Oracle Injects AI Into Database, Cloud Infrastructure and Applications

In June, Oracle entered the generative AI market, striking a deal to provide large language models (LLMs) to their customers and announcing billions of dollars in cloud computing commitments from startups training the complex systems.

Three months later, the software maker is delivering broad access to the powerful AI technology across their cloud infrastructure, electronic health record (EHR) software, data analysis tools, ERP, and customer service software. Signed business from AI developers has doubled to $4 billion.

“Generative AI showed up about a year ago now, and it changes everything. It’s certainly changing everything at Oracle,” Chairman and Chief Technology Officer Larry Ellison said September 19 during his keynote address at the company’s CloudWorld conference in Las Vegas.

During the conference, Oracle unveiled a generative AI cloud service, artificial intelligence enhancements to their software for healthcare practitioners, and new database capabilities for storing and rapidly retrieving the unstructured and semi-structured data used in AI systems. Oracle also announced customer management software enhancements that let marketers create campaigns or service agents summarize conversations, and a generative AI-powered assistant for business data analysts debuted.

Generative AI systems, which learn from patterns gleaned from the internet and private data sets and then help users arrive at new insights, have set off a flurry of development and funding—and a scramble for the computing power needed to train and run large language models. These models can aid in medical diagnoses, summarize market analysis, analyze conversations, create text and images, and write code. But LLMs need vast amounts of parallel processing power and fast data communication that exceeds the conventional computing capability found in most businesses’ data centers. That’s driving developers to run more of this work in public clouds, where the networks and specialized chips they need are more readily available.

To put generative AI tools in businesses’ hands, Oracle is expanding the use of chipmaker NVIDIA’s most powerful H100 AI processors in their cloud, touting the networking speed and price advantages the advanced chips bring to the Oracle cloud computing service. Customers are responding: AI development companies have signed contracts to purchase more than $4 billion of training capacity on Oracle Cloud Infrastructure (OCI), double the amount at the end of Oracle’s fiscal fourth quarter on May 31.

Oracle offers a unique cloud computing infrastructure that quickly moves data to where it’s needed for processing, then shuttles it rapidly among processors while AI models get trained, Ellison said.

Oracle is also catering to companies’ desire to quickly and affordably fine-tune standard models by adding their own data to improve predictive capabilities in specific fields. “They want to use their own training data,” he said. “The best way to do that is to put that supplemental training data into an Oracle vector database.”

“We are much faster and many times less expensive than the other clouds for training AI models,” Ellison said.

Businesses that turn generative AI to their advantage could reap big benefits. Generative AI could lift global GDP by 7%—or nearly $7 trillion—over the next decade by increasing office workers’ productivity, hastening drug discovery, and speeding software development, according to Goldman Sachs. Consultancy McKinsey & Co. estimates generative AI and related technologies could add up to $4.4 trillion to global GDP as half of today’s work in areas including sales, marketing, coding, and R&D becomes automated between 2030 and 2060.

Assistance and analysis

Oracle is delivering AI-powered tools on several fronts. In June, Oracle teamed up with Toronto, Canada-based large language model developer Cohere to offer OCI customers AI capabilities based on Cohere’s technology and took an investment in the startup. The first offerings based on the collaboration are coming to market now.

• AI Vector Search adds to Oracle Database 23c the ability to store semantic information about documents, images, and other unstructured files as vectors, or numerical representations of text that let large language models quickly find closely related terms. The collection of features also lets Oracle Database users perform retrieval-augmented generation (RAG), a technique that combines LLMs with private business data to deliver responses in specific fields of knowledge. AI Vector Search can also boost developers’ productivity by adding generative AI capabilities to Oracle Database development tools including APEX, so users can write applications and queries without coding.
• OCI Generative AI, which lets businesses integrate LLMs built by Cohere with their applications via an API, then run those models on Oracle’s fast and cost-effective OCI Supercluster service. The service will be available on a test basis to customers; once it’s generally available it is expected to work in tandem with AI Vector Search to let customers create RAG applications that combine pretrained models with their own data. The generative AI service also underlies new features in Oracle Cloud Applications, NetSuite ERP, and healthcare applications.
• More AI computing power: Oracle plans to make NVIDIA’s latest H100 chips more widely available on their cloud infrastructure; the GPUs offer up to 30 times better inference performance when serving up answers from users’ prompts and four times better training performance compared with NVIDIA’s previous-generation A100 chips. For customers training AI models, OCI Supercluster lets them connect tens of thousands of H100 chips over a high performance, low-latency network, initially available in Oracle’s Chicago, Illinois, and London, England, data centers. OCI also plans to offer NVIDIA’s L40S GPU for training smaller models or for AI inferencing.
• Oracle Clinical Digital Assistant, which works with the software company’s electronic health records, can cut down on paperwork by offering doctors a voice- and screen-powered user interface that automatically takes notes during conversations with patients and suggests actions, such as ordering medicines or lab tests and scheduling follow-up visits. Doctors will also be able to verbally call up MRI results or other elements from a patient’s EHR without typing. The software is expected to be available within the next year.
• Oracle Advertising and Customer Experience, aimed at marketers and service agents, adds the ability to draft AI-generated customer responses, write and search help articles, and summarize key information from customer interactions. Marketing staff can draft campaigns that adhere to their companies’ brand guidelines by walking through a step-by-step AI-driven process.
• Oracle Analytics Cloud, a platform for visualizing data and collaborating on it, now lets analysts interact with data using plain language, seeing and hearing responses generated by AI. The software is also being designed to automatically read and pull key information from text and image files.