Pricing power and competitive strength are not the same thing, and this is the cleanest available demonstration of the difference. Samsung raised prices because its 4nm lines filled, and the price list reads as a map of which customers have somewhere else to go.
Samsung's share of global foundry revenue fell to 7.3% in the second quarter of 2026, down from 11.5% a year earlier. Over the same period TSMC's share rose from 62.3% to 70.2%. Samsung's foundry division has been loss-making since 2022. In July, according to reporting by Reuters citing unnamed sources, it raised prices on advanced contract chipmaking for new orders by as much as 15%.
The attribution matters throughout this piece. The price increases are sourced reporting, not a Samsung announcement. Samsung has not confirmed the percentages, the regional differential or the timing. The market share figures are from published quarterly foundry revenue estimates and can be stated more firmly.
The tempting read is that AI demand has become so hot that even the weak player has pricing power. That is close enough to be plausible and wrong enough to matter, because it collapses two things worth keeping apart.
Pricing power and competitive strength are not the same property, and this is about as clean a demonstration of the difference as the semiconductor industry has produced. Samsung did not close the technology gap to TSMC. Its share of the market fell by roughly a third in twelve months while it was raising prices. What changed is not that Samsung's process got better; it is that Samsung's 4nm lines filled.
In foundry, the product being sold is not really a process node. It is capacity on a specific date. A customer taping out a design needs wafers in a particular quarter, and a fab that is fully booked has nothing to sell at any price. When the market leader's capacity is committed, the marginal buyer's choice is not between a better foundry and a worse one - it is between a worse one and nothing.
That is the mechanism generating Samsung's price increase. Per the reporting, 4nm SF4 rose by 10 to 15% for some customers, 5nm SF5 by 10 to 15%, and even the ageing 8nm line by close to 10%. An 8nm increase is the most telling of the three: nothing about 8nm got more competitive in 2026. Capacity displaced from advanced nodes cascades down, and mature lines that were previously price-takers stop being so.
Samsung is not being paid more because its chips got better. It is being paid more because the alternative got harder to reach. Those produce identical revenue and entirely different durability.
The most informative detail in the reporting is not the size of the increases but their distribution. Customers in China and the United States are reported to be taking increases of 10 to 15%. Customers in Taiwan are taking 5 to 10%.
Read that as a pricing schedule and it is unremarkable. Read it as a measurement of negotiating position and it is precise. A fabless firm headquartered in Hsinchu, with engineering staff who can be at a TSMC fab in an afternoon and an existing relationship with the market leader, negotiates from a different place than a firm that cannot easily reach that capacity. The five-to-ten-point differential is that fact expressed in dollars. Price discrimination of this kind is not a moral failing; it is what any supplier does when it can observe which customers have somewhere else to go.
US export controls on advanced chipmaking equipment to China were designed to slow the development of domestic Chinese advanced-node capability. Whether they have achieved that is a separate argument. What is observable is a second-order effect: constrained domestic capability increases Chinese firms' reliance on overseas foundries, and reporting indicates Chinese customers are among those accepting the steepest increases.
Stated as mechanism rather than verdict: restricting the equipment that would let a set of buyers supply themselves, without restricting their demand, concentrates that demand onto the remaining suppliers and transfers margin to them. In this instance a meaningful share of that margin is arriving at a South Korean foundry that had been losing money for four years. That is not an argument for or against export controls, which are made on national security grounds that pricing effects do not settle. It is an item that belongs on the ledger when the effects are being counted, and it is routinely left off.
There is a straightforward defence here and it deserves to be made at full strength. A business that has lost money since 2022 raising prices toward viability is ordinary corporate behaviour, not opportunism. Samsung has subsidised this division for years out of other earnings, on a strategic bet that a second credible advanced-node supplier is worth having. AI demand is genuinely tightening capacity across the whole industry, input and equipment costs have risen, and pricing that was set to win share during a share-winning phase was never going to persist into a shortage.
The counter-argument is not that the increase is unjustified. It is that customers should be clear about what they are buying. Nothing in this quarter suggests Samsung has closed the gap on yield or on process maturity, and a decision to move a design to SF4 made under capacity pressure is a different decision from one made on the merits - with different consequences when the pressure lifts.
Pricing power derived from a shortage is the most fragile kind there is, because it lasts exactly as long as the shortage. Samsung's leverage here is a function of TSMC's capacity being full, and TSMC is spending heavily to make that untrue. As new advanced-node capacity comes online through 2027 and beyond, the marginal buyer regains a choice, and a supplier at 7.3% share negotiating against a supplier at 70.2% will find that the differential runs the other way.
Which makes the interesting question not whether Samsung can hold these prices - it probably cannot - but what it does with the interval. Margin earned during a shortage can fund process development that closes a real gap, in which case the pricing power eventually becomes the competitive strength it was mistaken for. Or it can simply reduce a loss for a few quarters. The 2026 numbers will not distinguish between those two outcomes. The 2028 share figures will.
Because they are buying capacity on a date, not a process node in the abstract. When TSMC's advanced-node lines are fully booked, a customer needing wafers in a given quarter cannot choose the better foundry at any price - the choice is between the second-best foundry and not shipping. Price in that situation reflects availability rather than relative quality, which is why the increases are steepest for customers with the least practical access to alternatives.
DrafterDaily tracks chip supply, foundry pricing and the second-order effects of trade policy - including the ones nobody intended.
Explore Business coverageThe headline was the 94% drawdown from a $3.44 billion peak. The more useful arithmetic is the cash sitting inside the thing being bought — and the fact that everyone except the angels was paid in another private company's paper.
The biography is the least interesting part. What the board actually did was unbundle the CEO job — keeping Tim Cook on with a defined policymaker-engagement remit — and that separation says more about Apple's next decade than the appointment does.
Every outlet reported the size of Meta's teen-safety settlement and the screen-time limits. Almost nobody read the payment structure — where roughly $5.3bn is released only if two competitors act.