AMD Unveils New Hardware and Software at Advancing AI

Dr Lisa Su, CEO of AMD, and her team have revealed several major hardware and software innovations at AMD’s Advancing AI event today, all set to power the next generation of data centres, enterprise workloads and physical AI systems. 

Spanning hardware and software – from sixth generation EPYC CPUs and Instinct MI400 GPUs to the new ROCm.ai developer suite – the announcements mark a significant step forward in AMD’s open ecosystem strategy. 

“Our mission is to push the boundaries of high performance and AI computing,” Lisa said, during the opening of the Advancing AI keynote. “I often say AI is the most important technology in the last 50 years.”

Lisa pointed to the industry consuming more than 35 quadrillion tokens every month, with 60% of today’s capacity used for inference rather than training.

She put this down to the increased use of agentic AI: “Compute demand is growing at an incredible pace. When you ask the agent to do something, it has dozens of steps… so you need lots of GPUs to do all that reason and a lot of CPUs to orchestrate every step around it.”​​​​​​​

The EPYC 9006 Series 

The sixth generation of AMD’s EPYC Service CPUs deliver solutions for agentic AI, enterprise and high-performance computing (HPC) workloads.

With agentic AI depending on CPU-powered infrastructure to run agents, coordinate work around the model and connect to enterprise data and services, the sixth gen matches the right CPU to each stage of the agentic workflow to improve throughput, utilisation and efficiency across the data centre. 

The series – including EPYC 9006 SP7, EPYC 9006 SP8, 9006X SP7 and 9006 LP – are built to match the right processor to the right job.

“As agentic AI reshapes how enterprise infrastructure is designed and deployed, EPYC is evolving with our customers once again,” says Dan McNamara, Senior Vice President and General Manager of Compute and Enterprise AI at AMD.

“6th Gen AMD EPYC delivers workload-optimised compute for the full spectrum of modern data centre needs; from running more agents and feeding GPUs across rack-scale AI infrastructure to powering the business-critical applications and services organisations rely on every day.”

ROCm.ai 

AMD’s new AI-native developer experience – ROCm.ai – is designed to streamline AI development across AMD platforms from initial intent to full production. 

ROCm.ai unifies deployment, optimisation and AI-assisted tools through three core features:

• ROCm CLI: A new command-line interface for simplified installation, validation, serving and workload management
• AMD Skills: AMD-authored expertise integrated directly into popular AI coding assistants like Claude, Cursor and Codex
• Hyperloom: An open-source, agentic system built to automate end-to-end inference optimisation.

Together, these tools deliver a streamlined, unified workflow that allows developers to install, deploy and troubleshoot AI workloads faster on AMD hardware. According to AMD, ROCm.ai boosts performance with up to 3.3x faster inference and 2.4x faster training on average. 

“AMD is committed to helping developers move faster than ever,” says Anush Elangovan, Corporate Vice President of AI Software at AMD. “ROCm.ai brings agentic AI development to AMD platforms – agents that don’t just answer questions, but profile, debug and drive workloads toward peak performance on AMD hardware.”

The Instinct MI400 Series 

Meanwhile, the Instinct MI400 Series GPUs are designed to tackle large-scale frontier AI and HPC workloads.

The line-up features the MI455X, tailored for hyperscale AI factories and frontier training using advanced HBM4 memory, as well as the MI430X, purpose-built for sovereign AI and scientific computing. 

“The AMD Instinct MI400 Series extends our AI portfolio with purpose-built solutions optimised for the full spectrum of AI and HPC deployments, built on an open software foundation that gives customers the flexibility to innovate at every scale,” says Vamsi Boppana, Senior Vice President of AI at AMD.

Combining leadership compute, advanced security and flexible system architecture, the MI400 series aims to provide high-performance power needed to scale complex workloads with absolute confidence. 

Joining forces with AT&T and Microsoft for Otel 2.0

Following on from the Open Telco AI launch at MWC 2026, AMD, AT&T and Microsoft have developed OTel 2.0, an open-source model developed specifically for the telecommunications industry. 

Trained on dataset contributions from the Open Telco AI initiative, AT&T reports that OTel 2.0 is currently the largest and best-performing open model built for telco operations.

General-purpose AI models often struggle with complex domain-specific telecommunications standards and network data, AMD points out. OTel 2.0 addresses this gap by offering a domain-tailored model aimed at helping operators automate network management, accelerate troubleshooting and improve system resiliency.

Building the model was a team effort. AT&T used Microsoft Foundry, powered by AMD Instinct GPUs, to analyse over a trillion data points, picking out the top 400 billion most useful examples. AT&T then used that refined collection of data to train OTel 2.0 to handle real-world telecommunications tasks. 

Looking ahead, the collaboration moves from initial model publication into real-world evaluation and deployment testing. For telecom operators, OTel 2.0 represents a shift toward practical, open-source AI tools tailored for network infrastructure.

For AMD, it marks an important proof point for its hardware and ROCm software ecosystem across managed cloud environments.

Helios Rackscale for frontier AI

AMD also launched its Helios rack-scale solution at Advancing AI. 

The open infrastructure platform is engineered for foundation-model training and frontier AI deployment, integrating with the sixth generation of EPYC CPUs, Instinct MI455X GPUs, Pensando networking and ROCm software into a single architecture. 

Built for high efficiency, AMD reports Helios offers 50% more high-bandwidth memory capacity and up to 30% more tokens per dollar compared to competing systems. 

Designed as a modular building block, Helios allows data centres to scale seamlessly from individual racks up to gigawatt-scale AI clusters. 

Empowering physical AI

AMD has also introduced its Robotics Partner Network, bringing together hardware and software partners with the aim to accelerate the development and deployment of robotics and physical AI systems powered by AMD platforms. 

The programme is open to original design manufacturers, independent software vendors, simulation providers, sensor and perception companies, safety and benchmarking partners, and system integrators. 

“No single company will build the future of physical AI,” explains Amey Deosthali, Senior Director of Industrial, Robotics and Healthcare at AMD.

“Building robots requires bringing together AI models, middleware, sensors, simulation tools, system integrators and production hardware. Too often, those pieces come from disconnected vendors with little validation between them.

“With the AMD Robotics Partner Network, we are extending the AMD open ecosystem approach into robotics to help reduce integration efforts, shorten development cycles and allow roboticists to move from reference platform to production systems with greater flexibility, portability and confidence.”

Empowering physical AI

AMD has also introduced its Robotics Partner Network, bringing together hardware and software partners with the aim to accelerate the development and deployment of robotics and physical AI systems powered by AMD platforms. 

The programme is open to original design manufacturers, independent software vendors, simulation providers, sensor and perception companies, safety and benchmarking partners, and system integrators. 

“No single company will build the future of physical AI,” explains Amey Deosthali, Senior Director of Industrial, Robotics and Healthcare at AMD.

“Building robots requires bringing together AI models, middleware, sensors, simulation tools, system integrators and production hardware. Too often, those pieces come from disconnected vendors with little validation between them.

“With the AMD Robotics Partner Network, we are extending the AMD open ecosystem approach into robotics to help reduce integration efforts, shorten development cycles and allow roboticists to move from reference platform to production systems with greater flexibility, portability and confidence.”

The new Ryzen AI Embedded X100 Series is a single-chip processor engineered for physical AI systems like robotics, smart manufacturing and healthcare.

The system-on-chip combines up to 16 Zen 5 CPU cores, an integrated GPU and a dedicated Neural Processing Unit (NPU) into one package. 

By sharing a unified memory architecture, the processors minimise latency and boost real-time control, allowing autonomous machines – such as surgical robots or humanoid devices – to quickly perceive, reason and act in physical environments while operating under tight power, thermal and space constraints.

“The Ryzen AI Embedded X100 Series marks an exciting new chapter for physical AI, unlocking opportunities across a broad range of intelligent, real-time embedded applications,” adds Kirk Saban, Corporate Vice President of Product, Software and Solutions at AMD.

Furthermore, the new Kria AI Solutions – a turnkey platform powered by the Ryzen AI Embedded X100 Series – helps developers quickly build and deploy physical AI and autonomous robotics. 

Kria essentially bridges the gap between a robot’s body and its brain, enabling real-time sensing, reasoning and decision-making on an open, low-latency framework.