Vultr to Offer AMD Instinct MI455X GPUs and Helios Rack-Scale Architecture on Cloud Platform

Vultr has announced that it will make the AMD Instinct MI455X GPU available on its cloud platform, along with support for AMD's Helios rack-scale solution. The company is now accepting preorders for the hardware, with availability slated to begin in the fourth quarter of 2026. The move positions Vultr to address the growing demand for high-performance AI infrastructure as enterprises shift from experimentation to production-scale deployments.

AMD launched its MI400 series GPUs earlier this month, introducing the MI455X and MI430X models. The MI455X is designed for AI training and inference, while the MI430X targets sovereign AI and high-performance computing workloads. Built on AMD's new CDNA 5 architecture and a 2nm process node, the MI455X features 432GB of HBM4 memory, a 1.5x increase over the 288GB of HBM3E found in the MI355X, and delivers approximately 2.9x the peak memory bandwidth. The GPU also includes 3.6 TBps of low-latency scale-up bandwidth across 36 links.

On the same day, AMD unveiled the Helios rack-scale system, which integrates 72 AMD Instinct MI455X GPUs, 6th Gen AMD Epyc 'Venice' server CPUs, AMD ROCm software, and AMD Pensando networking into a unified rack-scale architecture. The system comprises 18 compute trays and six switch trays, offering a dense, scalable building block for AI workloads.

"Customers are rapidly moving from AI experimentation into production, and increasingly that means agentic and inference-heavy workloads that require infrastructure built for scale," said J.J. Kardwell, CEO of Vultr. "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."

"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," added Andrew Dieckmann, corporate vice president and general manager of the data center GPU business unit at AMD. "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."

The announcement follows a recent collaboration between Vultr, AMD, and the University of Cambridge on a project called TESSERA, an AI foundation model designed to monitor environmental change internationally. Using AMD Instinct MI325X GPUs on Vultr's cloud platform, the University of Cambridge generated global embeddings with a 10-meter resolution, leveraging data from the European Space Agency's Sentinel satellites covering the Earth's land surface from 2017 to 2025. The project aims to support agriculture monitoring, biodiversity conservation, and renewable energy infrastructure planning.

Professor Anil Madhavapeddy, Professor of Planetary Computing at Cambridge's Department of Computer Science & Technology and co-director of the Cambridge Centre for Earth Observation, said: "Our goal is to democratize access to planetary-scale environmental monitoring. By making TESSERA's embeddings freely available under a CC-BY license and publishing the complete training pipeline, we're ensuring that any researcher, government, or organization worldwide can deploy this technology for their specific conservation needs." He added, "This partnership exemplifies how cloud infrastructure can accelerate scientific research with genuine planetary impact. We're enabling a new paradigm for open, accessible environmental monitoring that can inform policy decisions and conservation actions in the UK and worldwide."

The addition of the MI455X and Helios to Vultr's platform underscores a broader industry trend toward rack-scale architectures that combine compute, memory, and networking into pre-integrated building blocks. This approach is expected to reduce deployment complexity and improve efficiency for AI workloads, particularly as demand for large-scale inference and frontier-model training continues to rise. By offering these technologies across its global infrastructure, Vultr aims to give customers the flexibility and performance needed to scale AI initiatives without infrastructure constraints.