Samsung wafer fab, price increase?

Samsung Electronics has for years been trying to address one of the biggest weaknesses in its semiconductor empire. It is one of the world’s largest manufacturers of memory chips, but its foundry business—producing chips for other companies—has long struggled to turn a profit.

Industry insiders estimate that Samsung has been operating at a loss since 2022. Although its financial situation has improved, its market position continues to decline. According to data from TrendForce, Samsung’s wafer foundry business grew 1.8% quarter-on-quarter from April to June, but as competitors expanded more rapidly, its global market share fell to 5.9%. TSMC’s market share, meanwhile, widened to 72.5%.

However, the situation that has kept Samsung in the red is beginning to change. Reuters reported in August that Samsung has raised prices by 15% on some new orders for its 4nm, 5nm, and 8nm processes (representing consecutive generations of its chip manufacturing technology). Its 4nm production line in Pyeongtaek is also now running at full capacity.

“The important thing isn’t the 15% price increase itself,” said Kim Rok-ho, Senior Analyst at Hana Securities, “but that Samsung can ask customers to pay more and still get the orders.”

Meanwhile, demand driven by artificial intelligence is causing TSMC’s advanced factories to operate at full throttle, with high-performance chip orders overflowing. TrendForce noted that TSMC’s 3nm and 5/4nm process capacities were fully booked in the second quarter, and reportedly, TSMC is preparing another price hike in 2027.

“The key issue isn’t that Samsung has suddenly gained market share,” said Lee Jong-hwan, Professor of Semiconductor Engineering at Sangmyung University, “but that customers are now asking not just which plant is best, but who else can actually manufacture this kind of chip when they need it.”

This matters because chip fabs are expensive to run regardless of capacity utilization rates. Even if a production line is underutilized, depreciation, labor, and maintenance costs persist. One of Samsung’s long-standing problems has been investing heavily in advanced capacity but not having enough customer orders to match demand.

Now, HBM substrate chips, artificial intelligence and high-performance computing chips, and overseas customer orders are taking up more capacity. Samsung said that after deducting incentive-related provisions, its “wafer foundry profits improved significantly in the second quarter," thanks to demand from U.S. customers and for HBM substrate chips (logic chips located under the high-bandwidth memory stacks used in AI systems).

Kim said: “The biggest change is that Samsung is finally starting to fully utilize the capacity it has already paid for. This lowers the cost per wafer, making higher prices more meaningful for profitability.”

The product mix is also changing. At an investor relations event held in Seoul this week, Samsung stated that high-performance computing (including powerful chips for artificial intelligence and data centers) now comprises 28% of the foundry’s revenue, up from just 5% in 2017.

Samsung Foundry executive Lu Mi-jeong said, “We expect high-performance computing to account for more than half of our revenue by 2029.”

However, not all fully loaded factories tell the same story. Orders from Samsung’s own memory business help cover costs but are far less significant than external chip designers choosing Samsung over TSMC.

The 15% price increase mentioned in reports does not apply to all of Samsung’s business. Contracts signed earlier may still be effective at lower prices. “As older contracts at lower prices are replaced by higher-priced orders, the benefits of the price hikes will gradually become evident,” Kim said.

Kim noted that higher factory utilization, more robust pricing, and a shift to higher-value chips mean Samsung is now closer to breaking even than it has been in years.

However, Samsung is also entering another costly phase. Its 2nm process is advancing rapidly, and its plant in Taylor, Texas will soon begin production, which will bring new engineering and operational costs before customers proceed to mass production. This means that some gains from the 4nm production line may be offset by the next round of investments.

There are many signs that customers are coming in. Samsung said in July that it has secured 2nm process projects from several major cloud computing and AI customers, and its long-term contract with Tesla has brought an important external customer to Taylor. Local media reported this month that trial production of Tesla’s AI chips has already started at the plant.

But winning a project does not equate to ensuring mass production. Qualcomm’s two new flagship Snapdragon chips are both made with TSMC’s 2nm process, indicating Samsung has not yet regained some of the industry’s most crucial advanced chip orders.

When capacity is tight, customers can’t simply shift the same chip from TSMC to Samsung. “Chips must be designed and tested for a specific foundry’s process,” Lee said. “So when customers truly start working with Samsung, it means they’ve invested time and engineering resources.”

In the long run, this supply shortage could change customers’ purchasing habits, offering Samsung a rare opportunity to turn a temporary capacity crunch into a long-term business. Companies that once depended almost entirely on TSMC are now looking for a second option.

Kim said, “There could be many reasons for the arrival of that first order. The real test is whether customers will come back and order the next generation of chips. That’s when it will show that Samsung’s products are good enough to win repeat next-generation orders.”