AI-based systems will be an integral aspect of future operations for many organisations. However, most public cloud services cannot meet strict data compliance regulations.
It is impossible to know where data entered into ChatGPT is being stored and processed for instance – which means users cannot guarantee it is not being transferred to another legal jurisdiction. And the same is true of virtually any public cloud platform – data may be hosted in a completely different jurisdiction. When this happens, users could be inadvertently breaking local laws regarding data protection and sovereignty.
Oracle’s “sovereign” AI solution
To help address this issue, Oracle’s Enterprise AI has a range of deployment options. The system can be operated across OCI Dedicated Region, Oracle Alloy, Oracle EU Sovereign Cloud, and Oracle Government Cloud. In their own words, this ‘offers customers a state-of-the-art AI solution that provides greater control over operations, location, and security to help support digital sovereignty’.
Regular readers of the WTL blog will already know about Oracle’s localised cloud data centres, specifically designed to help meet data sovereignty regulations. Across 66 cloud regions in 26 countries, customers can access more than 100 cloud and AI services without breaching data protection laws.
Equally importantly, these same AI services can be deployed in local data centres. OCI Dedicated Region allows you to bring a complete OCI cloud region into an on-premises data centre for instance. This allows your organisation to harness the power of the cloud at the same time maintaining ultimate management control.
More than just software
Oracle has also announced a partnership with NVIDIA that will see their new Grace Blackwell GPU technology being deployed across OCI Supercluster and OCI Compute. The GB200 chipset is of particular interest for AI purposes, delivering up to 30x faster real-time large language model (LLM) inference, 25x lower TCO, and requiring 25x less energy compared to the previous generation of GPUs.
NVIDIA DGX Cloud on OCI is also being upgraded with the new GPU. The full DGX Cloud cluster buildout will include more than 20,000 GB200 accelerators and NVIDIA CX8 InfiniBand networking. The cluster will consist of 72 Blackwell GPUs NVL72 and 36 Grace CPUs with fifth-generation NVLink.
According to NVIDIA, the Blackwell GPU will enable organisations to run trillion parameter scale generative AI models in real-time. With 208 billion transistors, this is claimed to be the world’s most powerful chip and has been specifically designed with the assistance of cloud service providers to deliver the compute potential required for next-generation AI learning and modelling.
Available now
Oracle’s Sovereign Cloud solutions are already available and will help you solve AI’s sovereignty problem. This means that your business can get started with fully compliant AI services immediately. Contact the WTL team today to learn more.