At the OCP APAC Summit, Ethernet and UALink advocates clashed over data center interconnect architectures. Kurtis Bowman, UALink board chair and AMD's director of architecture and strategy, questioned Ethernet's latency performance in scale-up scenarios, arguing that its technical limits hinder high-performance application scaling

Nvidia's framing of the next AI wave as "physical AI" has fueled investor and industry excitement around humanoid robots, accelerating development. But according to a DIGITIMES report, 2025: The first year of humanoid robots — Global market trends and key technologies, humanoids will represent only 0.2% of the global robotics market in 2025, with usage largely confined to logistics, warehousing, and manufacturing. Broader adoption remains a distant prospect

Tesla and Samsung have struck a US$16.5 billion chip deal seen as a win-win, with Samsung's Texas fab set to produce Tesla's next-gen AI6 chips. Tesla gains customized supply chain support, while Samsung strengthens its foundry credibility and expands its external client base

As automotive computing platforms increasingly demand higher performance, longer lifespans, and near-zero defect rates, chip architecture faces the challenge of balancing system reliability with cost efficiency under stringent automotive standards. Chiplets are emerging as a pivotal pathway to meet these requirements