Vultr has begun offering AMD's Instinct MI455X GPU and support for the AMD Helios rackscale system, making it one of the first cloud providers to offer the new hardware.
The move adds another option for customers seeking large-memory infrastructure for artificial intelligence and high-performance computing workloads. Vultr is taking pre-orders for reserved capacity, with the new GPU due in the fourth quarter and deployments on AMD Instinct MI455X and AMD Helios planned across 2027 and 2028.
Vultr is positioning the new systems for model training, inference, fine-tuning and agentic AI, as businesses start moving from pilot projects to production use. Helios is AMD's rackscale design, and Vultr will serve as a global provider for the system.
The offer centres on AMD's latest Instinct MI455X graphics processor, built on the chipmaker's CDNA architecture. Vultr said it includes HBM4 memory, wider memory bandwidth and support for low-precision AI data types, features that are becoming more important as operators run larger models and more frequent inference workloads.
AMD Helios is designed to support 72 GPUs in a rackscale architecture. Vultr said direct liquid cooling options will also be available for denser AI infrastructure, reflecting a broader industry shift as cloud operators and data centre groups manage rising heat and power demands from AI clusters.
Infrastructure shift
The announcement comes as cloud providers and chipmakers compete to capture AI computing demand beyond large language model training. Enterprises are placing greater emphasis on inference, where models respond to user prompts or feed automated software agents. That shift is driving demand for systems with substantial memory and more efficient rack-level design.
J.J. Kardwell, Chief Executive Officer of Vultr, said customers were changing how they buy AI infrastructure. "Customers are rapidly moving from AI experimentation into production, and increasingly that means agentic and inference-heavy workloads that require infrastructure built for scale," Kardwell said.
"The AMD Instinct MI455X and Helios rackscale architecture give our customers the flexibility, control, and price-to-performance they need to scale their AI initiatives. Making this architecture available across Vultr's global infrastructure enables teams to build and deploy without constraints," he added.
Agentic AI refers to a fast-growing part of the market in which software systems can complete tasks across multiple steps with limited human intervention. Those workloads often require repeated calls to AI models and access to large amounts of data, increasing pressure on memory, networking and power use inside data centres.
Research firm IDC said it expects a tenfold increase in the number and complexity of enterprise AI agents over the next five years. That forecast has become part of the sales case for vendors seeking to show that AI infrastructure needs are expanding beyond the first wave of chatbot development.
Dave McCarthy, Group Vice President, Cloud and Datacenter Infrastructure at IDC, linked Vultr's move to that shift in demand. "As enterprises transition into the agentic era, the massive influx of complex, data-intensive workflows requires a fundamental shift in infrastructure design," McCarthy said.
"Vultr's deployment of the AMD Instinct MI455X GPU provides the massive memory capacity and rack-scale efficiency necessary to support the continuous inference loops that define next-generation AI workloads," he added.
AMD push
For AMD, the agreement provides another public cloud route to market for its latest data centre AI products as it seeks to compete more directly with Nvidia in AI accelerators. The chipmaker has been expanding both its GPU line-up and its networking offering, including the role of AMD Pensando in scale-out and scale-up systems.
Vultr said Helios combines AMD compute products with AMD Pensando networking in a standards-based design intended to help organisations expand AI infrastructure from rack to cluster. It also highlighted ROCm software, AMD's platform for AI training and inference, as part of the offer.
Andrew Dieckmann, Corporate Vice President and General Manager, Data Centre GPU Business Unit, AMD, said the combined system was aimed at larger AI deployments. "AMD Helios rackscale solution brings the best of AMD compute and networking technology together into a single rack-scale building block, delivering leadership performance, performance per watt, and cost per token for the next wave of AI," Dieckmann said.
"As demand accelerates for large-scale inference and frontier-model training, AMD Helios gives customers a clear path to scale from rack to cluster with the efficiency, economics, and openness needed to power the most demanding AI workloads," he added.
Vultr said it serves hundreds of thousands of active customers across 185 countries and offers cloud compute, cloud GPU, bare metal and storage services. The company disclosed an equity financing in late 2024 that valued the business at USD $3.5 billion.