Red Hat, Yaskawa and Academia Highlight Multi‑Model AI and Physical‑AI Growth

Red Hat’s Greater China leadership announced a new AI platform strategy that supports multiple models, GPUs and cloud environments. The company stresses that Chinese enterprises need the freedom to choose among domestic and foreign models, varied hardware, and on‑premises, public or hybrid clouds, while maintaining data sovereignty and security. Red Hat AI 3.4 aims to handle distributed inference across any Kubernetes system and adds tools for data tracking and agent governance.

Professor Meng Jun of Wuhan University examined the rise of AI‑generated video models such as Sora, Stable Video Diffusion and others, describing how they enable full‑text‑to‑video creation and may reshape cinematic realism, while questioning whether AI‑driven film can still uphold traditional notions of truth.

National Taiwan University professor Xu Hong‑min projected that embodied or “Physical AI” robotics could reach a commercial sweet spot around 2030. He highlighted the importance of proprietary scene‑level data, advances in vision‑language‑action models, and hardware challenges such as actuators and dexterous hands, noting that cost reductions tied to electric‑vehicle production may drive a “golden cross.”

Japan’s Yaskawa Electric reported a 29 % rise in quarterly orders – the strongest in 15 quarters – driven by AI‑enabled semiconductor and data‑center demand. While revenue grew 10.6 %, profit fell short of expectations due to a core‑system migration and restructuring costs.

Market analyses show AI image generators gaining rapid adoption across marketing, media, gaming and e‑commerce, promising faster, personalized visual content and lower production costs. Parallelly, vendors are developing automated prompt‑extraction tools that convert existing video footage into detailed, structured prompts, aiming to cut creative‑workflow time and improve consistency for AI‑generated visuals.