Electronics prices for 2027 are being negotiated right now — the pass-through path of a triple chip-cost hike

Note: This article was automatically translated from Korean by AI. All content refers to South Korea unless otherwise stated. See Korean original for the most accurate reading.

Bottom line — The cost base for the laptops, cars and appliances you will buy in 2027 was already settled at negotiating tables in June and July of this year.

On 21 July, reports emerged that the world’s largest foundry had wrapped up talks with customers on raising 2027 contract manufacturing prices by up to 10%. One week earlier, on 14 July, an industry association forecast that the semiconductor equipment market would reach its largest size on record, while memory contract prices jumped nearly 30% in three months. Chip costs are rising not on one layer but on three simultaneously. This piece traces where those increases travel before they reach a consumer price tag, and where South Korea stands in that chain.

1. Negotiations began in June and closed in July

The reporting has a simple skeleton. The world’s largest contract manufacturer is raising the wafer prices it will apply from 2027. Talks started in June, a final proposal was settled in July, and the new prices take effect from early 2027. Multiple Taiwanese outlets carried the same report that day.

By segment, it breaks down as follows. Leading-edge nodes at 7nm and below face a baseline 5–10% increase; mature nodes up to roughly 10%. On top of that, high-performance computing customers placing orders beyond their original forecasts face an additional 10–15% above the baseline hike. This is not an even sharing of the cost burden — it is a structure where whoever urgently demands more pays more.

One important caveat. These figures were not officially announced by the company; they became known through media reporting. That is why every mention of an increase rate in this article carries a “per media reports” tag. Until officially confirmed, they should be treated as an observation shared by the market, not as established fact.

The decision to defer the increase to 2027 is not a detail to skip over either. It gives customers time to redo their designs, inventories and pricing — which, flipped around, means set makers will spend the whole of the second half of 2026 quietly rewriting their cost sheets.

2. The increase is not one layer — it is three at once

This is where the real problem starts. It is not only foundry prices that are rising. Costs are climbing at every layer a chip passes through before it exists.

One thing must be made explicit about this table. These four lines are not values that can be added together. One is an increase rate scheduled for 2027, one is a forecast of total market size, one is a price already transacted, and one is a specific company’s quarterly unit price. They measure different things over different periods. If you see arithmetic of the form “add them all up and chip prices rise by X%”, treat it as a manufactured number.

The direction of all four lines, however, is the same. Up.

Bar width is the ratio against the maximum value (USD 40). The two products have different specifications, so this shows each one’s rate of increase rather than an absolute comparison.

3. Record earnings — so why raise prices?

There is a reason the increase reports sound odd. The company raising prices is posting the best results in its history.

The second-quarter figures disclosed at its mid-July earnings call read as follows. Revenue of TWD 1.27 trillion, after-tax net income of TWD 706.56bn, earnings per share of TWD 27.25, and a gross margin of 67.7%. Revenue rose 36% year on year and net income 77.4%. By node, 2nm accounted for 3%, 3nm 30%, 5nm 33% and 7nm 11% — meaning leading-edge nodes at 7nm and below made up 77% of wafer revenue.

Bar width is the ratio against the largest share (33%). Based on disclosures at the Q2 2026 earnings call.

The answer lies on the capital expenditure side. At the same event, the 2026 annual capex plan was revised up from USD 52–56bn to USD 60–64bn. Management indicated that capital spending over the next three years would rise far more than over the past three. The annual revenue growth guidance was also lifted from “above 30%” to slightly above 40%.

Spend that much and depreciation follows. New plants being built overseas — in the United States and Japan — carry higher construction and labour costs than existing fabs in Taiwan, and yields run lower in the early operating phase. In other words, the point at which depreciation lands on the cost sheet is approaching. Rising material and equipment prices compound it. The justification for raising prices is not “profits are insufficient” but “far more money is going out from here.”

Frankly, when the reports first appeared, a company posting record results raising prices did not quite add up. But placing the capex figures alongside it, the sequence turns out to run the other way. This does not read as raising prices to defend profits; it reads as raising prices while profits are strong, to secure three years’ worth of investment funding in advance.

4. Cheap chips are disappearing — why mature nodes are rising too

Leading-edge increases are easy enough to accept. AI accelerator demand is overflowing. The harder part to understand is mature nodes. Older processes such as 28nm and 40nm have long been assumed to get cheaper over time.

The structure has changed. An AI server does not contain only compute chips. Power management chips, analogue chips and interface chips attach in volume. All of these are mature-node products. The larger AI investment gets, the more mature-node capacity it absorbs. The era of competing on price with spare capacity is ending.

Movements in China illustrate the shift well. The mainland’s largest contract manufacturer saw Q1 shipments fall 0.2% quarter on quarter, while the average selling price per wafer rose 2.5%. It was a quarter built on unit price rather than volume. Local media report that mature-node prices could rise a further 5–10% in the second half of 2026 — this part is a forecast, not a confirmed figure.

When mature-node prices rise, they actually reach consumers faster. A single car contains hundreds of mature-node chips, and so do refrigerators, washing machines and lighting. Leading-edge increases get buried inside the price of expensive servers; mature-node increases come down into everyday living costs.

5. South Korea is both a seller and a buyer

South Korea’s position in this chain is peculiar. One foot stands on the selling side, the other on the buying side.

Start with selling. June semiconductor exports reached USD 44.82bn, crossing USD 40bn for the first time on record and rising 199.5% year on year. The first-half cumulative total was USD 192.4bn, exceeding the full-year 2025 figure of USD 173.4bn in just six months. Total June exports came to USD 102.3bn, the first monthly reading above USD 100bn on record. The surge in memory contract prices is the main force behind those numbers.

The buying side is no small matter either. Korean fabless firms absorb the foundry price increase directly. Korean companies making finished goods — appliances, cars, telecom equipment — buy both memory and mature-node chips at higher prices at the same time. Rising memory prices inflate the country’s total export value, but for the domestic firms buying that memory to build finished products, it is a cost increase.

On the foundry side there is a spillover-benefit scenario. If the number one player raises prices, the relative price competitiveness of the second tier is highlighted, and US fabless firms gain more incentive to diversify supply chains. Reports that a Korean player’s 2nm yields have reached the 55–60% range, along with news of large customer orders, are cited as grounds for that scenario. That said, the yield figure too is industry talk rather than an official company announcement.

When I first saw the June export statistics, my eye went straight to the 199.5% growth rate. It was only after looking at the contract price table in the same release that the nature of that number became clear. Volumes did not double; the selling price rose sharply. The fact that “exports hit a record high” and “domestic set makers’ costs are rising” stem from the same cause is hard to see from a single line of statistics.

Taiwan’s numbers reveal how heavily this chain is tilted to one side. June exports of USD 74.83bn, up 40.3% year on year; first-half cumulative exports of USD 416.66bn, up 47.1%. Information and communication products together with electronic components accounted for 78.7% of first-half exports. Those figures show where the power to raise prices comes from.

6. The counter-view — “we do not adjust drastically”

That covers the case for increases. Here is the other side.

Five days before the increase reports appeared, management at the same company told its earnings call, in effect, that it does not make drastic price adjustments. That was read as a message that it will not raise prices enough to unsettle relationships with long-term customers. In other words, there is a temperature gap between the company’s official position and the media reporting.

There is also an absorption argument. The share of the final consumer price accounted for by chips varies by product. In products such as AI servers, where chips are most of the cost, an increase shows up immediately in the price; for smartphones and cars, non-chip costs are far larger. A 10% rise in foundry prices does not produce a 10% rise in finished-goods prices. On top of that, some observers note that AI-related customers have thick enough margins to absorb the increase themselves.

The third counterargument is timing. The increase applies in 2027. If the AI investment cycle turns over before then, the increase itself can be renegotiated. Price is a contract, not a law of physics.

Personally, I think the absorption argument holds up considerably well for leading-edge nodes and poorly for mature nodes. The thick margins sit with AI customers, not with the company making refrigerators. Then again, this is genuinely ambiguous, because the size of the mature-node increase is still at the forecast stage — how much actually comes through will only be known from second-half contracts.

7. Five inflection points to watch through 2030

Not much can be settled at this point. Instead, here are the places that will determine the direction of this trend over the next several years.

First, the depreciation peak at new overseas fabs. Once the plants being built in the United States and Japan enter full operation, depreciation lands on costs. When that burden peaks and turns determines how long the upward pressure lasts.

Second, second-tier foundry yields and order wins. For the spillover scenario to become real, yields must stabilise and large customers must actually move volume. It is more accurate to watch the point at which it registers as revenue, not the point of an order announcement.

Third, mature-node capacity and trade policy. Continued Chinese expansion in mature nodes is a downward force on prices over the long run — but if tariffs and export controls isolate that capacity from the market, it works the other way.

Fourth, the turning point of the memory cycle. Contract price surges eventually reverse. Given how large memory pricing’s contribution to Korean exports is, that reversal would change the direction of the entire export statistic.

Fifth, the pace of AI investment itself. A plan to raise capex sharply over three years rests on the premise that demand holds. If the premise wobbles, excess capacity and falling prices arrive together. By around 2030 it should be clear whether today’s increase phase was near the peak of the cycle or the start of it.

FAQ

Q1. Will laptop and smartphone prices really rise in 2027?

It is not settled. The foundry price increase is an observation per media reports, and the share of a finished product’s price accounted for by chips varies widely by category. That said, memory contract prices have already risen in fact, and mature-node prices are rising too, so the direction of cost is genuinely upward. Whether the increase flows into selling prices or is absorbed into manufacturers’ margins depends on the competitive situation of each company.

Q2. If memory prices rise, isn’t that good for South Korea?

In the export statistics, yes. June semiconductor exports crossing USD 40bn for the first time is the result. But South Korea also has many companies that buy memory to build finished products, and for them it is straightforwardly a cost increase. National-level figures and individual company profit and loss need to be looked at separately.

Q3. Why does a rise in mature-node chip prices matter more?

Because the distance to the consumer is shorter. Most chips in cars, appliances and lighting are mature-node products. Leading-edge increases get buried inside the price tag of expensive servers; mature-node increases come down into everyday living costs.

Q4. How likely is volume shifting to second-tier foundries?

The incentive has grown. When the leader raises prices, there is reason to look for alternatives. But as of 2025 the share gap is very wide — 69.9% against 7.2% — and chip designs are optimised to a specific process, so switching suppliers costs time and money. It is more accurate to watch the point at which mass-production revenue is actually recognised than the order announcement.

Q5. How should semiconductor-related assets be viewed right now?

This article is not material for judging individual stocks. The structure, though, is worth setting out. Price increases grow the seller’s revenue and grow the buyer’s cost. Even within the single word “semiconductor”, the direction of profit and loss reverses depending on which box you stand in. Binding an entire sector in one direction on the strength of a single news item is the most common misreading.

Closing — knowing the order of price increases reveals the timing

This article is not a buy or sell recommendation — it is an analytical memo tracing, point by point in time, the path along which chip costs rise.

To summarise: cost increases that began in equipment and materials are moving into foundry prices and then down into set costs, and in the meantime memory has already risen. The figures at each layer are different in nature and cannot be merged into one, but the direction is the same. That the increase applies from 2027 also means the second half of 2026 is the window for negotiation and inventory adjustment.

South Korea stands at both ends of this chain at once. The selling-side numbers keep breaking records; the buying-side costs rise quietly. Depending on which number you look at, the same news reads as good or bad. Deciding carefully, in line with the character of your own funds and time horizon, matters most.

Primary sources and company IR
SEMI Mid-Year Total Equipment Forecast 2026 (14 Jul 2026; global equipment revenue forecast of USD 165.9bn)
TSMC Q2 2026 earnings call materials (revenue, net income, gross margin, node mix, capex guidance)
Korea’s Ministry of Trade, Industry and Energy (산업통상부 / MOTIE), “June and First-Half 2026 Export and Import Trends” (semiconductor exports of USD 44.82bn; DDR5 and NAND contract prices)
Korea Policy Briefing, June export performance press release (monthly exports of USD 102.3bn)
Taiwan Ministry of Finance, June 2026 trade statistics (exports of USD 74.83bn; H1 USD 416.66bn)

Media referenced
Nikkei Asia · TechNews · Liberty Times Finance · Wealth Magazine · CNA-affiliated outlets · Economic Daily News (Taiwan) · The Korea Economic Daily · Yonhap-affiliated economic media · China Securities Times · 21st Century Business Herald