model development to high-performance runtime execution—and powering AI scenarios across our portfolio. NVIDIA NeMo™,  which consists of open libraries such as NeMo Gym and NeMo RL, helps accelerate large-scale model training across distributed RL environments. It enables teams to build and refine enterprise-grade AI models faster.

Models are hosted through SAP AI Core and generative AI hub, where our customers and partners leverage those best suited to their use cases. NVIDIA NIM microservices optimize inference performance, and we have observed up to a 20% improvement compared to another popular open source serving engine. Enabled by NVIDIA GPUs and NVIDIA NIM, the increased performance allows organizations to combine advanced AI models with trusted SAP business data and processes to ensure that AI operates within the workflows that drive business operations.

Modernizing the business logic that runs the enterprise

AI models trained on SAP knowledge and accelerated using NVIDIA technologies are already helping customers tackle some of their most pressing modernization challenges. For example, evolving business logic embedded in the SAP systems that run their operations.

For decades, organizations have extended SAP applications with custom ABAP code that reflects how their businesses operate. That logic captures years of operational knowledge across finance, supply chain, service processes, and more. But modernizing these environments for the cloud and preparing them for the next generation of AI-driven innovation can be complex.

To help accelerate this journey, SAP developed SAP-ABAP-1. a foundation model trained exclusively on real-world ABAP code and the business logic used across SAP environments. The solution incorporates specialized models for code-related tasks, including StarCoder2 for code completion, and Codestral for deeper code understanding and explanations. These models are served through NVIDIA NIM microservices to deliver high-performance inference.

SAP Joule for Developers brings these capabilities into the developer experience, helping teams analyze existing ABAP code, understand how customizations interact with core business processes, and generate new code when needed. By making decades of embedded business logic easier to interpret and update, we help organizations accelerate modernization and preserves the knowledge that makes their operations unique.

Connecting AI to business operations

The collaboration between SAP and NVIDIA also explores how AI can operate within enterprise workflows to help organizations apply intelligence across both physical operations and complex planning environments. One emerging area is embodied AI, in which intelligence extends beyond software systems into the physical world. By combining AI reasoning with sensors, robotics, and enterprise data, organizations can connect real-world observations directly with digital business processes.

For example, predictive maintenance alerts from SAP Asset Performance Management can trigger robotic inspections that analyze equipment using thermal, visual, and acoustic signals. These signals are evaluated alongside asset histories and maintenance records to identify potential issues. Joule then orchestrates follow-up actions through SAP Field Service Management, prioritizing work orders and guiding technicians with the right operational context. By linking physical-world insights with enterprise workflows, organizations can turn physical-world signals into coordinated enterprise actions.

The same principle applies to complex planning environments. Supply chains today must manage a constantly shifting web of constraints, from supplier availability and transportation disruptions to evolving customer demands. With NVIDIA, we are exploring technologies, such as NVIDIA Metropolis and NVIDIA Cosmos, to bring the latest AI advancements into warehouse management, safety, and asset inspection.

Together, we are also bringing new capabilities to SAP Integrated Business Planning that combine agent-based reasoning with the NVIDIA cuOpt GPU-accelerated optimization engine. This enables planners to simulate complex supply chain scenarios and evaluate alternatives with more speed and accuracy. By integrating advanced optimization with SAP’s supply chain planning capabilities, organizations can dynamically model constraints, adapt plans as conditions change, and make more confident decisions in increasingly complex environments.

Collaboration with large-scale SAP customers helps identify real operational bottlenecks, paving the way for AI-driven solutions. At NVIDIA GTC, SAP has unveiled a collaboration with Foxconn. The Taiwan-based global electronics manufacturer and manufacturing solutions provider will work with SAP to develop AI-powered innovations for manufacturing and supply chain operations.

By combining SAP’s enterprise applications and business context and Foxconn’s manufacturing expertise, organizations can enhance operational efficiency, increase resilience, and advance decision-making across complex production and supply networks.

Experience agentic AI at NVIDIA GTC

SAP is enabling Joule Agents across its application portfolio, helping organizations automate tasks and coordinate complex workflows within business processes. At NVIDIA GTC, visitors will see how these capabilities are extended using Joule Studio on SAP Business Technology Platform to build agents tailored to specific enterprise scenarios.

And because SAP’s AI architecture is model-agnostic, organizations can bring their own models into these workflows, in addition to those deployed through SAP AI Core. The hands-on experience at NVIDIA GTC will demonstrate how organizations can build AI-driven workflows that operate directly within the enterprise systems that run their business.