Samsung raises chip prices as AI demand tightens capacity.AI demand is tightening capacity across supporting chips.

 

Samsung Electronics has raised prices for some advanced chipmaking services as AI-related orders increase demand for its foundry capacity.

Prices for new orders using Samsung’s 4-nanometre process increased by as much as 15% in July, Reuters reported, citing people familiar with the matter. Samsung declined to comment on the pricing, saying it does not disclose details about operational matters.

According to the report, customers in China and the US faced increases of 10% to 15% for Samsung’s SF4 process, while Taiwanese customers saw increases of 5% to 10%. Prices for its 5nm process also rose 10% to 15%, while its 8nm process increased by close to 10%.

The July increases follow earlier changes in advanced-node pricing. TrendForce reported in March that Samsung had informed customers in the fourth quarter of 2025 that it would increase prices for its 5nm and 4nm foundry services.

The research firm expects global foundry revenue to increase 24.8% year over year to about $218.8 billion in 2026, with AI GPUs and custom processors driving demand for 5nm, 4nm, and more advanced processes. It also expects TSMC’s capacity at those nodes to remain fully utilised through the end of the year.

Advanced-node capacity tightens

Samsung’s 4nm capacity is being used for both external foundry orders and its own HBM production. Reuters, citing a person familiar with Samsung’s operations, reported that its Pyeongtaek SF4 line produces logic chips for customers, including Qualcomm as well as HBM base dies for Samsung.

The company said its foundry performance improved in the second quarter before incentive-related provisions, supported by demand for high-bandwidth memory base dies and orders from US customers.

Samsung plans to increase sales of 4nm Language Processing Unit chips and HBM base dies during the second half of 2026. It is also ramping production using its second-generation 2nm process as it adds AI and high-performance computing design wins.

Samsung’s HBM4 uses a logic base die manufactured on the company’s 4nm process, linking demand for advanced memory directly with foundry capacity. The company began commercial HBM4 production earlier this year and is also using 4nm technology for the base die in HBM4E.

Samsung plans to increase 4nm HBM base-die sales in the second half, alongside other logic products manufactured for external foundry customers.

Samsung said it expects foundry revenue to record double-digit year-over-year growth in the second half of 2026. It cited demand across process nodes from customers in the US and China, alongside HBM base dies and additional AI and high-performance computing projects.

TSMC’s second-quarter results provide another measure of where foundry demand is concentrated. Processes at 7nm and below accounted for 77% of its wafer revenue during the quarter, including 33% from 5nm, 30% from 3nm, 11% from 7nm, and 3% from its newly ramping 2nm technology.

High-performance computing accounted for 66% of TSMC’s second-quarter revenue and increased 20% from the previous quarter. TSMC chief executive C.C. Wei said AI-related demand remained strong and that the company was increasing capital expenditure and manufacturing capacity to support customers’ production plans.

Wei said AI infrastructure is also increasing demand for processors beyond accelerators, including CPUs used in data centres. Separate foundry market research has also identified custom processors from cloud providers and AI companies as a source of advanced-node demand.

Samsung remains considerably smaller than TSMC in contract chip manufacturing. TrendForce estimated that Samsung Foundry generated slightly more than $3.2 billion in revenue in the first quarter of 2026, giving it 6.5% of the global market, compared with 72% for TSMC.

Separately, TrendForce News, citing Chosun Biz, reported that overall utilisation at Samsung Foundry had recovered to around 70%–80%. The report said utilisation at key 4nm and 5nm nodes had fallen below 50% in 2025; Samsung has not disclosed utilisation figures for those lines.

AI demand reaches mature nodes

Foundry capacity pressure is also appearing in older manufacturing processes used for components that support AI servers. These include power-management chips, power devices, interposers, programmable logic, and optical communications components.

TrendForce said foundries are allocating more wafer capacity to AI-related products, including power-management ICs, power discrete devices, interposers, photonic integrated circuits, and optical communications components. Some manufacturers are reducing capacity allocated to lower-margin products such as image sensors and display-driver ICs.

The research firm said average utilisation across the world’s 10 largest foundries’ 8-inch fabs reached 88% in 2026 and is expected to reach 90% in the second half. Average 8-inch foundry prices increased by between 5% and 15% from the first quarter to the second.

Global 8-inch foundry capacity is expected to decline 2.4% year over year in 2026 as companies including TSMC and Samsung scale back older production lines. Power-management ICs and power discrete devices remain heavily dependent on 8-inch manufacturing, while demand from AI servers and other computing systems is raising utilisation.

The same research also found higher AI-related wafer consumption in parts of the mature 12-inch market, including power devices, silicon interposers, and FPGAs. Capacity is tighter in selected processes, while utilisation remains uneven across other mature-node products.

Longer lead times are appearing for some power-management and server-management chips as suppliers allocate more manufacturing capacity to AI-related orders.

In August, TrendForce raised its forecast for 2026 AI server shipment growth from 28% to nearly 31% and projected that combined capital expenditure among nine major cloud service providers would rise about 90% year over year. The firm linked the spending with continued demand for AI servers and related semiconductor components.

Samsung expands foundry orders

Samsung is also expanding its customer base for manufacturing processes beyond the nodes covered by the reported July price increases.

The company signed a memorandum of understanding with Broadcom in July covering memory, foundry services, and advanced packaging. The companies estimate their planned collaboration across memory and foundry technologies at more than $200 billion over the five years through 2030.

The foundry portion covers Samsung’s 2nm and smaller process technologies, along with 2.3D and 2.5D advanced packaging. Broadcom also plans to work with Samsung on HBM for future AI accelerators.

Samsung has also secured a 22.8 trillion won semiconductor manufacturing agreement running through the end of 2033. Tesla chief executive Elon Musk said Samsung’s Taylor, Texas, plant would manufacture Tesla’s next-generation AI6 processor.

 

 

 

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

Tech Wire Asia is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.