Samsung Foundry's new chief reveals plan to take on TSMC and keep SMIC away
The former coach of the New York Giants, Bill Parcells, once said "You are what your record says you are." Well, if you're Samsung Foundry that isn't good. TSMC is the world's leading foundry with a 65% market share and a star-studded client lineup that includes Apple, Nvidia, MediaTek, Qualcomm, and even Intel to name a few. While Samsung Foundry currently has a market share of approximately 18%, good enough for second place after TSMC, its lead over the third-place foundry has been slipping.
Despite its inability to obtain an extreme ultraviolet lithography machine (U.S. and Dutch sanctions, you know), China's SMIC is in third place and was just 3.3 percentage points behind Samsung Foundry during the third quarter of this year. That compares to the 5.8% lead Samsung Foundry held over SMIC during the second quarter of 2024.
Even though Samsung Foundry was the first to use Gate-all-around (GAA) transistors with its 3nm node (TSMC will start using GAA next year with its 2nm production), Samsung Foundry hasn't seen much demand for its services at 3nm. GAA transistors use vertically placed horizontal nanosheets allowing the gate to cover the channel on all four sides. This reduces current leakage and improves the drive current resulting in chips with enhanced performance and better energy efficiency.
Samsung Foundry has a two-part plan to help it battle TSMC and stay ahead of SMIC. | Image credit-Samsung Foundry
Samsung Foundry's main problem is its low yield which means not enough of the chips produced from silicon wafers are usable. This hikes the cost of the chips to clients since more wafers are required to produce the quantity of chips ordered. Samsung Foundry's 2nm yield has reportedly been in the 20%-30% area while TSMC's test runs at 2nm have been in the 60% range, closer to the 70% typically required to start mass production.
Samsung has promoted Han Jin-man to president of its company-wide foundry business from his executive role related to Samsung's U.S. chip facilities. He has a "two-track strategy" to help Samsung Foundry take on TSMC while holding off SMIC. Han's plan is to increase Samsung Foundry's yield on its 2nm production. That sounds easier said than done, and he says a higher yield will keep clients from leaving Samsung Foundry for TSMC. He also says that he will bring in more business for legacy nodes of 10nm and higher in order to make sure that the fabs are running closer to full capacity.
While it sounds like a plan, whether both parts can be accomplished is not known. And even if both parts can be done, it still isn't known whether this plan can make Samsung Foundry more of a threat to TSMC while allowing it to pull away from SMIC.
Things that are NOT allowed: