Samsung Electronics Co. plans to up synthetic intelligence chip foundry gross sales to about 50% of its whole foundry gross sales in 5 years with its superior nanometer silicon course of know-how as extra tech firms go for self-designed AI chips.
Samsung Foundry, the semiconductor foundry enterprise at Samsung Electronics, has lately set a aim to beef up foundry gross sales from high-performance computing (HPC) chip orders to 32% in 2028 from 19% in 2023 and automotive chip orders to 14% from 11%.
Over the identical interval, it’s going to scale back cellular chip foundry gross sales to the low 30 % vary from this yr’s estimated 54%.
With diversification of its chip foundry orders, the Korean chip large is predicted to enhance profitability with high-value added chip manufacturing.
Each HPC and automotive chips are thought of AI chips.
For foundry order diversification, the corporate plans to double the variety of non-Samsung clients over the identical interval.
With cellular chip foundry orders making up greater than half of its whole foundry gross sales, Samsung Electronics depends closely on its in-house chip builders and associates, comparable to Samsung System LSI that develops Samsung’s software processor Exynos and picture sensor ISOCELL for smartphones.
MORE TECH FIRMS OPT FOR OWN AI CHIPS
Samsung Electronics is assured about its foundry enterprise because it has lately seen an increase in AI chip foundry orders, comparable to graphics processing items (GPU) and central processing items (CPU) for AI servers and knowledge facilities.
There are rumors that the Korean chip large has already secured a world HPC main as its new foundry buyer.
In response to overseas media, CPU main Superior Micro Gadgets Inc. (AMD) is reportedly reviewing commissioning Samsung to provide its next-generation chips with its 4-nanometer class node.
As Samsung and AMD in April prolonged their strategic partnership in growing next-generation AP chips, the Korean chip maker may win the order, some analysts forecast.
Samsung, which nonetheless lags far behind international chief Taiwan Semiconductor Manufacturing Firm (TSMC) Ltd. in foundry orders, can be stated to have superior the manufacturing yield of its 4-nm chips to the extent of TSMC.
The Korean firm can be anticipated to profit from the latest transfer by Google, Microsoft Corp. and Amazon.com Inc. which might be speeding to develop their very own AI chips. As they’re fabless, they should order their chip manufacturing to foundry firms, like Samsung and TSMC.
TSMC at the moment produces Microsoft’s AI chips, utilizing its 5-nm node. Nonetheless, Samsung may win foundry orders from the US tech large later, contemplating that fabless chip firms can have the higher hand in worth negotiations with a number of foundry gamers.
“(Samsung’s) status and functionality are bettering (in foundry),” Jeong Ki-bong, vice chairman of Foundry Enterprise at Samsung Electronics, stated throughout the firm’s third-quarter earnings convention name on Oct. 31, anticipating “sturdy annual progress.”
Samsung’s foundry enterprise recorded its highest-ever quarterly orders final quarter because of a flood of orders from AI chip clients.
BEYOND 4-NM PROCESS TECHNOLOGY
The important thing to the regular progress in Samsung’s foundry orders is its nanometer chip processing know-how, which is a should to provide high-performance, low-power and high-efficiency AI chips.
Samsung plans to advance its 3 and smaller nanometer course of applied sciences to woo extra AI chip clients.
It pins excessive hopes on its next-generation gate-all-around (GAA) structure, which is predicted to enhance chip efficiency and power effectivity considerably.
“The GAA has gained nice consideration from the HPC business, and we’ll plan its manufacturing relying on demand,” stated Jeong.
It additionally plans to provide automotive and HPC chips, utilizing the 2-nm course of node beginning in 2026 and unveil the 1.4-nm node, thought of a dream know-how, in 2027.
Write to Jeong-Soo Hwang at hjs@hankyung.com
Sookyung Search engine optimization edited this text.
