Samsung raised the price of its advanced chipmaking services in July, by 10 to 15 percent on the 4 nanometre SF4 and 5 nanometre SF5 processes and by nearly 10 percent on its older 8 nanometre node. Reuters reported the increases on Tuesday, August eighteenth, 2026, citing people familiar with the matter. The reason is not a Samsung story so much as an industry one. The market leader's advanced capacity is effectively full, which turns the second source into a price setter rather than a discount. Here is what changed and how it reaches your hardware budget.
The short answer
Reuters reported on Tuesday, August eighteenth, 2026 that Samsung raised foundry prices in July. The 4 nanometre SF4 process rose 10 to 15 percent for customers in the United States and China and 5 to 10 percent in Taiwan, the 5 nanometre SF5 process rose 10 to 15 percent, and the older 8 nanometre node rose by nearly 10 percent. Samsung's SF4 line at Pyeongtaek has run at full utilisation since late 2025, and TSMC's advanced capacity is effectively saturated.
A second source usually competes on price. When it starts competing on availability instead, the market has changed underneath you.
What Samsung changed
Reuters reported on Tuesday, August eighteenth, 2026, citing people familiar with the matter, that Samsung raised contract prices for its advanced chipmaking services during July.
The 4 nanometre SF4 process went up 10 to 15 percent for customers in the United States and China, and 5 to 10 percent for customers in Taiwan. The 5 nanometre SF5 process went up 10 to 15 percent. The older 8 nanometre process went up by nearly 10 percent.
The regional split is worth noticing. Chinese customers took the steepest increases, which is a straightforward reading of demand rather than anything more complicated. Where buyers have the fewest alternatives, the price moves furthest.
Samsung has not issued a price list or a statement, so these are reported contract terms on new orders rather than a public announcement. The named customer base is not small. Qualcomm, Tesla, Apple, Broadcom and Nvidia all appear on it, and Google has been reported as in talks over SF4 manufacturing.
Why the second source stopped discounting
The market share numbers explain the whole thing. TSMC accounted for more than 70 percent of global foundry revenue in the first quarter of 2026. Samsung accounted for roughly 7 percent. That is the shape of a market with one dominant supplier and one meaningful alternative.
Under normal conditions the alternative wins business by being cheaper, because it has capacity the leader does not need to fill. That logic collapses when the leader has no capacity left. TSMC's advanced process lines are effectively saturated by AI chip demand, so a customer turned away there arrives at Samsung without a fallback to negotiate against.
Samsung is not sitting on idle tools either. Its SF4 line at the Pyeongtaek campus has run at maximum utilisation since late 2025. A fab running flat out has no reason to discount, and every reason to reprice.
The loss making division that suddenly works
Samsung's foundry business has lost money since 2022, and the reason is structural rather than a run of bad quarters. A leading edge fab is one of the most capital intensive things a company can build, and the depreciation runs whether the tools are busy or idle. Only a full order book at decent prices covers it.
Both halves of that condition are now met for the first time in years. Lee Min-hee of BNK Investment and Securities put it directly, saying that if Samsung raises prices from here, its foundry business could potentially become profitable as early as next year. Samsung itself has pointed to a return to profitability by early next year on the current trajectory.
There is a longer term consequence in this that is easy to miss while looking at the price rise. A profitable second source is a healthier industry than a loss making one. The alternative to expensive Samsung capacity is not cheap Samsung capacity, it is a foundry market with one viable supplier at the leading edge. Nobody buying silicon benefits from that.
Where it reaches you
This is the third notable price move in the chain in recent months, after Qualcomm's double digit increase, and the pattern is now hard to dismiss as noise.
Timing first. A wafer price agreed in July does not appear on an invoice in August. It passes through fabrication, packaging, test and assembly, then works through whatever finished inventory the vendor already holds. Six to twelve months to list prices is the normal shape, and it usually arrives disguised. A discount that no longer stretches as far. A promotion quietly not renewed. A configuration that used to be included and is now an option.
The detail worth planning around is the 8 nanometre increase, because mature nodes are not where the headlines are but they are where a great deal of infrastructure silicon is actually made. Switch ASICs, storage controllers, baseboard management controllers, power delivery components and network interface silicon largely live on nodes nobody calls leading edge. Equipment you would never describe as cutting edge is exposed to this.
Three responses are worth the effort. Stop modelling hardware that gets cheaper every refresh cycle, because that assumption has been wrong for roughly a year now. Quote earlier and hold quotes longer on anything with a long lead time, particularly switching and storage. And revalue the capacity you already own, because consolidation, right sizing and keeping working equipment for another year all get better as replacement cost rises, and none of them need a purchase order.
Sources and further reading
- Samsung hikes chipmaking prices by up to 15% on demand spike, Reuters via Free Malaysia Today, August 19, 2026
- Samsung Raises Foundry Prices Up to 15% as TSMC Capacity Fills, Seoul Economic Daily, August 19, 2026
- Samsung raises 4nm and 5nm chipmaking prices as AI demand grows, SammyFans, August 19, 2026
- Samsung Foundry Raises Prices on Some New Orders by Up to 15%, BigGo Finance, August 2026
Frequently asked questions
Exactly which prices went up, and by how much?
Three nodes moved, and the increases took effect in July 2026. The 4 nanometre SF4 process rose 10 to 15 percent for customers in the United States and China, and 5 to 10 percent for customers in Taiwan. The 5 nanometre SF5 process rose 10 to 15 percent. The older 8 nanometre process rose by nearly 10 percent. Reuters reported the figures on Tuesday, August eighteenth, 2026, citing people familiar with the matter. Samsung has not published a price list, so these are reported contract prices on new orders rather than a formal announcement.
Why is a second source raising prices instead of undercutting?
Because there is nowhere else to go. TSMC accounted for more than 70 percent of global foundry revenue in the first quarter of 2026 against roughly 7 percent for Samsung, and its advanced process capacity is effectively saturated by AI chip demand. When the leader cannot take another order, the alternative supplier stops competing on price and starts competing on availability. Samsung's own SF4 line at Pyeongtaek has run at full utilisation since late 2025, so it is not selling spare capacity either. A discount only makes sense when you have idle tools.
What does this mean for Samsung Foundry as a business?
It is the first credible route out of a four year loss. Samsung's foundry division has been loss making since 2022, largely because leading edge fabs carry enormous fixed costs that only a high utilisation, well priced order book can cover. With the line full and prices up, the arithmetic finally works. Lee Min-hee of BNK Investment and Securities said that if Samsung raises prices from here, its foundry business could potentially become profitable as early as next year. The named customer list includes Qualcomm, Tesla, Apple, Broadcom and Nvidia, with Google reported to be in talks for SF4 manufacturing.
When does this actually reach the price I pay for hardware?
Not immediately, and rarely in a form you can trace. A wafer price increase agreed in July reaches finished silicon after fabrication, packaging, test and assembly, then works through inventory the vendor already holds. Six to twelve months to list prices is the usual shape, and it lands as a smaller discount or a quietly withdrawn promotion far more often than as a visible increase. Note also that the 8 nanometre rise matters more than it looks, because mature nodes are where network silicon, controllers, management processors and power components are made. Those parts sit in equipment nobody thinks of as leading edge.
Is there anything useful I can do about it?
Three things, none of them dramatic. First, lengthen your refresh assumptions. If you have been modelling hardware that gets cheaper each cycle, that assumption has been wrong for about a year and this reinforces it. Second, quote earlier and lock longer on anything with a long lead time, particularly switches and storage controllers built on mature nodes. Third, treat capacity you already own as more valuable than a spreadsheet says. Consolidation, right sizing and squeezing another year out of working equipment all improve in value when replacement costs rise, and none of them require a purchase order.