Companies around the world rely on SAP applications every day to manage the processes that keep their businesses running. Today, 84 percent of global trade volume runs through an SAP system.

Over decades, our customers have built a powerful digital basis with SAP to consistently handle business processes in their companies. They often adapt and expand these systems to meet their individual requirements. Many companies are now entering the next phase of transformation: They are modernizing their SAP landscapes to unlock the full potential of artificial intelligence (AI).

These companies are moving their SAP systems to the cloud and simplifying their IT structure. Additionally, they are preparing to integrate smart features directly into their daily operations. This creates new opportunities for automation: AI agents can work across different systems and take on increasingly demanding tasks.

Modernizing these systems and introducing AI on a large scale is a daunting task. This requires technologies that work with existing applications, function reliably in important processes and meet corporate governance standards.

That’s why we’ve partnered with over the last few years NVIDIA constructed. Our common goal: combine powerful AI technology with real business knowledge. In this way, we help companies advance their modernization and use AI where it really counts – in the applications and processes that determine business success. This collaboration was presented to a wide audience at the NVIDIA GTC conference.

AI for companies – creating the basis

Together with NVIDIA, we accelerate the entire AI lifecycle in the enterprise: from model development to runtime execution. This opens up a wide range of AI applications across our entire portfolio. An important building block here is NVIDIA NeMo – a collection of open libraries like NeMo Gym and NeMo RL. These tools make it possible to train large AI models much more quickly using reinforcement learning (RL). This allows teams to develop and optimize AI models for corporate use more quickly.

The models are in Generative AI Hub from SAP AI Core hosted. There, our customers select exactly the language models that best suit their requirements. The microservices NVIDIA NIM ensure noticeably higher performance when executing models – in our tests, up to 20 percent more than comparable open source solutions. With NVIDIA GPUs and NVIDIA NIM, companies can connect modern AI models with their proven SAP data and processes. In this way, they ensure that AI becomes an integral part of their workflows that control business operations.

AI models that fit business logic

AI models trained with SAP knowledge and based on NVIDIA technology are already helping companies implement their most important modernization projects faster. For example, the business logic that is in the SAP systems and controls daily operations can be further developed and adapted more easily.

For decades, companies have expanded their SAP systems with their own ABAP code – tailored to their very individual business processes. This logic contains the accumulated knowledge from years of daily work: from finance, supply chain, customer service and many other areas. But modernizing these system landscapes for the cloud and preparing them for the next generation of AI-driven innovations can prove extremely complex.

To drive this change, SAP has introduced a new AI model called SAP ABAP 1 developed. This model was trained exclusively with real ABAP program code and the business logic used in SAP systems. The solution combines various specialized AI models for different tasks: StarCoder2 helps to automatically complete code, while Codestral analyzes the code and explains it in an understandable way. These models run on NVIDIA NIM microservices, enabling fast and powerful processing.

With SAP Joule for Developers These functions are integrated directly into the development environment. Development teams can use it to: analyze and understand existing ABAP code, understand how individual adjustments are related to central business processes, and have new code created if necessary. This makes it easier for companies to interpret and update business logic that has evolved over decades – without losing the knowledge that makes their processes unique.

Connecting AI with business processes

SAP and NVIDIA are working together on how AI can be integrated into business processes – both in physical operations and in complex planning environments. One promising area is so-called Embodied AI. Here AI goes beyond pure software and interacts directly with the physical world. By combining AI with sensors, robotics and company data, real-world observations can be directly linked to digital business processes.

An example: If SAP Asset Performance Management When a warning message predicts a need for maintenance, robot checks are triggered that analyze systems using thermal, visual and acoustic signals. This data is compared with asset history and maintenance history to identify potential problems. joules then coordinates the next steps SAP Field Service Managementprioritizes work orders and provides technicians with all relevant information. When companies connect real-world data to their workflows, they can respond to changes in their environment and take appropriate action.

The same principle applies to complex planning environments. Today, supply chains have to take constantly changing factors into account: from supplier availability to transport interruptions to changing customer requirements. Together with NVIDIA, we are investigating how the latest AI technologies – such as NVIDIA Metropolis and NVIDIA Cosmos – can be used in warehouse management, security and asset testing.

We are also developing new functions for SAP Integrated Business Planningthe agent-based AI with GPU-accelerated optimization from NVIDIA cuOpt connect. Planners can use it to simulate complex supply chain scenarios and evaluate alternatives more quickly and accurately. By combining advanced optimization with SAP’s supply chain planning capabilities, companies can dynamically model constraints, flexibly adjust plans, and make informed decisions even in complex situations.

Working with large SAP customers helps identify operational bottlenecks and develop AI-powered solutions. At the NVIDIA GTC, SAP announced a collaboration with Foxconn. The Taiwanese electronics manufacturer will work with SAP to develop AI-powered innovations for production and supply chains.

By combining SAP’s enterprise applications and business context with Foxconn’s manufacturing expertise, companies can work more efficiently, become more resilient and make better decisions across complex production and delivery networks.

Agentic AI on the NVIDIA GTC

SAP is bringing Joule agents across its entire application portfolio – so companies can automate tasks and coordinate complex operations within their business processes. The NVIDIA GTC conference demonstrated how these capabilities apply Joule Studio on the SAP Business Technology Platform can be expanded to build agents for specific business scenarios.

Since SAP’s AI architecture is model-independent, companies can also integrate their own models – in addition to those provided via SAP AI Core. Hands-on training explained how companies can develop AI-powered workflows that run directly within the corporate systems they use to manage their business.

All of these innovations were introduced on the NVIDIA GTC which took place March 16-19, 2026 in San Jose, California. Visitors were able to experience live how SAP and NVIDIA help companies modernize their systems, adopt AI faster, and become AI-native companies:

Further information about SAP at the NVIDIA GTC.

Brenda Bown is Chief Marketing Officer of Business AI at SAP.