Samsung Foundry has raised prices on new orders across its 4nm, 5nm and 8nm contract chipmaking processes by as much as 15%, according to a Reuters report citing two people familiar with the matter, corroborated this week by Tom's Hardware, Yahoo Finance and South Korea's MK Business News. The increases took effect with orders placed in July, marking Samsung's first broad round of foundry price hikes after years of discounting to win business away from TSMC.
The steepest adjustments hit Samsung's 4nm SF4 line. Customers in China and the United States saw quotes climb 10% to 15% from June, while Taiwanese customers were offered a smaller increase of 5% to 10%, one source told Reuters. Wafers built on the 5nm SF5 process rose by a comparable 10% to 15%, and even the older 8nm node — a workhorse for automotive and lower-power chips — climbed by close to 10%.
A supply squeeze years in the making
Samsung's 4nm line at its Pyeongtaek complex in South Korea has reportedly been running at full utilization since late 2025. That capacity crunch is the direct trigger for the new pricing, and it lands at a moment when Samsung's foundry division needed the relief: the unit had been losing money for years while it chased market share with aggressive discounts, and it has reportedly reached monthly profitability for the first time in three years as utilization and pricing power improved together.
Chinese chip designers are reportedly accepting the steepest increases of the group. Cut off from the most advanced chipmaking tools by US export controls, many have limited alternatives to Samsung or TSMC for anything beyond mature nodes, which leaves them with little room to negotiate. Some of that demand is spillover: TSMC's leading-edge lines are booked out by AI orders, and industry reports describe overflow business — particularly from AI accelerator designers and Chinese customers — moving toward Samsung simply because there is nowhere else with open capacity.
Samsung's customer list for the 4nm line reportedly includes Qualcomm, alongside expanding orders tied to AI accelerators and HBM-adjacent logic chips. That mix matters for how the increase lands: Qualcomm and similar fabless customers design flagship mobile and PC chips years in advance and cannot easily shift a committed design to a different foundry mid-cycle, which gives Samsung room to raise prices without an immediate exodus. Switching a qualified design to TSMC or a different node takes quarters of requalification work, not weeks.
Undercutting TSMC, then catching up to it
The new pricing follows a period in which Samsung tried the opposite approach. Earlier this year, the company reportedly cut its 2nm wafer price to around $20,000, roughly a third below TSMC's quotes, in an attempt to pull advanced-node customers away from the market leader. That discount strategy reflected Samsung's weak negotiating position: Counterpoint estimated Samsung's share of global foundry revenue at roughly 7% in the first quarter of 2026, against more than 70% for TSMC.
TSMC, for its part, had already moved in the other direction. The company notified customers in January of 5% to 10% price increases across all of its sub-5nm nodes, with some services reportedly set to rise around 25% in 2027 as its most advanced capacity remains booked solid by AI customers. Samsung's latest round of hikes lands beneath that ceiling — its prices are rising, but they still sit below where TSMC's quotes are headed, preserving at least part of the discount that has been Samsung's main selling point against the market leader.
- SF4 (4nm): quotes up 10%–15% for China and US customers, 5%–10% for Taiwan
- SF5 (5nm): wafer prices up 10%–15%
- 8nm: prices up close to 10%, affecting automotive and lower-power orders among others
What the pricing shift signals for Samsung's foundry business
Samsung is positioning the increases as part of a broader pivot in its foundry strategy. The company expects nodes at 5nm and below to account for more than half of its foundry revenue this year, with AI and high-performance computing applications potentially making up more than 30% of that advanced-node business on their own. Full utilization at Pyeongtaek's 4nm line, paired with the first monthly profit in three years, gives Samsung a case for holding firmer prices going forward rather than returning to the discount strategy that defined its foundry business for much of the past decade.
Whether that pricing power survives past this cycle depends heavily on execution at 2nm and beyond, where Samsung has trailed TSMC on yield. A customer accepting a 15% price increase on SF4 today is doing so because there is no cheaper seat available anywhere else — not because Samsung has closed the technology gap. If TSMC's A16 process, targeted for mass production in the fourth quarter of this year, delivers on its promised 8%–10% performance gain over N2P, the pressure on Samsung to keep proving itself at the leading edge only grows.
Packaging and components are tightening at the same time
The foundry price increases are not happening in isolation. TSMC's own advanced packaging lines, which turn finished wafers into the stacked chips AI accelerators require, have become a separate bottleneck this year, with CoWoS capacity constraints pushing customers toward newer specialization and capacity-sharing arrangements across the packaging supply chain. A wafer that clears Samsung's or TSMC's fabs still has to wait in line for packaging capacity before it becomes a finished chip.
High-end multilayer ceramic capacitors, a component present in nearly every AI server board, are seeing a similar squeeze. Lead times on high-capacitance MLCCs have stretched to five to ten months, against 14 to 18 weeks for standard parts, and some Samsung Electro-Mechanics products have reportedly reached roughly 40 weeks. Replenishment for certain high-end MLCC lines is not expected until late 2026 or January 2027. None of this is Samsung Foundry's doing directly, but it illustrates why chip designers have little leverage to push back on a wafer price increase: even if they found cheaper capacity elsewhere, the components needed to finish the board are just as constrained.
The wider East Asia foundry market
The Samsung increases are part of a broader repricing across East Asia's chip manufacturing base this year. Contract chipmakers up and down the value chain — from advanced-node leaders to mature 8-inch fabs serving automotive and industrial customers — have been raising prices as AI-related demand consumes capacity that used to go begging. STMicroelectronics, for instance, is preparing its third 2026 price increase, effective in late August, after earlier rounds in March and May pushed costs higher on automotive microcontrollers, power ICs and NFC chips; some automotive MCUs have already risen 15% to 20%, and lead times on certain power semiconductors have stretched past a year, with some reportedly reaching 52 weeks.
Memory suppliers have followed a similar arc. Kioxia raised NAND flash prices earlier this month as AI data center demand strained supply, and SK Hynix and Samsung have both been racing to expand HBM4 output for AI accelerator customers. The common thread across foundries, memory makers and passive-component suppliers is the same: a market that spent years competing on discounts has flipped into one where capacity, not price, decides who gets served first.
For chip designers in China, Taiwan and the US placing orders for late-2026 and 2027 delivery, that flip means budgeting for foundry costs that no longer move in one direction. Samsung's willingness to raise prices even on an 8nm line — a node with no shortage of alternative suppliers in China alone — suggests the company sees the current demand environment as durable rather than a short-term spike it needs to ride out quietly. Samsung declined to comment on the pricing when contacted by Reuters.