Google has introduced TurboQuant, a compression algorithm that reduces large language model (LLM) memory usage by at least 6x while boosting performance, targeting one of AI's most persistent bottlenecks: memory. The breakthrough lowers inference costs and expands deployment across cloud and edge environments.
The global DRAM industry is approaching a structural inflection point, as traditional scaling methods struggle to deliver the performance gains required by artificial intelligence workloads. With next-generation architectures such as 4F-squared (4F²) and 3D DRAM facing rising complexity and potential delays, manufacturers are being forced to reassess near-term roadmaps and rely more heavily on incremental and material-level improvements.
Semiconductor manufacturers are racing to secure critical materials as Middle East tensions disrupt supply chains, with the risk of production disruption outweighing rising costs.
SMIC on March 26 outlined an action plan to strengthen core operations and identify new growth drivers in 2026, targeting revenue growth above the industry average as it reinforces its role in China's semiconductor self-sufficiency strategy.
As Nvidia and Coherent signal a shift toward all-optical networking to solve the AI power crisis, Redmond-based Lumotive has announced a milestone that could redefine data center scalability. In an interview with DIGITIMES Asia, Gleb Akselrod, Co-founder and CTO of Lumotive, detailed the success of the world's first programmable 2D optical beamforming chip.
