In a move to capitalize on the growing demand for generative AI services, Oracle Corp., renowned for its expertise in database technology, announced a strategic partnership with Toronto-based startup Cohere. The collaboration aims to develop a new cloud service that enables enterprise customers to create and train large language models (LLMs) using their private data while ensuring data privacy and security.
Oracle’s founder and chief technology officer, Larry Ellison, confirmed the partnership during the company’s recent earnings call, where he highlighted Oracle’s substantial growth in the cloud business. The longstanding relationship between Oracle and Cohere, along with Oracle’s participation in Cohere’s recent $270 million Series C funding round, has fueled speculation about this collaboration.
Ellison emphasized the vision behind the partnership, stating, “Cohere and Oracle are working together to make it very, very easy for enterprise customers to train their own specialized large language models while protecting the privacy of their training data. Over the next few years, lots of companies are going to train their own specialized large language models.”
As part of the collaboration, Oracle’s internal application development teams have already started utilizing the Cohere AI cloud service on the Oracle Cloud Infrastructure (OCI). The service leverages Oracle’s private data to fine-tune and expand existing Cohere LLMs. Notably, this supplementary training has resulted in the creation of two new specialized LLMs, catering to medical professionals and first responders, respectively.
With Oracle’s significant presence in the healthcare sector following its acquisition of healthcare giant Cerner in 2022, the company recognizes the potential of specialized large language models to enhance the productivity of highly trained professionals. Ellison emphasized the impact of these models, stating, “Specialized large language models will be instrumental in helping highly trained professionals use their precious time more efficiently.”
The collaboration between Oracle and Cohere signifies Oracle’s commitment to exploring new avenues within the enterprise cloud space and capitalizing on the potential of generative AI services. As the partnership progresses, enterprises can look forward to an easier and more secure way of training their own customized language models, tailored to their specific needs and domains.
Oracle is no stranger the World of AI
While the upcoming service with Cohere is new, Oracle is quite familiar with the world of AI. In fact, Ellison made sure to emphasize during the earnings call that Cohere is using Oracle Cloud for training LLMs.
Ellison said that Oracle has an edge over its competitors because it had more experience and expertise in handling large amounts of data securely and efficiently. Other vendors that have publicly revealed they use Oracle Cloud for training LLMs include Adept AI Labs, which raised $350 million in March for a generative AI service for using software. Oracle also has a cloud AI partnership with Nvidia, that involves Nvidia GPU hardware and Nvidia using the Oracle Cloud to help with ongoing AI development. All told, Ellison boasted that Oracle Cloud is already a multi-billion business for AI workloads.
“In the aggregate, our generative AI cloud customers have recently signed contracts to purchase more than $2 billion of capacity in Oracle’s Gen2 Cloud,” Ellison said.
While the numbers are large and growing, in the cloud business Oracle still trails behind the big three hyper-scalers which all have their own generative AI services. Amazon Web Services (AWS) announced its Bedrock generative AI services in April, Google has a host of its own services and models that were updated at its recent I/O conference, and Microsoft benefits from its tight partnership with OpenAI.