The episode opens by discussing the timeline for CPO mass production, noting that while Broadcom, Marvell, and MediaTek have CPO developments, NVIDIA might accelerate faster due to customer pressure. It then reports that Samsung Electro-Mechanics has become the primary substrate supplier for Groq's LPU, which is built on Samsung's 4nm process and serves as an inference accelerator for NVIDIA's Verrubin platform. Next, it covers Intel joining Elon Musk's TerraFab initiative, which could help Intel validate its 18A technology and potentially disrupt traditional fabless-foundry models. The episode shifts to memory industry restructuring, highlighting how YMTC plans to start mass production at a new Wuhan line in late 2024, potentially making it the third-largest NAND producer globally, while CXMT focuses on HBM production in Shanghai. It then details Apple's multi-year agreement to license Google's Gemini AI for Siri, indicating Apple's reliance on external AI technology as it aims to move more AI processing to on-device chips. The episode also mentions Samsung Electronics' upcoming foldable phone launches, including the Galaxy Z Widefold with a 4:3 screen ratio. Finally, it discusses the expansion of Robotaxi competition into Europe, with companies like Vamo planning services in London to test autonomous systems in complex urban environments.
