Samsung Electronics is nearing a technical milestone for its 2nm process, but yields are reportedly still below the level required for stable mass production, raising questions about its ability to secure major foundry customers for the next-generation node.
AI workloads are scaling faster than traditional CPU architectures were ever designed to handle, exposing clear limits in performance per watt. SiFive's latest funding round highlights a shift in industry thinking, with RISC-V moving from niche alternative to a credible option for next-gen computing.
SK Hynix is reportedly considering reducing its planned 2026 shipments of high-bandwidth memory (HBM4) to Nvidia by about 20-30%, amid delays in ramping up Nvidia's next-generation Vera Rubin platform, according to ZDNet Korea.
China's imports of semiconductor manufacturing equipment from Southeast Asia surged in 2025, surpassing direct shipments from the US, according to a Nikkei Asia analysis, as Washington moves to tighten export controls on advanced chipmaking tools.
Broadcom and Meta have unveiled a sweeping multi-year, multi-generation partnership aimed at scaling Meta's AI infrastructure, signaling a deeper shift toward custom silicon and vertically integrated AI systems. The collaboration centers on Meta's Training and Inference Accelerator (MTIA) chips, with Broadcom providing the underlying technologies and co-design expertise through its XPU platform.
