Cloudera partners with NVIDIA to introduce Apache Spark GPU acceleration

This page was machine-translated and may differ from the original. View original

Cloudera partners with NVIDIA to introduce Apache Spark GPU acceleration

Published2026.08.20 11:39

Up to 4x acceleration compared to CPU without code modification, supporting public, private, and sovereign clouds

Enterprise data engineering capabilities have been launched that accelerate existing Spark workloads without code changes while reducing cloud infrastructure costs. With the expansion of AI adoption, accelerating data preparation speed has emerged as a critical task, and market response is focused on enhancing processing performance while maintaining consistent governance across hybrid environments.

Cloudera announced on the 20th that it is introducing native GPU acceleration capabilities for Apache Spark 4.1 based on NVIDIA's CUDA-X library 'cuDF' into Cloudera Data Engineering.

The company explained that up to 4x workload acceleration on NVIDIA GPUs compared to existing CPU infrastructure is possible without modifying PySpark and SQL code, and cost savings in cloud infrastructure expenses are expected through reduced computing execution time.

This feature is being provided starting today as part of Cloudera Anywhere Cloud™, which was announced at EVOLVE26 Singapore.

Enterprise Spark acceleration capabilities in detail

The key features of the GPU acceleration capabilities being introduced include △zero-code GPU acceleration for Apache Spark 4.1 workloads △enhanced ETL and data preparation speed △embedded deployment requiring no manual driver configuration △enterprise-grade security and governance through Cloudera Unified Data Fabric.

It is reported to deliver consistent performance across public, private, sovereign clouds and on-premises environments.

Unlike existing GPU acceleration solutions limited to a single cloud provider, Cloudera's approach expands functionality across hybrid environments while maintaining consistent governance and operational environments.

According to Cloudera's 'The Great AI Re-Architecture' report, 84% of respondents stated that infrastructure costs have increased due to AI workloads.

Statements from executives of both companies

Leo Brunnick, Chief Product Officer at Cloudera, stated, "For many enterprises, the bottleneck in AI is not the model itself, but how quickly raw data can be transformed into trustworthy and actionable insights. With Spark acceleration in Cloudera Data Engineering, we can eliminate this bottleneck."

Pat Lee, Vice President of Strategic Enterprise Partnerships at NVIDIA, said, "With the native integration of the CUDA-X library into Cloudera Data Engineering, it is now possible to reduce costs and increase speed in Spark pipelines without modifying a single line of PySpark or SQL code."

#AI 인프라 #GPU 가속 #NVIDIA #Cloudera #Apache Spark #CUDA-X #cuDF #데이터 엔지니어링 #ETL

본 기사에 대한 정정·반론·추후보도 청구는 보도 청구 안내를, 그간 게재된 보도문은 정정·반론보도 모아보기를 참고해 주세요.

명세환 기자

All articles by this reporter

관련뉴스

“AI Agents, Private AI, Key to Corporate AI Transformation”

Cloudera, a hybrid platform company for data, analytics, and AI, held "Partner Day 2025 Korea" at Seoul Oakwood Premier Coex, targeting domestic partners and customers, and announced key strategies for this year and enterprise AI market trends.

2025.04.18by 배종인 기자

Cloudera Unveils NVIDIA-Powered On-Premises AI Inference Capabilities

Cloudera announced that it is providing 'Cloudera AI Inference' powered by NVIDIA technology in on-premises environments. Enterprises can deploy and scale large language models (LLMs) and models for anomaly detection, computer vision, and speech AI in their data centers. The 'Cloudera Data Warehouse' based on Trino also expands data center support to centralize security, governance, and visibility while enhancing data accessibility. Data visualization includes feature updates such as AI annotation, query logging and tracing, and simplified administrative configuration.

2026.02.10by 명세환 기자

Cloudera Extends Platform Support Until 2032

Cloudera announced that it will extend support for its hybrid data and AI platform until 2032. The company explained that it has strengthened cloud scalability, Apache Iceberg table optimization, and data sharing capabilities with external platforms while reducing the burden of frequent upgrades and rebuilds. Gartner projects that global AI spending will reach $2.5278 trillion in 2026 and $3.3367 trillion in 2027. This announcement aligns with a trend where enterprises are beginning to weigh operational stability and data utilization efficiency alongside infrastructure replacement during a phase of expanded AI investment.

2026.04.09by 명세환 기자

Cloudera Named Leader in 'Data Lakehouse' Category of Forrester Wave

클라우데라가 포레스터의 2026년 3분기 데이터 레이크하우스 부문 평가에서 리더로 선정됐다. 하이브리드·멀티클라우드 환경에서 에이전틱 AI 워크플로우, 데이터 거버넌스, 실시간 분석을 지원하는 역량이 인정받았으며, ‘비전 및 로드맵’ 항목에서 최고 점수를 획득했다. 분산된 엔터프라이즈 데이터를 단일 플랫폼에서 통합 관리하며 신뢰할 수 있는 AI 확장을 가능하게 한다.

2026.08.03by 명세환 기자

The Problem with AI is Not Performance, but Trust

CEO Kang Min-seung emphasized at Snowflake's World Tour Seoul that generative AI has entered a phase where "trust" rather than accuracy becomes the key competitive advantage. As AI is applied to various tasks including consultation, risk management, and operational automation, incorrect data, erroneous responses, and bias are creating actual business risks. He pointed out that enterprises must build a trust layer above AI. To achieve this, accurate data management, semantic structuring, governance, and prompt security are essential. He explained that AI is not merely an automation tool but a technology that reorganizes the entire organization's workflows and decision-making systems. Snowflake provides a platform that supports this trust-based AI operation, emphasizing that enterprises should build "trustworthy AI" rather than "intelligent AI."

2026.08.28by 배종인 기자