AMD made several key announcements that underscore its growing AI, enterprise, and client computing leadership. Lisa Su, AMD’s CEO since 2014, was named one of TIME’s 100 Most Influential People of 2025 for her transformative leadership, including strategic moves such as the Zen architecture, Ryzen and EPYC processors, and the Xilinx acquisition, all of which have reshaped AMD’s role across PCs, gaming, data centers, and AI. AMD unveiled a $10 billion global AI infrastructure partnership with Saudi Arabia’s HUMAIN among its significant strategic developments. This initiative will use AMD’s full-stack compute portfolio to build scalable, open AI platforms across regions, from the U.S. to the Middle East.
Looking ahead to Advancing AI 2025 on June 12, AMD plans to showcase next-generation Instinct GPUs, updates to the ROCm software ecosystem, and its overarching AI vision, with Lisa Su and partners delivering a keynote via livestream. The launch of the EPYC 4005 Series brings enterprise-class features to small-to-medium businesses and hosted IT providers, broadening AMD’s leadership in entry-level enterprise compute. At Computex 2025, SVP Jack Huynh will highlight AMD’s innovations in Ryzen, Radeon, gaming, and AI PCs during the press conference on May 21 in Taiwan.
Strategically, AMD finalized its acquisition of ZT Systems, a move aimed at capturing opportunities in the AI accelerator infrastructure market, which is expected to reach $500 billion by 2028. AMD also announced plans to divest ZT’s U.S. manufacturing business. Additionally, AMD continues to deepen its collaboration with Microsoft, bringing new Copilot+ AI features like Cocreator and Live Captions to AMD-powered PCs, and powering Dell’s newly launched Plus 14 2-in-1 notebook with Ryzen AI 300 Series processors.
AMD’s “Venice” EPYC CPUs are the first HPC products to tape out on TSMC’s 2nm process in the data center and server space, reinforcing its roadmap and Arizona fab partnership. AMD-powered instances are also expanding in the cloud, with Oracle Cloud E6 shapes and Google Cloud’s C4D and H4D VMs now leveraging 5th-gen EPYC CPUs and Zen 5 architecture for improved AI and HPC performance. AMD’s Instinct MI325X GPUs led the MLPerf Llama 2 70B benchmarks in a multi-node setup with MangoBoost, and a collaboration with Ansys, Baker Hughes, and Oak Ridge National Lab used 1,024 MI250X GPUs to achieve a 96% reduction in CFD simulation time.
On the software front, ROCm 6.4 has launched, offering pre-optimized containers for Llama 3 and Gemma 3, a modular GPU driver stack, enhanced support for PyTorch and SGLang, and improvements for vLLM inference and Kubernetes orchestration via GPU Operator. AMD also expanded its partnership with Rapt AI, integrating workload automation with Instinct GPUs to streamline AI inference deployment and enable multi-cloud scalability.
AMD introduced the Pollara 400, a programmable AI NIC supporting Ultra Ethernet standards and congestion-aware routing in embedded and edge computing. At the same time, Japan’s KDDI selected 4th-gen EPYC CPUs to power its 5G network virtualization efforts. On the client side, the new Ryzen 8000HX Series mobile CPUs deliver high performance for gaming laptops. Amuse 3.0 Beta, in collaboration with Stability AI, accelerates local generative AI performance on Radeon and Ryzen AI hardware by up to 4.3x.
AMD continues to drive industry standards through the UALink Consortium, which released the UALink 200G 1.0 specification—capable of supporting up to 1,024 accelerators per AI pod with 93% bandwidth efficiency. Additional highlights include a fireside chat with SVP Salil Raje on 2025 priorities such as adaptive compute, SAM, edge AI, and market expansion, and a conversation between AMD CTO Mark Papermaster and Absci CEO Sean McClain on the transformative role of generative AI in drug discovery.
AMD’s 2025 roadmap is gaining momentum across enterprise, edge, and AI computing—from Instinct GPU scale-out and EPYC cloud dominance to Ryzen AI innovation in PCs. As AI becomes the new compute frontier, AMD increasingly positions itself as the engine powering that transformation.
The Cloud Comes Home: Oxide Reimagines On-Prem Computing
Why Enterprise Storage Is Still Stuck in 2010—And How Pure Storage Plans to Fix It
From Chaos to Clarity: HPE’s CloudOps Vision for Hybrid Control
NVIDIA’s Spectrum-X Becomes the Nervous System of Giga-Scale Data Centers
NVIDIA’s Public-Sector Push Turns AI Strategy into National Infrastructure
