NVIDIA Accelerates Enterprise Data Processing with cuDF and cuVS

At NVIDIA GTC 2026, NVIDIA announced advancements in accelerating corporate data processing through its accelerated data libraries, cuDF and cuVS, based on NVIDIA CUDA-X. These libraries are designed to reduce costs and improve performance for both structured and unstructured data processing across various industries.

The Growing Require for Accelerated Data Processing

Organizations are generating hundreds of zettabytes of data annually, driving a competitive need to extract valuable insights. NVIDIA cuDF and cuVS aim to address this challenge by providing accelerated data processing capabilities integrated into leading enterprise data platforms.

cuDF for Structured Data

NVIDIA cuDF accelerates the performance of open-source data processing engines like Apache Spark, Presto, DuckDB, Polars, and Velox. According to NVIDIA, cuDF delivers up to 5x faster processing speeds compared to CPU-only environments.

cuVS for Unstructured Data

With unstructured data comprising 80% of enterprise data and continuing to grow, NVIDIA cuVS accelerates key engines such as FAISS, Amazon OpenSearch Service, and Milvus. This enables faster extraction of context, facts, and recommendations from large volumes of text, image, and video data.

Industry Adoption and Results

Several companies are already leveraging NVIDIA cuDF and cuVS to improve their data processing capabilities:

• Google Cloud: Integrating NVIDIA cuDF to accelerate Apache Spark performance within Dataproc, reducing large-scale ETL task processing time from hours to seconds.
• Snap: Reduced daily data processing costs by 76% by adopting NVIDIA cuDF on Google Kubernetes Engine (GKE), enabling analysis of 10 petabytes of data within 3 hours. Snowflake reports that Snap CIO Saral Jain stated this collaboration allows them to innovate faster for Snapchat’s more than one billion users.
• Nestlé: Initial experiments with IBM’s watsonx.data, accelerated with NVIDIA cuDF, delivered 5x faster throughput and 83% lower costs in an order-to-cash mart workload. Nestlé CIO Chris Wright highlighted the potential for faster decision-making in manufacturing and logistics.
• Dell: The NVIDIA and Dell AI Data Platform, equipped with an accelerated data engine, provides up to 3x faster performance and 12x higher vector indexing throughput compared to CPU-based systems. Dell Chairman and CEO Michael Dell stated the platform transforms multimodal data into AI-ready states in hours instead of days.
• Oracle: Oracle Private AI Service Containers can now leverage NVIDIA cuVS to accelerate vector index creation in the Oracle AI Database, enabling faster AI-driven decisions. Oracle CEO Clay Margoyoke noted that this technology allows applications to run that were previously impossible.

Broad Platform Support

NVIDIA cuDF and cuVS are supported on leading enterprise data platforms, including EDB Postgres AI, NetApp, Snowflake, Starburst, and Bust Data, facilitating the future of AI-driven data processing.

About NVIDIA

NVIDIA established itself as a pioneer in accelerated computing with the invention of the GPU in 1999. The company has evolved into a full-stack computing company reshaping industries with data center-scale products and leading the modern era of artificial intelligence. NVIDIA continues to drive industrial digitalization across markets.