Oracle, Nvidia partner to add AI software into OCI services

Oracle and Nvidia are partnering to make the latter’s AI Enterprise software stack available via Oracle Cloud Infrastructure (OCI) services.

Nvidia’s AI Enterprise stack will be available natively through the OCI Console and will be available anywhere in OCI’s distributed cloud while providing enterprises access to over 160 AI tools for training and inference, including NIM microservices, the companies said in a joint statement at Nvidia’s annual GTC conference.

The AI-focused software stack will enable enterprises to combine them with OCI services for building applications and managing data across deployments, they added.

Analysts believe that the integration of Nvidia’s AI software stack into OCI will provide developers and enterprises with a variety of advantages.

“Developers stand to be able to seamlessly leverage technologies such as Nvidia NeMo, NIM, and RAPIDS, all of which are part of Nvidia AI Enterprise stack. They can use NeMo to build, train, and fine-tune large language models, or use NIM to facilitate the deployment of AI models as microservices,” said Arnal Dayaratna, research vice president at IDC.

Moor Insights and Strategy principal analyst Jason Andersen believes that the integration will drive a better user experience, which is important for users working in DevOps, as well as provide consistency in terms of support and billing.

Further, Charlie Dai, vice president and principal analyst at Forrester, said that the integration will drive efficiency for developers and enterprises by reducing setup time and simplifying access to advanced AI tools.

As an extension to Oracle’s strategy to help enterprise users deploy faster with minimal setup, the cloud services provider has integrated Nvidia’s NIM microservices into OCI Data Science.

“Data scientists can access preoptimized microservices from Nvidia NIM directly in OCI Data Science to support real-time AI inference use cases without the complexity of managing infrastructure,” Oracle said in a statement.

Explaining how NIM would reduce complexity for enterprise users, Dai said that NIM provides containers to GPU-accelerated inferencing microservices for pre-trained, customized AI models and the integration will aid data by simplifying AI model deployment, enhancing scalability, and accelerating time-to-insight.

NIM’s pre-built container instances and standard APIs are also expected to simplify hosting and accelerate integration with upstream and downstream applications, said Brian Alletto, director of West Monroe’s technology and experience practice.

What will Oracle and Nvidia gain out of the partnership?

The partnership, according to analysts, is a symbiotic one, to say the least.

While making software such as the AI Enterprise stack and NIM available through its services increases customer adoption of Oracle’s core compute, and storage offerings, the partnership provides Nvidia first-class citizenship to a major cloud provider, driving additional usage of GPUs and related software, resulting in more stickiness, Alletto explained.

Additionally, Moor Insights and Strategy’s Andersen pointed out that the integration may boost Oracle’s image as a provider of AI offerings.

“While Oracle has an excellent history with enterprise data, it is not as well known for its AI tools or model capabilities, so this is a way for Oracle to promote a solution that represents the most credible authorities in both data and AI,” Andersen said.

“It’s a good combo versus some of Oracle’s competitors, such as AWS, Microsoft, and Google Cloud, who have built a lot more of their own capabilities,” the analyst added.

On the flip side, Andersen believes that the partnership will make Nvidia’s software stack more credible to other cloud service providers since they have their own tooling and capabilities and don’t need the chipmaker’s software services.

Oracle adds AI Blueprints to OCI

In order to further help its enterprise customers simplify and accelerate their AI deployments, Oracle is adding AI Blueprints — no code deployment recipes — to OCI.

OCI AI Blueprints supports automatic scaling (autoscaling) of inference workloads to handle varying traffic loads efficiently, Oracle said.

“When demand increases, OCI AI Blueprints can spin up more pods (containers running inference jobs) and, if needed, provision additional GPU nodes. When demand decreases, it scales back down to save resources and cost,” it explained. However, West Monroe’s Alletto pointed out that while tools such as AI Blueprints enable rapid deployment and can be optimized later as product use cases, balancing the offering velocity and cost remains a challenge.