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.
SEMICON China 2026 spotlighted the scale of the AI investment boom, with Handel Jones, CEO of International Business Strategies (IBS), estimating that global AI and data center capital expenditure has surged from about US$110 billion in 2020 to roughly US$600 billion in 2026.
espite a surge in demand driven by generative artificial intelligence, the fundamental economics of the memory industry remain largely intact. While high-bandwidth memory (HBM) has created a premium segment, the broader market continues to operate on standardized, high-volume production rather than structural product differentiation.
The surge in artificial intelligence demand is pushing the semiconductor industry toward a supply crunch, as Elon Musk outlined a plan to build what he described as the largest chip manufacturing effort in history.
Samsung Electronics' next-generation high-bandwidth memory (HBM4) program is entering a customer validation phase, according to NewDaily, as major technology firms reportedly begin on-site audits of its production lines.
OmniVision Group said on March 20 it will invest CNY1 billion (US$145 million) in Rong Semiconductor (Ningbo) Co. (RongSemi) via a capital increase, taking a 5.88% stake based on a CNY4 billion funding round. The move targets tighter upstream integration, aiming to secure wafer capacity and improve supply chain resilience.
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.
