VAST Data and AMD Expand Collaboration to Advance AI Infrastructure

VAST Data announced an expanded collaboration with AMD on July 27, 2026, to develop high-performance AI infrastructure for cloud providers and enterprises.

The partnership integrates the VAST AI Operating System with 6th Gen AMD EPYC CPUs and AMD Instinct GPUs. Consequently, this integration supports scalable training, inference, and agentic workloads across distributed environments.

Expanding the Hardware Integration

As organizations transition from model training to operationalizing AI agents, the demand for efficient data management increases. Specifically, success depends on how systems manage data, memory, context, and compute resources together. VAST Data addresses this through its Disaggregated Shared Everything architecture, which provides multi-tenancy support and secure workload isolation. The architecture delivers native multi-protocol access and a unified global namespace, which simplifies data access and allows AI clouds to run massive concurrent workloads.

Optimizing AI Infrastructure Performance

The collaboration introduces several technical integrations to improve AI infrastructure efficiency. Notably, VAST selected 6th Gen AMD EPYC processors to power its 6th-generation CBox and 3rd-generation EBox platforms. These processors support PCIe Gen-6, which doubles the input-output bandwidth generationally to improve file and object storage performance.

“AI is entering an operational phase where infrastructure efficiency matters as much as model performance.”

John Mao, Vice President of Global Technology Alliances at VAST Data

Furthermore, early testing of the AMD Instinct MI355X GPU showed a 9X speedup in time-to-first-token. The system also achieved 9.7X more token throughput when using VAST for key-value cache offloading. These results are critical for delivering low-latency and energy-efficient deployments.

Reference Architectures and Ecosystem Growth

In addition, VAST, AMD, and DriveNets developed a reference architecture featuring AMD Helios rack-scale AI infrastructure. This architecture provides guidance for model training, inference, and reinforcement learning workloads. Meanwhile, software innovators like TensorMesh and EmbeddedLLM are collaborating to accelerate production-ready deployments.

“Our expanded collaboration with VAST combines AMD EPYC CPUs and Instinct GPUs with the software foundation customers need to accelerate inference.”

Derek Dicker, Corporate Vice President at AMD

The integration also includes automated key-value cache lifecycle management. This feature uses native data lifecycle policies to automatically delete cached data containing sensitive information. As a result, enterprises can maintain regulatory compliance without manual operational overhead. The system manages these tasks through its core database and event streaming capabilities.

Industry Adoption and Future Outlook

Several AI cloud providers, including Core42, Crusoe, and Vultr, are deploying these joint technologies. These companies require architectures that combine accelerated computing with context management. Ultimately, the unified platform helps organizations transition from pilot projects to production environments.