Samsung Electronics has raised prices on new orders across its most advanced foundry processes by as much as 15%, according to two people familiar with the matter, as surging demand for AI chips eats up available manufacturing capacity. The increases, applied in July across the 4nm, 5nm, and 8nm nodes, are the clearest sign yet that the South Korean giant is finally getting pricing power in the contract chip business it has been losing money on since 2022.
Chinese chip designers are absorbing the biggest hikes. U.S. export controls on advanced chipmaking tools have cut them off from domestic leading-edge production, so they are lining up at overseas fabs willing to take their wafers — and paying premium prices for the privilege.
Chinese Customers Accept the Steepest Increases
Customers in China and the United States saw quotes for Samsung's 4nm SF4 process climb 10% to 15% from June, while Taiwanese customers faced smaller increases of 5% to 10%, one of the sources said. Wafers on the 5nm SF5 node rose 10% to 15%, and the older 8nm process went up by close to 10%. Samsung declined to comment, citing its policy of not discussing operational details.
The 4nm line at Samsung's Pyeongtaek complex has reportedly been running at full capacity since late last year. It builds logic chips for Qualcomm alongside the base dies that sit underneath Samsung's own high-bandwidth memory stacks, which means outside customers are competing for wafer starts with the company's memory division — the part of the business that has been printing record profits through the AI boom.
Chinese orders now exceed what Samsung can take on. U.S. customers get served first, and a slice of capacity stays reserved for Samsung's own silicon. That captive-demand situation is not new. The Financial Times reported last year that Samsung's chip exports to China grew 54% between 2023 and 2024, including a deal that supplied Baidu's Kunlun division with more than three years' worth of logic dies for AI accelerators.
TSMC's Crowded Order Books Give Samsung Room to Price
Counterpoint puts Samsung at roughly 7% of global foundry revenue in the first quarter of 2026, against more than 70% for TSMC. That gap is doing Samsung a favor on pricing. TSMC's leading-edge capacity is booked out by AI orders, and the market leader has notified customers of 5% to 10% increases across all sub-5nm nodes starting in January, with some services reportedly rising around 25% in 2027.
Samsung is raising prices beneath that umbrella and still landing below where TSMC's quotes are headed. The company also cut its 2nm wafer price to $20,000 earlier this year in a reported attempt to undercut TSMC by roughly a third, giving customers another reason to qualify its fabs as a second source.
Lee Min-hee, an analyst at BNK Investment & Securities, told Reuters the pricing pressure is a downstream effect of TSMC's constraints. As customers shift work to rivals such as Samsung and Intel, Samsung has room to raise its own prices. Lee said that if the company keeps pushing prices up, its foundry business could potentially become profitable as early as next year — earlier than previously expected.
A Foundry Turnaround Is Taking Shape
The foundry division has been in the red every year since 2022, and the gap with TSMC remains wide. But the customer list behind that forecast has grown over the past year. It now includes Tesla's $16.5 billion AI chip contract, an Apple manufacturing deal, a Broadcom AI chip agreement, and Nvidia's inference processor. Google is also in talks with Samsung to manufacture chips using the SF4 process.
Samsung told investors in July that the division was on track to return to profitability, pointing to improved fab utilization, yield gains, and stronger average selling prices as the main reasons. The company expects advanced manufacturing processes to generate more than half of its foundry revenue this year, with AI and high-performance computing workloads representing more than 30% of the total — roughly double their share at the end of 2025.
The pricing shift is part of a wider squeeze across the semiconductor supply chain. TSMC, Samsung, and other manufacturers are all pushing wafer prices higher as AI demand outstrips supply, and the cost is flowing through to chip buyers everywhere. For Samsung, though, the arithmetic has an extra layer: every wafer sold to an outside customer at a higher price is also a step toward ending a five-year losing streak in the business.
The context inside Samsung matters too. The memory division, led by HBM sales to Nvidia and other AI hardware makers, has been the profit engine through the AI cycle, and it wants wafer starts at Pyeongtaek just as badly as any external customer does. That internal competition for capacity is one reason Samsung can raise prices without worrying about idle tools: even if an outside order walks away, the line stays full. Analysts tracking the company's quarterly reports note that foundry utilization has climbed steadily through 2026 as AI chip orders from Tesla, Apple, Broadcom, and Nvidia have piled up alongside the existing Qualcomm and Google discussions.
Samsung's approach to pricing has also shifted in tone. Earlier in the year it cut the 2nm wafer price to $20,000, a move widely read as an aggressive bid to pull designs away from TSMC before its own next-generation capacity comes online. Raising prices on mature and leading-edge nodes while keeping 2nm cheap looks contradictory on the surface, but the strategy has a logic: charge what the market will bear for scarce capacity today, and use the 2nm price anchor to lock in design wins that will fill the fabs of tomorrow. Whether that mix works will show up in the division's financials over the next two quarters.
What the Hikes Mean for Chip Buyers
Higher foundry prices do not stay inside the fab. They show up in the cost of the processors, memory controllers, and networking silicon that companies buy for AI servers, and eventually in the price of the systems themselves. Design houses that committed to fixed-price contracts early are better placed than those buying capacity quarter to quarter.
Chinese customers accepting the largest increases shows how export controls have reshaped the market. Cut off from advanced tools, Chinese chip companies now depend on overseas foundries for leading-edge wafers, and that dependence is being priced into every order. Samsung's position between U.S. clients, its own products, and Chinese demand puts it in an unusual spot: it cannot serve everyone, so it can charge more for the capacity it does allocate.
The memory side of Samsung's business adds another twist. Because the foundry shares Pyeongtaek's cleanrooms with HBM production, wafer allocation decisions are partly decisions about which product line earns the most. With memory prices still elevated, every advanced wafer that goes to an external foundry customer carries an opportunity cost that Samsung's pricing team is now recovering.
A Longer View on Pricing Power
Foundry customers are notoriously quick to move designs between suppliers when prices climb, and TSMC's next-generation capacity, once it comes online, will add supply to a market that is currently starved of it. Samsung's own roadmap depends on winning enough leading-edge orders to keep its advanced fabs busy and fund the transition to 2nm-class production, where it has already shown willingness to price aggressively.
The broader foundry race has been a running theme in recent coverage on this site. China's SMIC posted a profit surge as AI demand fueled the foundry race, and Intel's CEO put $10 million of his own money into the foundry turnaround — two signs that the competitive picture is shifting after years of TSMC dominance. Samsung's price hikes fit the same pattern: when the industry's biggest supplier runs out of capacity, the second-tier players get their turn at the pricing table.
For now, Samsung has something it has not had in years: a waiting list. Whether that lasts through 2027 depends on how much new capacity TSMC brings online and how quickly customers start qualifying alternatives. The people familiar with the matter said the July increases are not a one-off; further adjustments are possible if demand keeps running ahead of supply.
Related coverage: Semiconductors · SMIC posts profit surge as AI demand fuels the foundry race · Intel's CEO backs the foundry turnaround with his own money · Nokia buys NXP's Arizona fab for indium phosphide chips
Sources: Reuters · Tom's Hardware · Quartz