analiza aiqbench
This deal is about AI infrastructure capacity: scaling Oracle's GPU clusters to multi-gigawatt levels on NVIDIA Vera Rubin NVL72 systems. The most useful detail for AI readers is the split between GPU-heavy training and inference and CPU-heavy work that agents do between model calls, with the two scaled independently. Whether that independent scaling reduces idle capacity in practice will depend on real utilization data, which the announcement does not provide.
NVIDIA and Oracle Announce Expanded Partnership
De la 36Kr
RezumatNVIDIA and Oracle announced plans to expand Oracle Cloud Infrastructure (OCI) Supercluster to multi-gigawatt scale based on NVIDIA Vera Rubin NVL72 systems, to meet growing demand for AI training, inference and agentic applications. The NVIDIA Vera Rubin NVL72 systems and NVIDIA Vera CPUs are now fully available on OCI. The core of the expanded collaboration is providing customers with infrastructure that supports two types of workloads: GPU-intensive training and inference for frontier models, and CPU-intensive tasks that agents perform between model calls. Customers can scale CPU and GPU resources independently, precisely matching compute configurations to their workloads, thereby reducing idle capacity and improving infrastructure utilization. (Caixin)
Subiecte conexe

Uber is building an autonomous vehicle empire, and here’s every company it’s using to do it
TechCrunch · 1 aug. 2026

Pomerantz Law Firm Announces the Filing of a Class Action Against Datavault AI Inc. and Certain Officers
prnewswire.com · 24 sept. 2026

Pomerantz Law Firm Announces the Filing of a Class Action Against Datavault AI Inc. and Certain Officers
prnewswire.com · 7 aug. 2026

Anthropic opens Seoul office and announces new partnerships across the Korean AI ecosystem
Anthropic · 17 iun. 2026