Remote-First-Company | NEW YORK CITY & SAN FRANCISCO, CA. – July 23, 2026 – VAST Data, the AI Operating System company, today announced an expanded collaboration with AMD to help AI cloud providers and enterprises build and operate high-performance AI factories at scale. The collaboration brings together the VAST AI Operating System with 6th Gen AMD EPYC™ CPUs and AMD Instinct™ GPUs to support scalable training, inference and agentic AI workloads with improved efficiency, flexibility and performance.
As organizations move beyond model training and begin operationalizing AI agents, reasoning systems and large-scale inference services, infrastructure requirements are rapidly changing. Success is increasingly determined not only by compute performance, but by how efficiently organizations can manage data, memory, context and compute resources as a unified system.
Crucial to this unified approach is VAST’s inherently versatile Disaggregated Shared Everything (DASE) architecture, which extends beyond standard storage to deliver robust multi-tenancy support for secure workload isolation, native multi-protocol access, and a unified global namespace that simplifies data access across distributed environments. By streamlining critical data operations like rapid model loading alongside VAST's core data platform values, the system allows AI clouds to run massive, concurrent workloads while maximizing hardware utilization.
Many of today’s AI environments were designed for training-centric workflows yet struggle to support the expanding persistent context of retrieval augmented generation (RAG) pipelines and long-thinking, multi-turn inferencing. As a result, many organizations face rising infrastructure and token costs, underutilized GPUs and increasing operational complexity as they scale AI into real-world applications.
To address these challenges, VAST and AMD are expanding their collaboration to deliver an open and flexible approach to AI infrastructure that combines accelerated computing, intelligent data services and optimized inference software into a unified platform for AI clouds and enterprise AI deployments.
The collaboration includes:
VAST has selected 6th Gen AMD EPYC processors, formerly codenamed “Venice,” to power the 6th-generation of CBox and 3rd-generation of EBox platforms to underpin the VAST AI OS. 6th Gen EPYC CPUs bring support for PCIe® Gen-6 that enables 2X the I/O bandwidth generationally for improved file and object storage performance, and helps improves CPU core performance by lowering latency for critical AI data services including database, data warehouse, and event streaming via VAST’s DataBase and DataEngine capabilities.
An AI Infrastructure Reference Architecture developed by VAST, AMD and DriveNets that features AMD Helios rack-scale AI infrastructure coupled with the VAST AI OS and DriveNets AI Fabric networking. These reference architectures document infrastructure support for model training, inference, reinforcement learning (RL) and KV cache workloads with sizing considerations and guidance for AI cloud providers and enterprises to simplify and accelerate high performance, highly available and efficient AI factory deployments.
Expanded ecosystem collaboration with software innovators including TensorMesh and EmbeddedLLM to accelerate deployment of production-ready inference architectures optimized for agentic AI applications.
New KV cache and inference optimizations that combine AMD Instinct GPUs, AMD Infinity Context, and AMD ROCm™ software with the VAST AI OS.
Early VAST testing utilizing an AMD Instinct MI355X GPU demonstrated 9X speedup in time-to-first-token (TTFT), and 9.7X more token throughput utilizing VAST for KV Cache offloading with high concurrency agentic AI workloads. *
These accelerated and high concurrency results are critical to deliver low latency, lower cost and energy efficient inference deployments.
Note on baselines: Because speedups are always relative to the underlying hardware, running these workloads on a GPU baseline with lower compute capacity may yield an even higher relative speedup compared to the latency of reading directly from storage.
Automated KV Cache Lifecycle Management: Purging cached data that contains sensitive or personal information is a critical enterprise compliance challenge. The integration leverages VAST’s native data lifecycle policies to automatically expire and delete KV cache data, helping ensure robust security, privacy, and regulatory compliance capabilities without manual operational overhead.
The AMD Pensando™ Pollara 400 AI NIC provides the high-performance data path connecting AMD Instinct GPUs to the VAST AI OS. Using NFS over TCP and NFS over RDMA, this integration efficiently moves data from GPU memory to the NVMe SSD-based VAST storage cluster, enabling the KV-cache and storage access that large-scale inference and agentic AI workloads depend on.
Proven deployments across leading AI cloud providers delivering AMD technology-powered AI services to customers around the world.
AI is entering an operational phase where infrastructure efficiency matters as much as model performance,” said John Mao, Vice President, Global Technology Alliances at VAST Data. “The industry is discovering that inference is fundamentally a data problem. Success depends on how effectively organizations can bring data, compute, memory and intelligence together as a single system. The VAST AI Operating System was built for this transition, giving AI cloud providers and enterprises a more efficient, scalable and open foundation for training, inference and the next generation of agentic AI applications.
The future of AI will be built on an open ecosystem that gives organizations the flexibility to choose the technologies that best meet their performance, operational and business requirements,” said Derek Dicker, Corporate Vice President, Enterprise Business Group at AMD. “Our expanded collaboration with VAST combines AMD EPYC CPUs and Instinct GPUs with the software foundation customers need to accelerate inference, improve infrastructure efficiency and deploy AI at scale. Together, we’re enabling AI clouds and enterprises to build high-performance AI factories without compromise.
The VAST AI Operating System serves as the data and execution layer for these environments, enabling organizations to unify storage, database, streaming and AI services within a single software platform. By treating model management as a core data service, VAST virtually eliminates cold-start bottlenecks with near-instantaneous model loading that maximizes GPU utilization. Combined with AMD accelerated computing, customers can deploy scalable infrastructure capable of supporting training, inference, retrieval-augmented generation and agentic AI workloads across distributed environments.
Growing Momentum Across the AI Ecosystem
Leading AI clouds are embracing VAST because AI infrastructure is evolving beyond GPU rental and training workloads. As providers build platforms for large-scale inference and agentic AI, they need architectures that combine accelerated computing, data infrastructure and context management into a unified system – delivering performance, flexibility and operational simplicity at scale.
“As AI moves into production, customers need infrastructure that can securely and reliably support increasingly complex workloads without sacrificing performance,” said Raghu Chakravarthi, EVP of Engineering and General Manager – Americas at Core42. “Our infrastructure strategy is built on a silicon agnostic approach to give customers the best optionality and output for their use cases. The combination of AMD accelerated computing and the VAST AI Operating System provides a powerful foundation for delivering enterprise-grade AI services across sovereign and commercial environments with the scale, efficiency and operational simplicity our customers expect.”
“Every frontier lab, every AI-native company building the future of AI needs infrastructure that scales as fast as their ambitions," said Erwan Menard, SVP Product Management at Crusoe. "Our collaboration with AMD and VAST gives Crusoe Cloud customers a validated foundation purpose-built for AI -- combining accelerated compute performance with the data and storage scalability that training and inference workloads demand, so they can spend less time integrating infrastructure and more time building their future.”
“Organizations want the freedom to build AI on the infrastructure that best fits their needs,” said Kevin Cochrane, CMO at Vultr. “The collaboration between AMD and VAST gives customers more flexibility in how and where they deploy AI, as well as the scale and performance required for enterprise workloads. That’s exactly the kind of flexibility customers are looking for as they scale AI across diverse environments.”
“Many organizations understand the potential of AI but are still working through the operational realities of deploying it at scale,” said David Bitton, Vice President, AI and Product Strategy at 5C. “Validated architectures developed by AMD and VAST help reduce complexity, accelerate deployment and give customers greater confidence as they move from pilots to production AI environments.”
“The AI market is increasingly looking for high-performance alternatives that provide both scalability and flexibility,” said Piotr Tomasik, President & COO and Co-Founder at TensorWave. “As an AMD-centric AI cloud, we’ve seen firsthand how rapidly the ecosystem has evolved. Collaborations like this help accelerate adoption by giving customers a validated architecture that combines modern AI infrastructure software with AMD accelerated computing.”
“Agentic AI moves the inference bottleneck beyond raw compute to context: long-running, multi-turn agents make cache, state, and data movement part of the performance path itself,” said Pin Siang Tan, CTO, Embedded LLM. “Embedded LLM works with AI clouds and enterprises to turn AMD accelerator capacity into production AI services, and we’re collaborating with VAST to make its data platform the persistent foundation of that work, from KV-cache reuse today to agent state, replay, and reinforcement-learning data ahead.”
“One of the biggest challenges in large-scale AI is efficiently managing large and concurrent context windows as agents retrieve, reason over, and generate information across massive datasets,” said Dhanaseker Kandhasamy, Co-Founder at Phanos.AI. “At Phanos.AI, we leverage VAST and AMD capabilities including purpose-built storage, compute, network, and software-stack observability to drive real-time decision-making on workload placement for optimal performance. The innovations AMD and VAST are bringing to inference infrastructure help address a critical bottleneck for organizations deploying agentic AI at scale.”
“Customers increasingly want the flexibility to build AI infrastructure using the technologies that best meet their requirements,” said Yossi Kikozashvili, Vice President, Head of Product & G2M, AI Infra at DriveNets. “The collaboration between DriveNets, AMD and VAST demonstrates how an open ecosystem can bring together best-in-class networking, accelerated computing and intelligent data infrastructure to support the next generation of AI factories.”
VAST will showcase the joint architecture, reference designs and inference optimization technologies at AMD Advancing AI 2026 in San Francisco on July 22-23.
Additional Resources:
Blog: Beyond HBM Limits: Accelerating Inference with KV Cache, AMD Instinct, and VAST Data
Partnership Overview: VAST + AMD
*All performance and/or cost savings claims are provided by VAST and have been reviewed, but not independently verified by AMD. Performance and cost benefits are impacted by a variety of variables. Results herein are specific to VAST and may not be typical.