Research Finder
Find by Keyword
VAST Data and Cloudera: Authorization Follows the AI Workload
The partnership addresses two requirements for production enterprise AI: the rate at which enterprise context reaches accelerated compute, and the scope of data each workload is authorized to retrieve.
8/15/2026
Key Highlights
- VAST combines a high-performance path to accelerated compute with Cloudera's governed enterprise data services, connecting data delivery and authorization in the same AI architecture.
- Retrieval augmentation and vector search make enterprise data part of runtime AI context, bringing authorization scope into the inference path.
- Cloudera Ranger and the Ranger Authorization Service (RAZ) extend policy enforcement to S3-compatible object storage at bucket, directory, and file granularity, allowing customers to preserve authorization as AI workloads access enterprise data.
- The initial integration supports VAST as an S3-compatible storage option for Cloudera, preserving customer choice and allowing VAST infrastructure to be applied where accelerated workloads justify it.
- The companies indicate that a joint financial services deployment already extends beyond traditional HPC workloads into AI inferencing, providing an early enterprise test of the combined architecture.
The News
Cloudera and VAST Data announced a strategic partnership that combines Cloudera's containerized data services with the VAST AI Operating System to support enterprise AI environments spanning on-premises and public cloud deployments. The architecture uses the NVIDIA AI Data Platform reference design and combines VAST storage, database, and global namespace capabilities with Cloudera data engineering, analytics, governance, machine learning, and AI services. The companies position the architecture for production AI workloads requiring high-performance access to governed enterprise data. For more details, read the official partnership announcement.
Analyst Take
Enterprise AI places new demands on data infrastructure. HyperFRAME Research Lens (1H 2026) finds that only 14% of organizations have data architectures they consider ready for AI, while 62% identify security and governance as leading AI infrastructure priorities. These findings connect two requirements that enterprises often address separately: delivering enterprise data to AI infrastructure at sufficient speed, and controlling which data an AI workload is authorized to use.
We view these requirements as rate and scope. Rate determines how quickly relevant enterprise context reaches accelerated compute. Scope determines which context a user, application, or AI workload is permitted to retrieve. Production AI requires both. Improving rate without preserving scope can simply allow an AI system to access the wrong data faster.
GPU utilization is therefore partly a data architecture problem. Accelerated compute can be constrained when enterprise pipelines cannot deliver the required context at the rate an AI workload requires. VAST addresses this part of the architecture through high-throughput access to persistent enterprise data and support for NVIDIA GPUDirect Storage, which reduces reliance on traditional CPU memory buffering and shortens the data path between storage and GPU memory.
Cloudera's role extends across the data lifecycle, with containerized services preparing and serving the enterprise data that applications consume. This is relevant for Private AI, where organizations bring compute closer to governed enterprise data while sensitive datasets remain in place.
Cloudera addresses scope through its data services and governance architecture. Apache Ranger provides centralized authorization policy. The Ranger Authorization Service (RAZ) extends those policies to S3-compatible object storage at bucket, directory, and file granularity, with policy administration remaining in Ranger where platform engineering teams manage it today. That enforcement point becomes crucial as enterprise data becomes context for AI. An inference workload may require customer transactions or operational records while remaining unauthorized to retrieve employee compensation or sensitive PII. Cloudera documents coarse-grained group-level enforcement as the condition preceding RAZ, which leaves enterprises choosing between broad access to an entire bucket and fragmenting data into separate buckets by sensitivity. Both approaches can carry significant operational costs that enterprises must absorb today.
Granular object authorization operates at the retrieval boundary. RAZ enforces whether a request may retrieve a file or object, and separate controls determine whether the retrieved information is accurate, appropriate for the task, and safe for an agent to use downstream. AI needs context, and access to that context remains governed by identity and policy. Cloudera has carried that position independent of the VAST relationship, which indicates the authorization boundary functions as an architectural commitment.
From Architecture to Adoption
Reference designs standardize the architectural path from persistent storage to accelerated compute while leaving enterprise policy implementation to the surrounding platforms. NVIDIA's new financing platforms, structured to mobilize more than $500 billion in third-party capital, could accelerate adoption of these common AI infrastructure designs. NVIDIA states that investments will be underwritten against demand, utilization, cash flow, and residual value, placing a financial premium on keeping accelerated infrastructure supplied with enterprise context that workloads are authorized to use. As more vendors align with common AI infrastructure designs, differentiation extends beyond the data path into how enterprise context is governed, authorized, and delivered.
The architecture also supports customer optionality. Cloudera consumes S3-compatible object storage through open interfaces, and platform teams retain existing pipelines and table definitions while the storage target changes underneath. VAST becomes an infrastructure option for selected workloads, applied where accelerated compute justifies its requirements and economics, while Cloudera customers keep their broader data estates where they sit today.
The companies describe the result as a unified AI factory. The initial integration delivers a complementary architecture, with customers procuring, operating, and supporting two products from two vendors. Platform engineering teams owning data services and infrastructure organizations procuring accelerated compute often hold separate budgets and buying authority. In our opinion, the strength of the architecture will be measured by whether joint customer deployments convert technical complementarity into repeatable product and go-to-market motions.
What Was Announced
The VAST-Cloudera architecture combines Cloudera's portable, containerized data services with the VAST AI Operating System. Cloudera provides data engineering, streaming, analytics, machine learning, AI, governance, and related services spanning hybrid and multi-cloud environments. VAST provides its Disaggregated Shared Everything architecture and AI Operating System, including high-performance storage, database capabilities, and a global namespace.
VAST integrates vector database services with NVIDIA cuVS for GPU-accelerated vector indexing and search. Cloudera AI Inference Service uses NVIDIA NIM microservices and supports NVIDIA Nemotron open models. The announced architecture supports Apache Spark acceleration through NVIDIA cuDF, allowing Cloudera Data Engineering workloads to use VAST data services with GPU-accelerated processing.
The companies base the data-platform layer on the NVIDIA AI Data Platform reference design. Their stated objective is to continuously ingest, refine, govern, and deliver enterprise data to AI models for training and inference while improving utilization of accelerated compute infrastructure. The architecture targets structured, unstructured, and multimodal datasets and supports deployment spanning data centers, private clouds, and public clouds.
Private and sovereign AI are also part of the announced positioning. The companies combine NVIDIA AI infrastructure and NVIDIA AI Enterprise software with VAST data infrastructure and Cloudera enterprise data and AI services. Cloudera and VAST state that their combined customer environments manage approximately 60 exabytes of data.
The solution is available immediately through the companies' enterprise sales teams and partner ecosystems. Cloudera and VAST said they plan to expand reference architectures, validated deployment patterns, and industry-specific solutions throughout 2026.
Looking Ahead
Customers purchase VAST from VAST and Cloudera from Cloudera, with each company supporting its respective platform. The companies are developing a joint co-sell motion around the combined architecture. There is room for deeper technical integration as customer requirements surface. Semantic retrieval and Cloudera consumption of VAST vector capabilities are two potential areas, although the companies have yet to define a comprehensive roadmap.
Authorization becomes more dynamic as enterprise AI relies on runtime context retrieval, and agentic workloads extend that requirement because retrieval decisions can change based on intermediate results. Whether identity and authorization context persist through retrieval, model reasoning, tool invocation, and downstream agent actions will be an area to watch as the architecture evolves.
HyperFRAME Research has learned that a joint financial services customer is already using the platforms together for workloads extending beyond HPC into AI inferencing. Financial institutions combine large regulated data estates with growing investments in accelerated AI infrastructure, creating simultaneous requirements for performance, authorization, lineage, and auditability.
VAST established its position with AI cloud providers and high-performance infrastructure and is now expanding into large enterprises through AI infrastructure attached to accelerated compute, reporting major global banks as customers in recent quarters. Cloudera holds established relationships with Global 2000 data, platform, and engineering teams, and with regulated market infrastructure operators such as the National Stock Exchange of India. The companies describe increasing convergence between the infrastructure organizations deploying GPUs and the data organizations Cloudera serves. As VAST moves up the stack, it will invariably invite comparison and competition from vendors outside traditional infrastructure. The Cloudera relationship demonstrates one approach to that expansion through partnership, with each company contributing its respective capabilities.
The companies describe performance benefits from the combined architecture but have yet to publish joint benchmarks. We will be watching for those results, along with customer adoption and demand for deeper integration around vectors and semantic context, as important measures of how the partnership develops.
Don Gentile | Analyst-in-Residence -- Storage & Data Resiliency
Don Gentile brings three decades of experience turning complex enterprise technologies into clear, differentiated narratives that drive competitive relevance and market leadership. He has helped shape iconic infrastructure platforms including IBM z16 and z17 mainframes, HPE ProLiant servers, and HPE GreenLake — guiding strategies that connect technology innovation with customer needs and fast-moving market dynamics.
His current focus spans flash storage, storage area networking, hyperconverged infrastructure (HCI), software-defined storage (SDS), hybrid cloud storage, Ceph/open source, cyber resiliency, and emerging models for integrating AI workloads across storage and compute. By applying deep knowledge of infrastructure technologies with proven skills in positioning, content strategy, and thought leadership, Don helps vendors sharpen their story, differentiate their offerings, and achieve stronger competitive standing across business, media, and technical audiences.
Stephanie Walter | Practice Leader - AI Stack
Stephanie Walter is a results-driven technology executive and analyst in residence with over 20 years leading innovation in Cloud, SaaS, Middleware, Data, and AI. She has guided product life cycles from concept to go-to-market in both senior roles at IBM and fractional executive capacities, blending engineering expertise with business strategy and market insights. From software engineering and architecture to executive product management, Stephanie has driven large-scale transformations, developed technical talent, and solved complex challenges across startup, growth-stage, and enterprise environments.