Oracle Cloud Infrastructure incorporates Nvidia GPUs and AI
Nvidia GPU chips will support Oracle Cloud Infrastructure. Also, OCI users will have access to Nvidia's Enterprise AI system and Clara healthcare-specialized platform.
Oracle and AI vendor Nvidia expanded their partnership to a multiyear agreement that brings Nvidia's full computing stack to Oracle Cloud Infrastructure.
OCI is adding tens of thousands of Nvidia GPUs, including the A100 and soon-to-be-released H100, the software and hardware giant said on Oct. 18 at its CloudWorld 2022 conference in Las Vegas.
An upcoming release of the Nvidia AI Enterprise platform will also become available on OCI, according to Oracle, which did not provide a date.
Oracle's expanded partnership with Nvidia signals the trend of building applications on cloud infrastructures with AI, Gartner analyst Sid Nag said.
Oracle and Nvidia are not the only vendors doing this. Tech giants Google and Microsoft recently introduced capabilities that also embed AI into cloud applications.
"The more AI functionalities are embedded into the application, that's good," Nag said.
Sid NagAnalyst, Gartner
Adding AI into cloud operations ensures to some extent that the day-to-day running of applications works smoothly, he added. "Any intelligence that's built into the process of running, operating and provisioning orchestration of the cloud data center ... is always helpful."
Oracle and Nvidia's partnership takes a multi-pronged approach, Nag said.
First, Oracle is bringing Nvidia's AI software into its AI strategy and main computing engine. Also, AI models require high-performance hardware, and Nvidia's widely used AI-specialized GPUs will enable customers to rebuild AI models at scale as the models change.
"The faster you can rebuild the models, the better you are," Nag said.
For Oracle, using Nvidia's GPUs will also make workloads run faster.
"Oracle as a database company also does a lot of analytics," said David Menninger, analyst at Ventana Research. "To the extent those analytics are AI and machine learning-based, GPU chips will help those workloads run faster."
Finally, Oracle is using the message passing interface (MPI) to bring AI functionality in real time, Menninger noted. MPI is a standardized and portable API for communicating data between distributed processes.
"AI is great, but if you really don't have the ability to do things in real time, then there's really no value," he said.
Questions and challenges
One possible question about the Nvidia-Oracle partnership is whether the relationship addresses both on-premises and OCI configurations, Menninger said.
"There are workloads where, maybe for governance reasons, they might have to be run on premises," he said.
Another question relates to the ongoing chip shortage and who's responsible for matching supply with demand, Menninger added. "Somebody has to make sure that there are enough chips available, configured and ready to go to process with workloads."
Due to the popularity of Nvidia's AI chips, it's probable that Oracle's customers have been seeking a strengthened partnership like this. "The GPU chips have been in the market long enough now and their value is well recognized. And so I would expect Oracle customers to be asking for this type of relationship, if not demanding it," Menninger said.
Oracle is also bringing Nvidia's Clara system -- a platform for AI applications and frameworks for healthcare developers and researchers -- to OCI.
The vendors said Clara will enable Oracle users to get faster results on X-rays, scans and genomics processing.
"This is this moment we've all been dreaming about, really, for decades since we started to do AI work in the old system days," Oracle CEO Safra Catz said during a streamed conversation with Jensen Huang, Nvidia's co-founder and CEO.
Using AI to make "shopping more efficient and routing trucks better and all those things are great," she said. "But really, one of the most critical areas is saving lives."