Cloudera and VAST Data have announced a strategic partnership to deliver a joint AI factory. The solution combines Cloudera’s containerized data services with VAST’s AI Operating System and is designed to provide GPUs with workloads so they are fully utilized. The companies are targeting both on-premises and public cloud environments.
Large companies are investing billions in GPUs, but many of these expensive chips sit idle much of the time. They are waiting for data that doesn’t arrive on time due to bottlenecks elsewhere. This phenomenon, known as GPU starvation, is what Cloudera and VAST Data are targeting with their collaboration.
The idea is an “AI factory,” a frequently mentioned concept explained here as a production environment where data is continuously ingested, enriched, managed, and delivered to AI models for training and inference. The solution runs on Nvidia’s AI Data Platform architecture and is available in all types of cloud environments where GPUs are used.
Two layers, one stack
The division of roles is clear. VAST provides the foundation with its Disaggregated Shared Everything architecture, enabling exabyte-scale storage and vector queries via NVIDIA cuVS. Cloudera then layers data engineering, analytics, governance, and AI services on top of that. Together, they aim to solve the classic problem that traditional architectures were never designed for continuously running AI pipelines.
This is becoming even more important now that organizations have been startled by the high costs of AI and are demanding further optimization in this area. For parties offering open-source models, it will be crucial to respond to this by reducing their own costs and charging competitive token fees.
Familiar territory for both parties
VAST has long positioned itself explicitly as a builder of an AI Operating System and is leveraging that market position to develop a complete full-stack AI infrastructure. Meanwhile, Cloudera is firmly committed to private AI. According to the companies, the two combine 60 exabytes of customer data. The Cloudera AI Inference Service, accelerated by NVIDIA NIM microservices, allows organizations to deploy any model, including the latest NVIDIA Nemotron open models.
The joint solution is available immediately. Over time, new architectures will emerge, including industry-specific solutions.