Vultr reports that it is among the first cloud providers to offer the new AMD Instinct MI455X GPU and support for the AMD Helios rackscale solution. Vultr is now accepting pre-orders available in Q4 to help customers access and operate high-memory rackscale accelerated computing for training, inference, fine-tuning, agentic AI, and HPC.
“Customers are rapidly moving from AI experimentation to production, and increasingly that means inference-intensive and agentic workloads that require infrastructure designed to scale,” said JJ Kardwell, CEO of Vultr. “The AMD Instinct MI455X GPU and Helios rackscale architecture give our customers the flexibility, control, and price-performance ratio they need to scale their AI initiatives. Making this architecture available on Vultr’s global infrastructure allows teams to build and deploy without restrictions.”
Much faster model training
Based on the next-generation AMD CDNA architecture, AMD Instinct MI455X GPUs deliver breakthrough performance and efficiency for the most demanding AI training, inference, and high-performance computing workloads. With expanded HBM4 memory capacity, leading memory bandwidth, advanced support for low-precision AI data types, and AMD Helios rackscale architecture with support for 72 GPUs, the MI455X accelerators enable faster model training, higher-performance inference, and efficient deployment of large-scale foundational models and scientific applications.
For higher density AI infrastructure, direct liquid cooling (DLC) options provide superior thermal management, enabling higher performance at the rack level and greater power efficiency at scale. Complementing the advanced hardware platform, the latest ROCm software innovations enhance AI training, inference optimization and framework interoperability, delivering exceptional performance, scalability and low-latency performance in modern AI and HPC environments.
Additionally, AMD Helios combines scale-up and scale-out high-performance AMD Thinking networks to help organizations build and expand AI infrastructure using open, standards-based technologies.
IDC predicts a tenfold increase in the number and complexity of agents in the enterprise over the next five years. To support agent adoption and scaling, enterprises require an accelerated computing architecture that can accommodate the data-intensive workflow phases and repeated inference loops inherent in running AI agents.
Rack-scale operational efficiency
“As enterprises transition into the agentic era, the huge influx of complex, data-intensive workflows requires a fundamental shift in infrastructure design,” said Dave McCarthy, Group Vice President of Cloud and Datacenter Infrastructure at IDC. “Vultr’s deployment of the AMD Instinct MI455X GPU provides the massive memory capacity and rack-scale efficiency needed to support the continuous inference loops that define next-generation AI workloads.”
AMD Instinct MI455X GPUs Deliver Performance and Efficiency for the Most Demanding AI Training Workloads
“The AMD Helios rackscale solution brings together the best of AMD computing and networking technology into a single rackscale building block, delivering leading performance, performance per watt and cost per token for the next wave of AI,” said Andrew Dieckmann, Corporate Vice President and General Manager of Data Center GPU Business Unit at AMD. “As demand for large-scale inference and frontier model training accelerates, AMD Helios offers customers a clear path to scale from rack to cluster with the efficiency, economics and openness needed to power the most demanding AI workloads.”
Vultr is now accepting pre-orders for reserved capacity for deployments on AMD Instinct MI455X and AMD Helios from 2027 to 2028. Contact the sales team to get started.
