VAST Data expands to AMD silicon to optimize AI video streaming

The streaming industry is shifting from model training to large-scale inference where data throughput determines efficiency. By diversifying from an exclusive Nvidia-centric stack to include AMD's EPYC and Instinct hardware, VAST is addressing the 'data starvation' problem that often bottlenecks real-time video intelligence and agentic workflows. For streaming providers, this offers an alternative to the supply-constrained Nvidia ecosystem, providing a high-performance path for deploying AI services that require low-latency access to massive datasets. Watch for the deployment of these 'AI factories' across cloud providers like Vultr and Core42 to see if compute-to-dollar savings drive a broader exodus from proprietary hardware stacks.

The VAST and AMD collaboration coincides with the 'Advancing AI 2026' summit, where AMD unveiled its broader roadmap to challenge Nvidia's dominance. Per GF Securities and Reuters (July 2026), AMD is positioning its MI455X accelerators and 'Venice' CPUs to compete directly with Nvidia’s 'Rubin' architecture. Analysts report that the industry is pivoting toward an 'AI factory' model, where the optimization of silicon, networking, and software stacks is now more critical than raw FLOPS. AMD recently projected that the market for AI accelerators will reach $1.4 trillion by 2030, with inference workloads increasingly dominating the total addressable market.

Related developments include Microsoft’s announcement (July 2026) that it will deploy the AMD Helios rack-scale solution on Azure to power frontier model inference, while Anthropic was named a major customer for upcoming gigawatt-level deployments. Concurrently, VAST Data has strengthened its position as a central platform for these builds; on July 14, 2026, VAST announced a similar strategic partnership with Cloudera to eliminate GPU starvation in hybrid cloud environments. These maneuvers signal a maturing market where storage vendors like VAST are no longer just repositories but are acting as the primary control plane for multi-vendor AI infrastructure.