MMarket PulseScope요청 / 문의
All market news
SEMICONDUCTORS

DB HiTek Lifts Foundry Prices by Up to 30% as Chinese Orders Fill 200mm Capacity

South Korean foundry DB HiTek has raised prices by as much as 30%, with Chinese customers facing the steepest increases as orders push 8-inch capacity toward full utilization. TSMC is meanwhile weighing a multi-fab Texas campus that could add tens of billions of dollars to its US expansion.

DB HiTek has raised foundry prices by as much as 30%, with Chinese customers facing the steepest increases as rising orders push the South Korean foundry's 8-inch capacity close to full utilization, according to The Elec. The scale of the increase is unusual for mature-node capacity, which had been in oversupply through 2024 and much of 2025, and it points to a genuine tightening at 200mm rather than a negotiated headline number. Chinese fabless customers, which have been consolidating orders into domestic and regional foundries, appear to be absorbing the steepest part of the increase.

Leading-edge capacity is tightening for the same reasons. TSMC is reportedly weighing a new manufacturing campus in Texas that could add tens of billions of dollars to its US expansion, potentially creating a second major American production hub alongside Arizona as AI chip demand continues to strain advanced-node capacity. Singapore is also in talks. The buildout matters for pricing because advanced-node wafer allocations are effectively rationed: customers with AI accelerator or AI server ASIC programmes are competing for the same limited leading-edge output, which pushes spill-over demand down into mature nodes and lifts pricing there.

Advanced packaging has become the second bottleneck. Data center AI accelerator package power has climbed from about 500W in the A100 generation to roughly 3,500 to 4,000W for Rubin Ultra-class products expected in 2027, pushing supply current into the thousands of amperes. Intel Foundry estimates that raising a single compute package from 1kW to 5kW increases power delivery network I-squared-R losses from 89W to 2,222W, cutting the share of input power reaching the transistors from 66.0% to 41.8%. That is why Intel, TSMC and Samsung are all converging on backside power delivery, making packaging rather than lithography the next competitive divide.

China's domestic chain is scaling in parallel. Chinese OSAT providers are forecast to take 13% of the global cloud AI accelerator market in 2026 as custom ASICs from Huawei, Cambricon and Alibaba enter volume production, with JCET Group, Tongfu Microelectronics and SJ Semiconductor handling the assembly. JCET signed an agreement on September 29 for a high-end advanced packaging and testing project in Wuxi. Equipment localization is accelerating too, with ACM Research's order backlog up 88% as Chinese capacity investment continues.

Sources