Why Samsung Stock Jumped After Raising Foundry Prices by Up to 15%

Samsung’s stock jump to 271,000 won, reported as nearly 9.5%, was less about the absolute size of a single price increase than what it implied about the foundry market: Samsung appeared able to raise prices on new orders because customers needed scarce AI-chip capacity. That suggested improved pricing power after years in which the company was widely viewed as competing from behind its larger rival, TSMC.

The signal is encouraging, but not conclusive. The reported prices were relayed by people familiar with the matter, and Samsung did not publish customer contracts or an official wafer-price schedule.

How much did Samsung raise foundry prices?

The reported July increases were for new orders and varied by process and region:

• 4 nm SF4: prices rose 10%–15% month over month for customers in China and the United States.
• 4 nm SF4 in Taiwan: prices rose 5%–10%.
• 5 nm SF5: prices rose 10%–15%.
• 8 nm: prices increased by nearly 10%.

The available reporting does not provide a separate regional breakdown for SF5 or 8 nm. It also does not establish that existing contracts were repriced; the reported changes concerned new orders.

Why the price increase mattered to investors

Foundries typically compete for customers by offering a combination of process technology, yield, capacity, delivery reliability and price. A supplier raising prices is therefore a useful market signal: it may mean that customers are willing to pay more because alternative capacity is difficult to secure.

That is the interpretation investors appear to have applied to Samsung’s move. Higher quotes could lift revenue per wafer, while fuller production lines could improve fixed-cost absorption. Together, those effects could support foundry economics even before Samsung demonstrates a sustained improvement in segment profitability.

The important distinction is that a price increase shows scarcity and bargaining power, not necessarily durable earnings. Customer mix, manufacturing yields, contract terms and the cost of adding capacity will determine how much of the increase reaches Samsung’s bottom line.

Why TSMC’s capacity squeeze creates an opening for Samsung

TSMC’s July earnings update provided evidence that demand for advanced semiconductor production remained strong. The company guided for third-quarter revenue of $44.6 billion to $45.8 billion, representing approximately 12% sequential growth and 37% year-over-year growth at the midpoint.

That outlook does not prove that every AI-chip customer is supply-constrained, but it is consistent with continued spending on data-center infrastructure. When leading-edge capacity is committed well in advance, customers may consider another foundry for designs that can be qualified on a different process—or pay more to secure an available slot.

For Samsung, the opportunity is most immediate in the advanced but established nodes represented by SF4 and SF5. Moving a chip to another process is not frictionless: designs must be qualified, software and packaging may need to be adjusted, and performance or power targets must be rechecked. But when the alternative is a long delay, a second source can become more valuable.

This is why Samsung’s reported ability to lift prices matters strategically. It suggests that at least some customers are valuing access to its capacity enough to accept higher quotes, rather than treating Samsung as a supplier that must win primarily through discounts.

Why Chinese customers reportedly faced the steepest SF4 increase

The reported regional difference points to an allocation problem as much as a pricing problem. Chinese demand was described as particularly strong, while Samsung was reportedly prioritizing some U.S. customers and reserving capacity for internal products.

Chinese chip designers may also have fewer practical alternatives for advanced production because of restrictions on access to certain semiconductor manufacturing technologies. That can make available Samsung capacity more valuable and reduce a customer’s ability to switch quickly to another supplier.

Still, the evidence does not show that geography alone determined the final price. Order size, product type, delivery timing, contract duration and allocation priority could also explain why quotes differed. The strongest supported conclusion is that Chinese buyers were reportedly willing to absorb the largest SF4 increases—not that every Chinese customer paid the same rate.

What the AI-demand indicators show—and what they do not

TSMC’s revenue guidance is a forward-looking indicator that AI-related semiconductor demand remained robust when the guidance was issued. The broader memory market provides another sign of strong infrastructure spending: Counterpoint Research forecast global memory revenue at roughly 1,500 trillion won in 2026, compared with about 360 trillion won in 2025.

Memory is a major part of an AI system’s cost. A data-center server requires more than a logic accelerator: it also needs high-bandwidth memory, server DRAM, storage, advanced packaging, networking, power delivery and cooling. If memory prices rise alongside foundry prices, the cost of building and operating AI capacity rises as well.

That creates a mixed signal for investors. Strong demand can improve semiconductor revenue and margins, but higher component costs can make data-center expansion more expensive. Customers may continue spending if the expected returns from AI services justify the investment; otherwise, rising hardware costs could eventually slow deployment.

Nvidia’s upcoming results are therefore an important confirmation point, not evidence already available. Strong data-center revenue, orders and guidance would support the view that demand remains tight. A weaker outlook, slower customer capital spending or signs of rising inventories would weaken that thesis.

The bullwhip risk behind the shortage

A supply shortage can reflect genuine end demand and still become exaggerated by precautionary ordering. Customers facing long lead times may order earlier, reserve capacity with multiple suppliers or purchase more components than they immediately need.

That behavior can amplify the cycle:

1. AI infrastructure demand rises and absorbs available capacity.
2. Customers place advance or duplicate orders to protect future supply.
3. Suppliers interpret those orders as evidence of even stronger demand and raise prices or expand production.
4. If end demand later normalizes, customers cancel, defer or consume existing inventory more slowly.
5. Utilization and chip prices fall, putting pressure on earnings and valuations.

Counterpoint’s separate memory research has already described customers pulling orders forward and raised its third-quarter DRAM price forecast to 10%–20% quarter over quarter. That does not establish that an inventory correction is imminent, but it illustrates why current demand data must be separated from orders placed defensively.

The bottom line for Samsung

The reported foundry price increases gave investors a concrete reason to believe Samsung’s advanced manufacturing capacity had become more valuable. SF4 prices reportedly rose as much as 15% for new orders in China and the United States, SF5 rose as much as 15%, and 8 nm increased by nearly 10%.

The rally was therefore a bet on improving pricing power, utilization and strategic relevance—not proof that Samsung has already solved its foundry challenges. TSMC’s strong outlook and the rapid expansion of the memory market support the near-term AI-demand case, while pull-forward orders leave open the possibility of a later inventory correction.

For Samsung, the key test is whether higher prices persist alongside better yields, reliable delivery and repeat customer demand. If they do, the July increases could mark more than a one-off pricing event: they could signal a more competitive role for Samsung in the AI supply chain.