AMD has poached Amazon Web Services’ (AWS) infrastructure leader as lead on its open-standard Helios rack platform.

Arvind Balakumar joined AMD in November as corporate vice president of engineering for AI infrastructure. In a LinkedIn post, the exec revealed he would lead on cluster-scale AI infrastructure solutions for Helios, “unifying EPYC CPUs, Instinct GPUs, Pensando DPUs, and the ROCm software stack into an open, AI-optimized rack and cluster-level architecture built to power the AI factories of the future.”

At AWS, Balakumar headed the hyperscaler’s global infrastructure scalability across compute, networking and data centers. Prior to AWS, he was head of product and technology partnerships for compute and AI infrastructure at hyperscale rival Google Cloud, following a nine-year stint at Intel, overseeing engineering and product for computer vision segment RealSense, which Intel spun out in July.

That Intel experience will likely come in handy working on Helios, AMD’s alternative to Nvidia’s Vera Rubin platform.

Helios saw its first commercial deployment via HPE last week, as built on HPE Juniper Networking hardware and software and Broadcom’s Tomahawk 6 networking chip.

HPE’s deployment connects 72 AMD Instinct MI455X GPUs in a single rack, as well as touting 31 terabytes of fourth‑generation high‑bandwidth memory (HBM4) and 1.4 petabytes per second of memory bandwidth, with HPE targeting high capacity and data movement from trillion-parameter model training, high inference throughput, and extremely large AI model sizes.

The AI use case was highlighted by Balakumar on Linkedin, who wrote: “The demand for AI infrastructure is insatiable, and meeting that demand requires breakthrough innovation across the entire compute stack from silicon architecture and high-performance interconnects all the way to power delivery and grid modernization.

AMD is uniquely positioned to lead this era, powered by its broad compute portfolio and its firm commitment to an open ecosystem [and] I look forward to working closely with … AMD’s world-class leadership teams to scale our AI infrastructure ambitions and tackle the industry’s most complex customer challenges at unprecedented scale.”