According to Pulsestacks on May 10th, citing industry sources, Samsung Electronics' AI inference chip Mach-1 is poised to undergo prototype production using the MPW (Multi-Project Wafer) method, potentially based on Samsung's in-house 4nm process.
The industry source also mentioned the possibility of Mach-1 utilizing 5nm process technology.
Samsung has set a timeline for Mach-1: mass production in the second half of this year, chip delivery by the end of this year, and delivery of servers based on this chip in the first quarter of next year.
Samsung has also secured a pre-order from Naver worth up to 1 trillion Korean won (approximately 52.8 billion RMB at the current exchange rate).
While Samsung Electronics currently offers 3nm foundry services, adopting the more mature 4/5nm process for Mach-1 can enhance project execution stability.
Mach-1 chip is seen as a crucial step in Samsung Electronics' AI strategy. Previously, Samsung Electronics' AI-related products mainly focused on high-performance storage and edge AI SoCs, without venturing into the server-level computing chip domain.
Mach-1 inference chip utilizes LPDDR external memory. Samsung Electronics' Co-CEO Kyung Kui-hyun has stated that the chip's development goal is to reduce the data bottleneck between external memory and computing chips to 1/8 through algorithms, while also improving energy efficiency eightfold.
Samsung Electronics is currently recruiting more talent for Mach-1 development: externally, Samsung announced 8 job positions related to the chip in April; internally, Korean media reports that Samsung is transferring some personnel from the System LSI department's automotive semiconductor team to the Mach-1 project.
