AMD used its Advancing AI 2026 event to introduce one of its broadest artificial intelligence portfolios to date, unveiling new hardware, software, and platform technologies designed to support AI workloads from hyperscale data centers to enterprise environments, personal computers, and autonomous robotics. Alongside the new product launches, AMD also highlighted collaborations with major technology companies including OpenAI, Meta, Anthropic, Cerebras, Cisco, Microsoft, and AT&T as it expands its open AI ecosystem.
At the center of the announcements is AMD’s strategy of delivering a full-stack AI platform built around open standards, allowing organizations to deploy infrastructure suited to a wide range of AI workloads without relying on proprietary ecosystems.
New AI Infrastructure Across Data Centers, Edge, and Robotics
One of the biggest launches was AMD Helios, the company’s first rack-scale AI infrastructure designed for frontier AI, large-scale inference, and foundation model training. The platform integrates AMD Instinct MI455X GPUs, 6th Generation AMD EPYC server processors, AMD Pensando networking, and the ROCm software platform into a unified architecture that can scale from a single rack to large AI clusters. AMD said Helios is designed to simplify deployment while providing the compute, networking, and software foundation needed for next-generation AI infrastructure.
Supporting the new infrastructure is the introduction of the AMD Instinct MI400 Series GPU family. The lineup includes the MI455X accelerator for frontier AI and AI factory deployments alongside the MI430X accelerator aimed at sovereign AI and high-performance computing. AMD positions the new GPU family as a solution capable of handling both AI inference and scientific computing through a combination of high-bandwidth memory, scalable architectures, security features, and support for the open ROCm software ecosystem.
AMD also expanded its data center processor lineup with the launch of the 6th Generation EPYC 9006 Series server CPUs. Rather than targeting a single workload, the processor family is optimized for different AI and enterprise scenarios, including running AI agents, supporting GPU-heavy AI infrastructure, traditional enterprise applications, and high-performance computing. AMD describes this as a workload-optimized approach that allows organizations to match processors with specific infrastructure roles as agentic AI becomes a larger part of modern data centers.
Software also played a major role during the event. AMD introduced ROCm.ai, a new AI-native development experience designed to simplify building, deploying, troubleshooting, and optimizing AI workloads across AMD hardware. The platform combines ROCm CLI, AMD Skills for AI coding assistants, and Hyperloom, an AI-powered optimization system that automates performance tuning. By integrating AMD-authored expertise into coding assistants such as Claude, Cursor, and Codex, the company aims to reduce the complexity of AI software development while improving workload optimization through AI-assisted tools.
Beyond cloud and data center deployments, AMD expanded its client AI strategy through Ryzen AI platforms. The company demonstrated how Ryzen AI Halo systems are intended to help developers, creators, and enterprises run increasingly capable AI models locally. AMD also announced a broader collaboration with Hugging Face to provide optimized models and development tools for Ryzen AI Halo while offering a year of Hugging Face Pro with future developer platforms.
AMD’s embedded computing portfolio also received significant updates with the introduction of Ryzen AI Embedded X100 Series processors. Designed for robotics, industrial automation, healthcare, aerospace, defense, and other intelligent embedded systems, the processors combine CPU, GPU, and neural processing capabilities within a single embedded platform. AMD said the processors are built to support AI perception, reasoning, and real-time control while operating under strict thermal, power, and reliability requirements found in industrial environments.
Building on those embedded processors, AMD introduced Kria AI Solutions, which extend the company’s robotics portfolio from real-time control systems into higher-level AI reasoning and autonomous decision-making. The portfolio includes Kria AI System-on-Modules, robotics carrier hardware, and the Kria AI Robotics Developer Platform, which combines CPU, GPU, NPU, and FPGA computing into an integrated robotics platform. AMD said the platform is designed to help developers move from prototyping to production while supporting AI perception, reasoning, control, and robotics workloads within a single open development environment.
To support broader robotics adoption, AMD also launched the AMD Robotics Partner Network. The program brings together hardware manufacturers, software developers, AI model providers, simulation companies, robotics specialists, sensor providers, and system integrators around AMD platforms built on open standards. The initiative is intended to simplify robotics development, reduce integration complexity, and provide validated hardware and software combinations that can help accelerate deployment of physical AI systems.
Partnerships Drive AMD’s AI Ecosystem
AMD highlighted several strategic collaborations that expand its AI ecosystem across infrastructure, software, cloud services, and large-scale deployments.
AMD and OpenAI showcased an expanded collaboration that spans silicon, software, and future AI infrastructure. The companies are working together on optimizing GPT-class workloads using AMD hardware while preparing AMD Helios deployments through multiple deployment partners. Their collaboration also extends into future infrastructure planning, with OpenAI providing insights into evolving AI model requirements that help shape AMD’s future hardware development.
Meta also detailed its growing collaboration with AMD across AI infrastructure. The companies are co-engineering AI systems covering processors, GPUs, networking, software, memory, cooling, and rack-scale deployments designed for gigawatt-scale AI infrastructure. Meta is expanding its use of AMD EPYC processors while continuing validation work on future AMD platforms and Helios systems for large-scale AI deployments.
AMD and Anthropic outlined a strategic partnership centered on deploying AMD Instinct MI455X GPUs through AMD Helios rackscale solutions. The companies are also collaborating on software development, using Claude to help optimize workloads for AMD Instinct GPUs and accelerate ROCm software development.
AMD also announced a technical partnership with Cerebras to create a disaggregated AI inference solution that combines AMD Helios with Cerebras Wafer-Scale Engine technology. Under the collaboration, AMD Helios will provide high-throughput processing while Cerebras focuses on ultra-low-latency token generation for AI workloads requiring both speed and scalability.
Enterprise AI Initiatives
AMD also introduced collaborations focused on helping enterprises adopt AI across local, hybrid, and industry-specific environments.
AMD and Cisco are collaborating to combine Ryzen AI Halo systems with Cisco networking, observability, governance, and security technologies. The goal is to help organizations deploy local and hybrid agentic AI while maintaining centralized management, visibility, and security across enterprise environments.
Meanwhile, AMD, AT&T, and Microsoft highlighted progress on Open Telco AI through the introduction of OTel 2.0, an open-source model developed specifically for telecommunications workloads. Using AMD Instinct GPUs and ROCm software within Microsoft infrastructure, AT&T processed more than one trillion tokens while training the updated model. The initiative aims to provide telecommunications operators with AI models better suited for interpreting network data, understanding telecom standards, and supporting operational workflows.
AMD’s Broader Vision for AI
Across the announcements, AMD consistently emphasized open standards and an open software ecosystem as central themes. From ROCm software and open networking technologies to robotics frameworks and AI development tools, the company positioned openness and interoperability as key elements of its long-term AI strategy across cloud infrastructure, enterprise systems, edge computing, client devices, and robotics.
With new AI infrastructure, processors, accelerators, software platforms, embedded solutions, and partnerships announced during Advancing AI 2026, AMD outlined an AI roadmap that spans nearly every layer of modern computing. Rather than focusing on a single segment of the AI market, the company presented a portfolio intended to support workloads ranging from frontier AI and enterprise applications to local AI experiences and autonomous robotics, reflecting its broader effort to build an integrated AI ecosystem across the data center, edge, and physical world.
You must log in to post a comment.