Samsung's July foundry price rise moves the AI crunch into your bill of materials

Samsung raised prices for some of its advanced contract chipmaking by as much as 15 percent on new orders, with the increases taking effect in July and centred on its 4-nanometre SF4 process, according to a Reuters report citing two people familiar with the matter ^[1][2]. The consequence for buyers is that the AI capacity shortage has stopped being a scheduling problem in TSMC's order book and become a unit-cost problem, because the second source is no longer the discount option: for years Samsung's foundry competed on price to win work TSMC could not or would not take ^[3].

The geography is the part worth reading twice. SF4 customers in China and the United States saw rises of 10 to 15 percent from the previous month, while customers in Taiwan, TSMC's home market, faced 5 to 10 percent ^[4][5]. SF5, Samsung's 5-nanometre node, went up 10 to 15 percent, and older 8-nanometre work by nearly 10 percent ^[6][7]. At the bottom of each band, a Taiwanese buyer absorbs half the increase a Chinese or American buyer does ^[8].

Samsung's sources describe demand from Chinese customers as especially strong, and say the company has not been able to fill every order because it also has to serve US customers and hold back capacity for its own chip production ^[9]. The reason Chinese buyers absorb the steepest rises is policy, not preference: US curbs on exports of advanced chipmaking equipment to China have pushed local firms towards foreign foundries such as Samsung, and with fewer leading-edge options they have less room to push back ^[10]. One of the sources said Chinese customers are among those accepting the largest increases even while paying more than buyers elsewhere ^[11]. Samsung declined to comment, saying it does not discuss operational matters ^[12].

This is a division that has lost money since 2022 on industry estimates ^[13], and that took 7 percent of global foundry revenue in the first quarter of 2026 against TSMC's more than 70 percent, according to Counterpoint ^[14]. TSMC's share is therefore at least ten times Samsung's ^[15]. Lee Min-hee of BNK Investment & Securities read the move as a knock-on effect: "As TSMC faces tight capacity and raises prices, customers are shifting to rivals such as Samsung and Intel, prompting Samsung to raise its prices as well" ^[16]. If the increases continue, he said, the foundry could turn profitable as early as next year, sooner than previously expected ^[17].

The physical constraint sits at Pyeongtaek, where the SF4 line has run at full capacity since late last year, making logic chips for customers including Qualcomm and the base dies for Samsung's own high-bandwidth memory ^[18]. That is the same silicon competing for the same tool time. Samsung said in July that it expected the foundry unit to return to profit soon on higher factory use, better yields and firmer pricing, and that second-half foundry revenue should rise by more than a double-digit percentage year on year ^[19]. It expects advanced processes to make up more than half of foundry revenue this year, with AI and high-performance computing above 30 percent, up from 15 to 20 percent in late 2025 ^[20] - an increase of between 1.5 and two times ^[21].

Watch whether the SF4 premium holds once TSMC's leading-edge capacity loosens, and whether the China surcharge widens or narrows as export controls shift. Watch Pyeongtaek allocation too: HBM base dies and external logic cannot both win.