Skip to content
NVIDIA Launches cuObject for Accelerated AI Storage Access

NVIDIA Launches cuObject for Accelerated AI Storage Access

First seen 2 Oct 2026, 00:06 UTC • •
API

Article Content

Browse articles
ThreatCluster AI
ThreatCluster •October 2, 2026 at 15:08 UTC
  • •NVIDIA cuObject libraries enable direct GPU-to-storage data transfers via RDMA.
  • •The SCADA Server SDK supports GPU-initiated requests, enhancing interoperability.
  • •Major cloud providers are joining NVIDIA's initiative for standardized storage access.

On September 30, 2026, NVIDIA announced the general availability of its cuObject client and server libraries, designed to enhance AI storage access by enabling direct data transfers between GPU memory and S3-compatible object storage via RDMA. This development aims to address the limitations of traditional file systems and improve interoperability among storage solutions. The cuObject libraries provide standardized APIs and eliminate the need for CPU intervention in data transfers, which can significantly enhance performance for AI workloads. The introduction of the SCADA Server SDK allows storage providers to respond to GPU-initiated requests, with IBM already demonstrating a prototype. While the technology is now available, NVIDIA's broader interoperability plans through the xio-sig consortium remain incomplete. The initiative aims to unify storage access protocols across various providers, with major companies like Google Cloud and Microsoft involved. This release is particularly relevant for organizations managing large datasets in object storage, as it promises to streamline data access and improve efficiency.

Start a free Starter trial for enhanced analysis

Ask AI about this cluster

Updated 10d ago How this analysis works

Timeline

2026-09-30
NVIDIA cuObject libraries launched
NVIDIA announced the general availability of cuObject client and server libraries for accelerated AI storage access.
Developer.Nvidia
2026-09-30
SCADA Server SDK introduced
NVIDIA introduced the SCADA Server SDK to enable storage providers to handle GPU-initiated requests.
Developer.Nvidia
2026-10-01
General availability confirmed
NVIDIA confirmed the general availability of cuObject client and server libraries, targeting improved AI infrastructure.
Remio.Ai

More articles in this cluster (5)

Extracted Entities (4)

1 / 2

Common questions

What is cuObject?
cuObject is a suite of libraries from NVIDIA that facilitates direct data transfers between GPUs and S3-compatible object storage using RDMA.
How does cuObject improve AI workloads?
By enabling zero-copy data transfers, cuObject reduces latency and increases throughput for AI training and inference.
Which companies are involved in the interoperability initiative?
Major companies like Google Cloud and Microsoft are participating in the xio-sig consortium to enhance storage access standards.