Efficient Computer

Korean chip giants Samsung Electronics and SK Hynix are moving into co-packaged optics (CPO) technology, which links AI semiconductors using optical connections to boost data transfer speeds and energy efficiency. As AI computing demand surges, Korean media describe the shift as a new race in which chips compete 'at the speed of light', alongside global players developing similar interconnect solutions.

Researchers report that hyperbolic wave attractors, structures that trap and concentrate light waves, could be used to build next-generation optical chips. By confining light more efficiently, the approach may improve photonic circuits used in computing and communications. Details of the research and its practical applications remain limited in the available reporting.

Retailers are increasingly turning to AI-powered shelf technology, with Modern Retail reporting on the growing adoption of automated shelf monitoring and inventory systems. The trend points to bricks-and-mortar stores using sensors and computer vision to track stock levels, pricing and product placement in real time, as chains look to cut waste and compete with the efficiency of e-commerce operations.

Open Source For You has published a feature on Infrastructure Processing Units (IPUs), describing them as the unsung heroes of the data centre. The piece highlights how these specialised processors offload networking, storage and security tasks from main CPUs, improving efficiency in modern server infrastructure. Interest in dedicated infrastructure processors continues to grow as data centres handle heavier workloads and cloud providers seek to free up compute capacity.

Software engineer Evan Schwartz has published a blog post urging makers of Jev-style AI models — lightweight open models whose efficiency drives heavier overall usage, echoing the Jevons paradox — to add prompt caching. Caching previously processed prompts would cut redundant computation, lower latency and reduce serving costs. The post is being shared among AI and open-source engineering communities, where efficiency and inference costs are active topics of debate.