Samsung Electronics has raised supply prices for new customers by about 15% on its foundry 4nm and 5nm processes, as well as part of its automotive-grade 8nm process. Chosun Biz reported this on July 9, 2026, citing industry sources. The same outlet had also reported on July 2 that Samsung began prioritizing production capacity for existing customers while selectively accepting new orders. Samsung has not officially announced any price revision. However, the company's latest earnings materials point to increased production of HBM4 base dies using 4nm and AI processors, alongside an outlook for full utilization of advanced nodes in the second half of 2026. This official outlook aligns with the tightening of order conditions suggested by the two reports.
The ~15% Increase Is Limited to New Customers and Specific Nodes
According to Chosun Biz, the roughly 15% price increase applies to the 4nm and 5nm nodes, where demand is concentrated, as well as part of the automotive-grade 8nm process. It does not apply uniformly across all existing customers or all processes. The same outlet reported that TSMC has also informed major customers such as NVIDIA, Apple, and AMD of plans to raise wafer supply prices for 3nm, 5nm, and 7nm by 5–10%. Samsung's revision represents a larger percentage increase than TSMC's. However, it is limited to new customers and specific nodes.
This difference reflects the two companies' respective positions. TSMC applies price increases broadly across multiple advanced nodes. Samsung, by contrast, is revising prices starting with the processes where orders are concentrated. For new orders, it can set prices without altering existing contracts, making it easier to select customers based on profitability and production efficiency. The roughly 15% figure does not represent a market-wide rate across the foundry industry—it is the figure reportedly offered by Samsung for new orders on certain lines, at a higher unit price than before. Whether customers have accepted these terms has not been disclosed.
The inclusion of automotive-grade 8nm in the price increase is also notable. Samsung mass-produces chips ranging from 14nm to 4nm for automotive infotainment applications, and supplies 8nm and 5nm for ADAS and autonomous driving applications. This increase extends beyond advanced processes for AI servers into automotive processes that require long-term supply and reliability. However, no figures have been released regarding supply-demand conditions for 8nm or the rate of acceptance of the price increase.
HBM4 and AI-Focused LPUs Filling 4nm Capacity
Samsung's January–March 2026 earnings materials specifically explain why demand is concentrating on 4nm. The company outlined plans to increase mass production of memory products using 4nm and AI/HPC-focused LPUs during the April–June quarter. It expects advanced node lines to reach full utilization in the second half of 2026, with earnings improving as supply of HBM4 base dies increases.
The HBM4 base die is a logic chip that handles data transfer between stacked DRAM and AI accelerators. Samsung began initial shipments of the 4nm version in the October–December 2025 quarter, and 2026 marks the stage of expanding mass production. Large-scale chips for HBM and AI/HPC applications are now joining the 4nm line, which had previously been centered on smartphone SoCs. As demand sources multiply, the priority shifts from discounting to fill idle lines toward allocating limited production capacity among competing orders.
The 4nm node is not a newly established process. Samsung's SF4 is a FinFET-generation process that began mass production in 2021, and the company treats it as a mature platform with stable yields and process control. Even so, larger AI dies face a higher probability of encountering defects as die area increases, making uniformity and yield management more challenging. Samsung plans to begin mass production of its next-generation LPU in the second half of 2026. Whether raising prices will retain customers depends on the company's ability to consistently deliver good yields.
TSMC's 74% Share Builds a Price Umbrella
Looking at competitor figures, the groundwork for AI demand to push prices upward is clear. In TSMC's January–March 2026 quarter, 3nm accounted for 25% of wafer revenue, 5nm for 36%, and 7nm for 13%. Combined, nodes at 7nm and below reached 74% of wafer revenue. By application, HPC accounted for 61% of revenue, far surpassing smartphones at 26%.
Wafer shipments (12-inch equivalent) for the quarter totaled 4.174 million units, up 28.1% year-over-year. Gross margin rose 3.9 percentage points from the previous quarter to 66.2%. TSMC attributed this to cost improvements, high capacity utilization, and foreign exchange effects. The company has not announced the price increases themselves either. However, with advanced processes accounting for three-quarters of revenue and HPC as the largest application, this earnings structure supports an environment where suppliers can more easily maintain prices.
As TSMC secures high utilization and margins across a broad range, Samsung faces less need to compete purely on price as an alternative supplier. For customers, adding Samsung as a second production source—despite the costs of design porting and verification—becomes more meaningful than concentrating production with a single supplier. That said, migrating the same design to a different foundry requires process-specific redesign work, so customers won't shift orders based on price differences alone. For Samsung to sustain higher unit prices, TSMC's supply capacity must remain constrained, while Samsung keeps its own yields and delivery times at levels customers can accept.
Yields and Utilization Matter More Than the Price Increase Rate
Even if the wafer price customers pay rises by 15%, the cost of a finished GPU or smartphone does not necessarily increase by the same rate. The number of good dies obtainable from a single wafer depends on die size and yield, with additional costs layered on afterward for advanced packaging, HBM, substrates, and more. For large AI chips, the effective cost per good die—rather than the wafer price itself—more strongly influences ordering decisions. If yields improve after the price increase, some of the added burden on customers can be offset.
The same conditions apply to Samsung. Foundry earnings for the January–March 2026 quarter deteriorated due to seasonal factors. The company does not disclose foundry-specific revenue or operating profit separately. The substantial profit recorded by Samsung's semiconductor division cannot be directly interpreted as a foundry recovery. Moreover, the full-utilization outlook applies specifically to advanced nodes, not to overall utilization across the business including mature processes.
Whether the roughly 15% revision marks a turning point will become clearer in the second-half 2026 earnings results—specifically, whether 4nm HBM4 base dies and AI-focused LPUs grow as planned, and whether full utilization of advanced nodes translates into improved profits. Additionally, whether pricing for new customers extends to existing contracts will need to be confirmed. Only once both points are verified can it be concluded that Samsung Foundry has moved beyond a phase of chasing orders through discounting.