Samsung eyes improved AI chips with "competitive advantage"

Featured image for Samsung eyes improved AI chips with

Samsung will respond to the growing demand for artificial intelligence (AI) semiconductor chips, according to a report from FN News via SamMobile. The company’s Samsung Foundry, an arm that focuses on chipmaking, will improve on AI chips with a “competitive advantage.” Samsung shared this information in Hong Kong at a forum for investors. “Samsung Electronics has a very unique competitive advantage,” said Jeong Ki-bong, the vice president of Samsung Foundry. “I call it GDP.”

According to the report, the so-called GDP advantage represents three aspects of Samsung chipmaking. First, it references the gate-all-around (GAA) chip fabrication technology that Samsung has employed in recent production. It also includes high-bandwidth memory and advanced packaging technology. That’s how you get the acronym GDP: GAA, DRAM, Packaging. While Samsung is not necessarily the only company to offer these technologies individually, Samsung executives believe they have an advantage by putting all of this together.


“Hyperscaler (large data center operators), automotive OEMs, Tesla, and other customers came to us because they wanted chips they designed,” Jeong said. “When asked why they came to us, they say, ‘Samsung has all three.'”

Growing demand for AI results in greater performance needs

Tasks and processes that utilize AI can require more performance than typical chips offer. That’s why Samsung says that it has received the largest amount of new orders in the high-performance computing (HPC) sector. The other notable sectors of Samsung Foundry development are the mobile and automotive markets. The mobile processing market is still the largest that Samsung serves, represents 54% of Samsung Foundry’s estimated sales this year. HPC now slots in at 19% for second-place, with automotive sales estimated to make up for 11%.

“It’s our mission to bring semiconductors, including power and memory, to life to the imagination, and some customers will be selling the 4 nanometer AI accelerators under development,” said Jeong. “The first electric vehicle company in the automotive industry, is also developing the next version of the 5 nanometer, a fully autonomous chiplet.”

Although Samsung is still posting slightly-sluggish sales, these positive statements come as the company goes all-in on AI chips and HPC. As more applications for AI-based products and services come to market, they may command better-performing equipment. We are already seeing large AI learning models require large amounts of onboard memory for on-device computing. If Samsung’s proclaimed “GDP advantage” really is appealing to OEMs, it could be the key to Samsung Foundry’s success. However, it will still face tough competition from Taiwan Semiconductor Manufacturing Corporation (TSMC), which is currently the premier chip fabricator.