SAP highlighted at Sapphire a host of technology partnerships to enable their AI strategy. In doing so, it seems to have learned its lesson of not wanting to develop everything itself. All major IT players in the AI field are on SAP’s partnership list.
SAP is one of the most important tools for the business user. In it, a user often has to perform many repetitive actions or design complex processes. This should become much easier in the future with SAP Joule, SAP’s Copilot.
SAP needs high-end technology to give Joule the knowledge, intelligence, and power to be a good CoPilot. They probably would have tried to make that technology themselves in the past, but now they are closing many partnerships. Among others, partnerships with AWS, Google Microsoft, Meta and Nvidia pass by during SAP Sapphire in Orlando.
SAP is partnering with AWS on multiple fronts
On his second official working day as CEO of AWS, Matt Garman is present at SAP Sapphire to explain the collaboration with SAP. This is a broad collaboration in which SAP will make generative AI models from Amazon Bedrock available within the SAP Generative AI Hub. Models from AI21 Labs, Anthropic, Cohere, Meta, Mistral AI, Stability AI and Amazon, among others, will thus be available to SAP customers.
Furthermore, AWS has made available a new instance type, the Amazon EC2 High Memory U7i, with 8 sockets and up to 32TB of RAM on which SAP S/4HANA can run in the AWS Cloud. According to Christian Klein, many customers were waiting for even larger instances.
The cooperation between the two IT companies goes a bit further in terms of chips. The SAP HANA Cloud uses AWS’ Graviton3 chips for many workloads. SAP claims that this reduces its carbon footprint by 45%. SAP will now partner with AWS to make the switch to Graviton4 soon.
SAP is also going to deploy the AWS Tranium and AWS Inferentia chips for AI and machine learning workloads. This will mainly involve future SAP Business AI solutions.
SAP uses Google’s AI for supply chain
Where SAP chooses AWS to fully support Bedrock, with Google, it chooses a more limited integration of Vertex AI, mainly focused on the supply chain. Although we read between the lines that both Joule and SAP Integrated Business Planning for Supply Chain will be integrated with Google Gemini AI Assistant and Google Cloud Cortex’s Data Foundation.
This makes it possible to generate more accurate forecasts and respond faster to signals from the supply chain. Google AI can detect and solve problems in the supply chain much faster. This is due to the integration between Cortex and Big Query. This makes a lot of additional contextual data available, such as the weather forecast and search trends.
Like AWS, Google is also developing additional instances for SAP S/4HANA workloads. The new SAP HANA X4 family comes in three versions with 16, 24 and 32TB of working memory. Up to 1920 vCPUs and Hyperdisk block storage can be used with speeds of up to 10,000 megabits-per-second throughput.
Finally, it will be possible to link SAP Datasphere directly to BigQuery. All data in SAP Datasphere can be brought to BigQuery for additional data analysis. SAP Datasphere is now available in the Google Cloud marketplace.
SAP and Meta Llama open source models.
SAP makes many AI models available through the SAP generative AI hub in SAP AI Core on the SAP Business Technology Platform. These include all models from Amazon Bedrock, as well as the Meta models. Meta Llama 2 and Llama 3 will be available for customers to generate content. In SAP Analytics, SAP will deploy Llama 3 for comprehensive analytics. In many large analytic workloads, Llama 3 will be deployed.
SAP and Nvidia partnership for new AI models
SAP has further expanded its partnership with Nvidia. For example, they also use Nvidia NIM microservices to train and deploy AI models. As just described, SAP has many AI partners for models, but it is also developing smaller models. For generating ABAP code, SAP also uses the NVIDIA Omniverse Cloud. Nvidia’s CEO was a brief virtual presence at the SAP conference.