Samsung plans to release a chip containing artificial intelligence technology under the name Mach-1 at the beginning of 2025, according to the company’s announcement during its fifty-fifth annual shareholder meeting, as reported by Sedaily newspaper.
With this move, Samsung will find itself competing with companies like Nvidia, despite not having plans to compete with advanced artificial intelligence solutions such as Nvidia’s H100 and B100 and B200 GPUs, at least not until now.
The Mach-1 chip from Samsung is designed for artificial intelligence recognition and features LPDDR memory, making it suitable for edge computing applications.
Reports indicate that the Mach-1 chip contains a unique feature that reduces the memory bandwidth requirements for inference by up to approximately 0.125 times compared to current designs.
Kyung Heon Kyung, Head of Samsung’s Device Solutions Division, mentioned that the chip design has been technologically validated in a Programmable Logic Gates Array (FPGA), and the final touches on the chip are advanced and will be completed soon.
Kyung affirmed that the chip will be ready by the end of the year, paving the way for the early launch of the artificial intelligence system with the Mach-1 chip.
The Mach-1 chip from Samsung is not aimed at competing with high-performance artificial intelligence processors such as AWS Trainium or Nvidia H100, but will certainly compete with other inference solutions, such as AWS Inferentia.
Samsung is likely to target the Mach-1 artificial intelligence chip towards edge applications that require low power consumption, compact dimensions, and low costs.
According to Kyung, Samsung is not only focusing on developing the Mach-1 chip, as the company has established a lab in Silicon Valley dedicated to general artificial intelligence (AGI) research, indicating Samsung’s desire to become a key player in the field of artificial intelligence technology in the future. The mission of this lab is to develop new types of processors and memories that align with the needs of general artificial intelligence systems.
