[ad_1]
Samsung plans to launch an AI accelerator chip next year. The confirmation came from the company’s President and co-CEO Kye Hyun Kyung during the recently held annual shareholders’ meeting. Kyung said the Korean firm has already completed the technological verification of the first-gen model, Mach-1. The chip will arrive in early 2025.
Samsung will revolutionize the AI chip market in 2025
Major tech companies are vying to be at the forefront of the ongoing AI revolution. Samsung, one of the biggest names in the tech industry, started making its move in 2023. It introduced an LLM (Large Language Model) called Samsung Gauss and followed up with Galaxy AI on the Galaxy S24 series. Gauss powers some of the AI features on the new flagships.
The Korean behemoth is now working on expanding these AI features to more Galaxy devices. It is simultaneously new ways to strengthen its grip on the AI market. It appears Samsung sees a big opportunity in AI chips, a market currently dominated by Nvidia. Samsung wants to break the stronghold by introducing its AI accelerator chip as early as 2025.
The company has already made significant progress in its development and plans to begin production later this year. “Technological verification for Mach-1 has been completed through programmable chip (FPGA), and system-on-chip (SoC) design is in progress,” Kyung said. “We will produce the chip by the end of the year and see the AI system made with our chip early next year.”
Kyung is one of the two CEOs of Samsung— Han Jong-hee is the other CEO. He is also the Director of SAIT (Samsung Advanced Institute of Technology), the company’s research and development unit, and the head of the DS Division, i.e., the semiconductor unit. The Korean firm announced a major leadership shuffle at the end of 2023 but didn’t change its CEOs.
Samsung may revolutionize the market with its first solution
Samsung’s first AI accelerator chip is said to address one of the biggest problems with these products. The current crop of AI accelerators struggle due to a bottleneck during AI inferencing. This leads to a drop in data transmission speed and power efficiency when exchanging information between the CPU, GPU, and memory. Samsung aims to solve these issues.
According to the Korean media, Mach-1 will feature an “unprecedented structure” capable of reducing this bottleneck by almost 90%, i.e., to 1/8th of current levels. Thanks to its “lightweight” chip design, it will also reduce HBM (high-bandwidth memory) use for AI semiconductors. Samsung hopes these advancements will help it topple Nvidia.
The American chipmaker currently supplies AI accelerators to Microsoft, Meta, and other big companies. Samsung, which has doubled down on its AI efforts lately, would be looking to steal a few customers from it. The Korean firm recently established an artificial general intelligence (AGI) computing lab in Silicon Valley, USA. “The specialized lab is working to create an entirely new type of semiconductor, a semiconductor designed to meet the processing requirements of future AGI,” Kyung said.
[ad_2]
Source link