Tag: #hbm-memory17 results
Samsung Electronics is in final negotiations with its labor union to avoid a massive strike involving 45,000 employees. A potential work stoppage threatens to disrupt the supply of memory chips critical for AI data centers and consumer electronics.
China unveils domestic HBM3e packaging with 960GB/s bandwidth, targeting AI chips on 3nm and below processes. Despite HBM4 emerging for AMD/NVIDIA flagships, HBM3e stays mainstream amid surging AI memory demand.
SK Hynix reports smooth progress supplying HBM4 products to customers. Company confirms no current plans for stock split. This update reassures AI hardware supply chain stability.
SK Hynix plans to list shares in the US this year to raise capital. The move addresses artificial intelligence's surging demand for memory chips. This strategy helps the company keep pace with AI-driven growth.
SK Hynix plans to spend 11.9 trillion won ($7.9 billion) on ASML's cutting-edge extreme ultraviolet lithography tools. This investment deepens its push into next-generation memory production. The move targets surging demand from AI applications.
Huawei launched the Ascend 950PR processor in the Atlas 350 AI accelerator card at its 2026 partner conference. It delivers nearly 3x performance of H20 with support for FP4 precision. Major upgrades include low-precision formats, vector compute, interconnect bandwidth, and self-developed HBM.
Korean chipmakers SK Hynix and Samsung reversed fortunes via HBM bets for AI GPUs, capturing 80% market and record $25B Feb exports. SK Hynix sold out 2026 HBM capacity after pioneering HBM3E mass production. Deep Nvidia collaboration via on-site engineers secured demand.
NVIDIA will unveil its next-gen Feynman GPU at GTC conference March 16-19. The architecture features a 1.6nm process and exceeds 1000W power consumption. It pairs with next-gen HBM memory, following the naming tradition after physicist Richard Feynman.
Nvidia has downscaled HBM4 performance specs for its next-gen Rubin architecture VR200 GPU as SK Hynix and Samsung can't hit the 22TB/s bandwidth target in production. Per SemiAnalysis report, this reflects supply chain challenges.
SK Hynix locked in 10% annual profit sharing with unions for 10 years, yielding massive AI-driven HBM bonuses. Samsung rejects similar formula, faces historic strike from May 21 impacting 3-4% global DRAM. Signals profit redistribution push in AI supply chain.