Oracle Corporation

04/09/2025 | Press release | Distributed by Public on 04/09/2025 14:58

Global Industry Analyst Views on GPU Expansion for Oracle Compute ...

Global Industry Analyst Views on GPU Expansion for Oracle Compute Cloud@Customer and Private Cloud Appliance
April 9, 2025 | 6 minute read

Michael Brown
Senior Director, Product Marketing, Oracle
Oracle Compute Cloud@Customer and Oracle Private Cloud Appliance let you run AI inferencing, high performance computing (HPC), and visual processing on OCI compute, GPU, storage, and networking services wherever you need. These modular, scalable systems deliver exceptional performance with up to 48 NVIDIA L40S GPUs, help address data residency concerns by giving you control over where data is stored and processed, and provide low-latency connectivity to on-site data sources such as Oracle Exadata Cloud@Customer and Internet of Things (IoT) devices. With cloud automation and cost-effective pricing, you'll find that these new Oracle distributed cloud capabilities make it easy and affordable to run GPU based applications anywhere.

What people are saying
Leading industry analysts had the following remarks on the availability of GPUs in Compute Cloud@Customer and Private Cloud Appliance:

Marc Staimer, Senior Analyst and Contributor, theCUBE Research
"As part of Oracle's distributed cloud strategy, Compute Cloud@Customer with NVIDIA GPUs delivers OCI Compute on-premises and takes the guesswork out of building an AI platform. Unlike competitive offerings disguised as mere boxes with financing contracts and no cloud supporting them whatsoever, Compute Cloud@Customer uses the same APIs, the same software and same control plane as OCI for an authentic cloud experience. This is a timely solution from Oracle that enables customers to start small and get into AI without having to commit to a massive upfront investment-plus, it's a perfect complement to Exadata Cloud@Customer deployments for Oracle Database 23ai."

Mike Leone, Practice Director for Data Management, Analytics & AI, Enterprise Strategy Group
"Oracle Compute Cloud@Customer brings the power of scalable AI and consumption-based cloud economics to organizations that can't move their data to the cloud. It's integrated architecture with scalable compute, storage, and up to 48 NVIDIA L40S GPUs makes it easy for users to run LLMs for inferencing and other AI use cases anywhere."

Alexei Balaganski, Lead Analyst & CTO, KuppingerCole Analysts
"In the race to adopt artificial intelligence for competitive advantage, businesses must balance innovation with security and compliance. However, many face challenges related to data sovereignty and residency requirements, raising concerns about where and how their data is processed. Oracle's Compute Cloud@Customer addresses these issues by offering an on-premises cloud solution that ensures data remains securely within a customer's data center. It can even function in a disconnected mode for greater sovereignty control. Equipped with NVIDIA L40S GPUs, it delivers the performance needed for AI, graphics, and video processing across industries. With its security, modularity, and consumption-based pricing, the platform provides a compelling foundation for enterprise AI adoption."

Ron Westfall, Research Director, Futurum Group
"Coming directly after the launch of the breakthrough Exadata X11M platform for Oracle Database 23ai workloads, Oracle now unveils Compute Cloud@Customer with NVIDIA GPUs for the application tier-to help organizations thrive with demanding workloads across AI, HPC, 3D graphics and 4K streaming. Another great use case for organizations is fraud detection, where the combination of Oracle's platforms and NVIDIA's GPUs work together to find anomalies and detect fraud in real-time-while helping organizations address data residency requirements."

Holger Mueller, Vice President and Principal Analyst, Constellation Research
"Oracle continues to deliver on its 'bring AI to your data' mantra starting with Oracle Database 23ai AI Vector Search, followed by Oracle Exadata X11M, the database platform in all of its variants, with impressive performance across AI functions. Having completed the data side of the equation, Oracle now announces an economically priced compute engine to run AI models on-premises-Compute Cloud@Customer with NVIDIA L40S GPUs-a high-performance, modular solution with a cloud consumption model. This trifecta provides a unified architecture that supports end-to-end workflows for many of the most popular AI use cases, including generative AI inferencing, fine-tuning AI models, RAG and others. CIOs, including those running Oracle environments, those with data residency requirements, or those taking their first steps into AI, should take a serious look at adding Compute Cloud@Customer to their portfolio."

Carl Olofson, Principal Analyst, DBMSGuru
"Most data is still on-premises and much of that resides on Oracle Database. Many of these customers want to get the benefits of AI, but some don't want to ship their data to an AI cloud service. Compute Cloud@Customer with NVIDIA GPUs provides a compelling solution. It enables customers to leverage a high-performance AI platform for running GenAI inference or LLM/SLM fine tuning on-premises, where their data is. never leaving their data centers, for maximum security control. Further, when combined with Exadata Cloud@Customer, the platform co-engineered with Oracle Database for extreme vector processing, customers can run their GenAI models on Compute Cloud@Customer and use RAG techniques on their internal data to get more relevant answers to their prompts. This combination delivers a powerful value prop for on-premises AI workflows."

Steven Dickens, CEO and Principal Analyst at HyperFRAME Research
"Oracle's Compute Cloud@Customer with NVIDIA GPU expansion offers a compelling solution for organizations needing powerful and scalable AI capabilities on-premises. The ability to run demanding workloads like Generative AI and LLMs while addressing data sovereignty concerns makes this a game-changer for industries like financial services, healthcare, and telecom. Plus, it's the ideal combination with Exadata Cloud@Customer, by addressing the AI application layer while Exadata manages the AI data layer with Oracle Database 23ai AI Vector Search."

Richard Winter, CEO, WinterCorp
"Getting started with AI can be a major challenge for organizations. Oracle Compute Cloud@Customer with NVIDIA L40S GPUs eliminates three key problems for the customer. First, it removes the barrier to entry with a modular architecture that starts small and can grow as needed-and still delivers the performance needed to run substantial AI models. Second, it follows a cloud subscription model, so customers get charged only for what they use. Third, security and compliance risks are greatly reduced, because Compute Cloud@Customer is deployed and runs on-premises, so that data never needs to leave the customer's data center."

Steve McDowell, Principal Analyst and Founder, NAND Research
"For manufacturing organizations looking to deploy digital replicas of their production lines, Oracle Compute Cloud@Customer with NVIDIA GPUs is the perfect on-premises solution. Organizations can reduce costs and time-to-market, improve quality control and improve energy efficiency and sustainability-all without impacting current manufacturing operations. And with a pay-as-you-go model that brings the benefits of OCI distributed cloud to on-premises environments, Oracle Compute Cloud@Customer addresses the needs of organizations who are required to store and manage their data in specific locations-and sets them up for a fast start in AI-driven workloads."

Matt Kimball, Vice President and Principal Datacenter Analyst, Moor Insights & Strategy
"AI Vector Search and retrieval-augmented generation (RAG) are two techniques enterprise organizations employ to more efficiently extract value from the enterprise data that resides in databases around an organization. However, operationalizing AI Vector Search and RAG-and continuous tuning-for absolute best performance and efficiency is a time-consuming and inexact science for many IT organizations. The combination of Oracle Database 23ai on Exadata X11M with Compute Cloud@Customer with NVIDIA L40S GPUs strikes that optimal performance-efficiency balance, making deriving value from AI more frictionless and more accurate. Further, consuming this through Oracle Compute Cloud@Customer delivers this capability in a more cost-efficient manner."

Tony Baer, Principal, dbInsight
"Oracle's extension of its distributed cloud computing line adding a smaller, lower cost option for NVIDIA racks is a huge step in clearing the way for generative AI workloads to come on-premises. Until now, other than Oracle's Dedicated Region and Alloy deployment options, practically the only alternative for running workloads with language models required going to a public cloud hyperscaler. But for many reasons, that could range from internal policies to data residency requirements, a significant proportion of the enterprise is likely to stay on-premises over the foreseeable future. Oracle Compute Cloud@Customer and the new Private Cloud Appliance with NVIDIA GPUs will bring gen AI workloads into the data center. While the NVIDIA L40S processors are smaller in size compared to the instances that run in OCI Superclusters, they may be right-sized for the types of domain-specific enterprise workloads that will run on the more compact models, which we expect to become commonplace in the future."