At this year’s Mobile World Congress 2026, AMD highlighted how its newest EPYC processors and AI technologies are powering telecom operators’ transition from experimental AI pilots to full-scale deployments. The company’s focus is on helping carriers evolve from traditional Radio Access Networks (RAN) to open, software-driven, and virtualized network architectures.
AMD showcased a complete technology stack designed to support AI at every stage — from enterprise software and high-performance CPUs and GPUs to advanced networking and adaptive computing platforms. According to AMD, building effective telecom AI isn’t just about having the right models or infrastructure. It requires an open and collaborative ecosystem, robust software to operationalize AI, and efficient compute solutions optimized for the distributed edge.
One of AMD’s major initiatives is its participation in Open Telco AI, a global program launched by the GSMA to accelerate the creation and adoption of “telco-grade” artificial intelligence. The project aims to develop open, shared tools that can understand telecom data and workflows — something general-purpose AI models often struggle with. Tasks such as network optimization, interpreting standards, and troubleshooting demand specialized models trained specifically for the telecom domain.
The collaboration includes contributions from major players: AT&T is providing datasets and foundational Open Telco models, AMD is delivering the compute power, and TensorWave is supplying the cloud infrastructure. Training of these models takes place on AMD’s Instinct GPUs, which use the open ROCm software platform for both training and inference.
To move AI solutions from testing to production, AMD integrates its Enterprise AI Suite — an enterprise-grade platform that bridges open-source AI frameworks with optimized AMD hardware. The suite includes modules for model deployment, governance, validated workflows, and developer operations, all running at scale on GPU clusters. Its Kubernetes-native architecture ensures seamless integration into modern DevOps and MLOps environments while maintaining strong security and team management capabilities.
AMD’s new EPYC 8005 server chips also play a key role. Built for edge computing, these CPUs deliver high compute density for virtual RAN (vRAN) workloads and handle Layer 1 network processing efficiently. They’re engineered for durability, supporting wide thermal ranges suitable for rugged outdoor and small-form-factor telecom systems.
In related announcements from MWC, Samsung Electronics unveiled new joint achievements with AMD across its network solutions portfolio, including advancements in 5G Core, vRAN, and private network offerings. The partnership mirrors similar collaborations in the industry, such as Nvidia’s alliance with Nokia, reflecting growing momentum around AI-powered, open network ecosystems.
