AMD executives are highlighting the company’s open-source strategy—especially its ROCm software—as a key edge over Nvidia’s more closed CUDA ecosystem. On the Q2 2026 earnings call, CEO Lisa Su called ROCm an “inflection point,” with open-source contributions surging 10x in the past year, enabling faster software development by tapping free community talent. AMD posted record $11.5 billion revenue (beating estimates), though the stock dipped 9% after hours amid broader chip-sector volatility.

Publication: Business Insider

AMD has launched the Pensando Vulcano 800 AI NIC, delivering up to 800 Gbps Ethernet and 2.4 Tbps scale-out bandwidth per GPU, directly challenging NVIDIA’s ConnectX-8 SuperNIC in the high-stakes AI networking market. It claims up to 13% faster AI job completion and 33% lower switching costs, targeting large-scale training, inference, and agentic AI workloads in next-gen Helios rack systems. The move intensifies competition with Broadcom (Tomahawk 6), Arista (7800R4 AI Spine), and Marvell (1.6T optics), as networking efficiency, cost, and reliability become as critical as GPU performance in scaling AI clusters.

Publication: Infotechlead

Dell and AMD are expanding on-premises enterprise AI by adding support for AMD Instinct MI350P PCIe GPUs in Dell PowerEdge XE7745 and R7725 servers (starting July 2026). This enables drop-in deployment of generative and agentic AI workloads inside existing data centers—no redesign needed—with high performance (up to 4,600 peak teraflops and 144GB HBM3e memory). The modular Dell AI Platform with AMD also scales from pilot to production, supporting the full AI lifecycle (training, fine-tuning, inference, agentic workflows) in a secure on-premises environment.

Publication: SME Channels