Vultr announced that it will become one of the first cloud providers to offer AMD’s new Instinct MI455X GPUs and support the AMD Helios rackscale solution, extending its AI cloud platform with infrastructure designed for large-scale AI training, inference, agentic AI, and high-performance computing (HPC). The company is accepting pre-orders now, with customer deployments scheduled to begin in Q4 2026 and reserved capacity available through 2028.
The deployment centers on AMD’s Helios rackscale architecture, which integrates up to 72 AMD Instinct MI455X accelerators with AMD Pensando networking into an open Ethernet-based AI infrastructure platform. Built on AMD’s next-generation CDNA architecture, the MI455X introduces HBM4 memory, higher memory bandwidth, expanded support for low-precision AI data types, and direct liquid cooling options for higher rack density and improved energy efficiency. AMD also highlighted continued ROCm software enhancements for AI training, inference optimization, and interoperability across leading AI frameworks.
Vultr said the new infrastructure targets enterprises moving production AI workloads beyond experimentation toward inference-intensive and agentic AI deployments. Alongside the infrastructure announcement, Vultr and AMD also highlighted their collaboration with the University of Cambridge’s Energy & Environment Group, which is using Vultr cloud infrastructure powered by AMD Instinct MI325X GPUs to develop TESSERA, a planetary-scale AI foundation model that analyzes European Space Agency Sentinel satellite imagery for applications including precision agriculture, biodiversity monitoring, and renewable energy planning.
“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,” said Andrew Dieckmann, Corporate Vice President and General Manager of AMD’s Data Center GPU Business Unit.
🌐 Analysis
The announcement extends AMD’s strategy of delivering complete rack-scale AI infrastructure rather than standalone accelerators. Helios combines GPUs, networking, cooling, and ROCm software into an integrated platform intended to compete with vertically integrated AI systems while preserving an open Ethernet-based architecture. Cloud providers such as Vultr are becoming key launch partners that can make next-generation GPU platforms available to enterprises before many organizations deploy on-premises infrastructure.
The accompanying TESSERA collaboration demonstrates another emerging trend in AI infrastructure: cloud providers increasingly pair infrastructure announcements with high-profile scientific workloads that showcase large-scale training and inference capabilities. Environmental modeling, satellite imagery analysis, and geospatial foundation models are becoming important HPC and AI use cases alongside enterprise generative AI.
