For four years, Samsung Electronics' contract chipmaking arm has been the semiconductor industry's most expensive cautionary tale: a distant second place that kept losing money while TSMC banked the profits of the AI boom. On Wednesday, Reuters reported the clearest evidence yet that the story is turning. Samsung has raised prices on new foundry orders by as much as 15%, and customers are paying it.
The increases took effect on July orders, according to two people familiar with the matter who spoke to Reuters on condition of anonymity. Quotes for chips built on Samsung's 4-nanometre SF4 process rose 10% to 15% from June for customers in China and the United States, while customers in Taiwan, TSMC's home market, saw smaller increases of 5% to 10%. Wafers on the 5-nanometre SF5 node also climbed 10% to 15%, and even the ageing 8-nanometre line went up by close to 10%. Samsung declined to comment, saying it does not discuss operational matters.
The geography of those numbers is the story within the story. Chinese chip designers, cut off from advanced chipmaking equipment by U.S. export controls, have few alternatives for leading-edge logic, and they are absorbing the steepest hikes. Chinese demand is strong enough that Samsung cannot fill all of it: U.S. customers are served first, and part of the capacity stays reserved for Samsung's own chip production, the sources said. Scarcity is being rationed by policy as much as by price.
Capacity is genuinely tight. Samsung's SF4 line at Pyeongtaek has been running at full capacity since late last year, a person familiar with the company's operations told Reuters. It does double duty, producing logic chips for customers including Qualcomm alongside the base dies beneath Samsung's own multi-layer high-bandwidth memory stacks. External customers are therefore bidding for wafer starts against the memory division that just delivered Samsung's record profits, exactly the kind of internal competition that gives a sales team room to quote higher.
None of this makes Samsung a peer of the market leader. Counterpoint Research put Samsung at 7% of global foundry revenue in the first quarter of 2026, against more than 70% for TSMC. But the gap is precisely what creates the opening. AI orders have booked out TSMC's leading-edge capacity, and the Taiwanese firm has already notified customers of increases across its sub-5nm nodes starting in January, with some services reportedly rising by as much as 25% in 2027. Samsung is raising prices under that umbrella and still landing below where TSMC's quotes are headed. Earlier this year the Korean firm cut its 2nm wafer price to roughly $20,000, about a third below TSMC's reported $30,000.
Samsung's own guidance points the same way. The company expects advanced processes to account for more than half of foundry revenue this year, with AI and high-performance-computing applications making up more than 30%, up from 15% to 20% in late 2025. In July it said the unit should return to profit in the near future on higher factory utilisation, better yields and firmer pricing, and that second-half foundry revenue should rise by more than double-digit percentage points from a year earlier.
"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," said Lee Min-hee, a Seoul-based analyst at BNK Investment & Securities. "If Samsung raises prices from here, its foundry business could potentially become profitable as early as next year, earlier than previously expected."
Why It Matters
Price increases are the most reliable tell in a capital-intensive business. Samsung spent three years buying share with discounts, most visibly on 2nm, because it had capacity to fill and a credibility problem on yields. Charging more for established nodes like SF4 and SF5, and getting it, means the discount era is ending for reasons that have little to do with Samsung's roadmap and everything to do with TSMC being sold out.
The consequences run beyond Suwon. For AI chip designers, the second source is no longer automatically the cheap option, and cost inflation is spreading from the leading edge down into 8nm, a node old enough that nobody budgeted for a 10% hike on it. For Samsung, a unit loss-making since 2022 by industry estimates gets a path to break-even that does not require first winning the technology race. And for Washington, the report is an awkward data point: export controls meant to slow Chinese AI development are, at the margin, subsidising a Korean foundry's pricing, with Chinese buyers paying the largest premiums for capacity they cannot get at home.
The risk is symmetry. Pricing power borrowed from a rival's constraints evaporates when those constraints ease. TSMC is ramping N2 aggressively, and Intel Foundry is chasing the same overflow demand. Samsung's roster now carries real weight, spanning Tesla's $16.5 billion AI chip contract, deals with Apple and Broadcom, and Nvidia's inference processor. But a roster is not a structural advantage; the test is whether those wins survive the moment leading-edge supply loosens.
What To Watch
Samsung's third-quarter results, due in late October, are the first hard test of whether July's increases show up as margin rather than mix. Watch Google's reported talks to build chips on SF4, which would validate the node as more than a captive HBM line. Watch the Taylor, Texas fab, where the first 2nm output is expected around year-end and where discount pricing still applies, an odd split between a node Samsung is cutting and nodes it is marking up. And watch TSMC's 2027 quote sheet: if the reported 25% increases hold, Samsung's ceiling rises with them, and this week's hike will look less like a peak than a first move.
“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.”— Lee Min-hee, Analyst, BNK Investment & Securities