The foundry division (chip contract manufacturing), the "aching finger" of Samsung Electronics DS (semiconductor) institutional sector, is seen as possibly swinging to a profit next year. On top of mass-producing Tesla's next-generation artificial intelligence (AI) chips announced last year, the odds have risen for orders from global big tech corporations such as Qualcomm and AMD. As the Production yield of advanced processes at 3 nm (nanometer, one-billionth of a meter) and below, which had long underperformed, is gradually stabilizing, the company is said to be tightening the reins in the global big tech order race, with the start of operations at the Taylor fab (plant) in the United States this year serving as a turning point.
According to the industry on the 26th, prospects are emerging that the Samsung Electronics foundry division, which has posted multi-trillion-won losses since 2022, will turn a profit next year. Kim Dong-Won, head of research at KB Securities, said, "Samsung Foundry's results are expected to swing from a 7 trillion won loss last year to a profit next year through higher utilization rates driven by expanded supply of Tesla AI chips." A person familiar with Samsung Foundry's situation also said, "We are speeding up orders with the goal of turning a profit next year. In addition to the stabilization of Production yield in advanced processes at 3 nm and below, utilization for the 4–8 nm nodes, which have high profitability, is also at its maximum level."
Expectations for Samsung Foundry's profit turnaround are tied to the timing of the Taylor fab's ramp-up. Samsung Electronics is building a large-scale foundry production base in Taylor, Texas, investing more than $37 billion (about 54 trillion won) by 2030, and plans to begin mass production in the second half of this year. It is in the final stages of dispatching engineers and bringing production lines online for pilot production in the first half. Samsung Foundry aims to win and mass-produce chips from U.S. big tech corporations.
If Samsung Foundry mass-produces Tesla AI chips through the Taylor fab, profitability is expected to improve. Samsung Foundry plans to mass-produce next year the AI5 and AI6, AI chips that serve as the core brain for Tesla's Humanoid Robot Optimus, robotaxi, and Autonomous Driving. In particular, the AI6 chip is said to be exclusively produced by Samsung Foundry.
The likelihood of winning orders for products from Qualcomm and AMD is also said to be rising. Qualcomm is a longtime customer of Samsung Foundry and is in a "special relationship," with the Samsung Electronics mobile business (MX) institutional sector as a client. However, it departed due to low Production yield in Samsung Foundry's advanced nodes. As the 2 nm process stabilizes, the chances of winning orders are growing. In addition, AMD, a longtime customer of TSMC, is reportedly pushing for dual-sourcing its foundry.
Another reason order prospects are improving is that TSMC's production lines, which account for 70% of the foundry market share, have reached saturation. By stably providing advanced processes at 3 nm and below, TSMC has drawn big tech corporations' production volumes. However, as utilization on its lines hits limits, prices have skyrocketed and the risk has grown that customers will not receive chips when they want them, prompting moves to consider Samsung Foundry as an alternative. Using Samsung Foundry's Taylor fab also appears to help align with the Trump administration's "Made in USA" stance, which has worked favorably.
Higher utilization in the 4–8 nm nodes also appears to have contributed to stronger profitability. The 4–8 nm nodes are mature processes with stabilized mass-production Production yield compared with advanced nodes, and are considered competitive with TSMC in performance and process pricing. Samsung Electronics uses these nodes to manufacture logic dies for high-bandwidth memory (HBM), as well as chips for IBM and Nintendo.