3 months ago
96
ARTICLE AD BOX
Oracle's Larry Ellison believes AI's adjacent frontier is backstage endeavor data, not conscionable nationalist internet. He argues existent models are commoditized and existent worth lies successful securely accessing and reasoning implicit delicate institution information. Oracle is investing heavily, leveraging its immense database beingness to go the endeavor AI backbone, facing contention from unreality rivals.
Oracle laminitis and CTO Larry Ellison has identified what helium calls the captious regulation holding backmost AI models from Gemini to ChatGPT: they're each trained connected the aforesaid publically disposable net data, making them progressively commoditized with small existent differentiation betwixt them.Speaking during Oracle's fiscal Q2 2026 net telephone successful December, Ellison argued that portion grooming foundational models connected nationalist information has created "the largest and fastest-growing concern successful quality history," the existent worth lies up successful a 2nd phase: enabling these models to securely crushed implicit backstage endeavor information without compromising security."For these models to scope their highest value, you request to bid them not conscionable connected publically disposable data, but marque privately owned information disposable for those models arsenic well," Ellison said.
He estimates this adjacent signifier volition beryllium "even larger and much valuable" than the existent GPU and information halfway boom.
Oracle bets large connected becoming the endeavor AI backbone
Oracle is positioning itself arsenic uniquely equipped for this shift. The institution points retired that astir of the world's high-value backstage information already lives successful Oracle databases. Its AI Data Platform uses techniques similar Retrieval-Augmented Generation to fto immoderate large AI exemplary query backstage information successful real-time portion maintaining security.
The institution is backing this imaginativeness with assertive spending, projecting astir $50 cardinal successful superior expenditures for the afloat year, up from $35 cardinal estimated successful September. At Oracle AI World successful October, the institution announced partnerships including a 50,000-GPU AI supercluster with AMD MI450 chips launching successful Q3 2026, and the OCI Zettascale10 supercomputer connecting hundreds of thousands of NVIDIA GPUs.
The contention for backstage information entree heats up
Ellison's thesis faces challenges from aggregate angles. Synthetic information procreation could trim dependence connected exclusive proprietary information sets; real-time idiosyncratic enactment information from user apps could crook retired to beryllium much invaluable than static endeavor records. At the aforesaid time, rivals similar Amazon Web Services, Microsoft Azure and Google Cloud are racing to make comparable endeavor AI capabilities, though Oracle argues its established grip connected endeavor databases gives it a strategical advantage.
By precocious 2025, the company's unreality backlog had topped $500 billion, driven mostly by request for AI.
