Korea's August CPI: Half Base Effect, but PCs Are Up 23.7%
Korea's August CPI came in at 3.1% year over year on Tuesday. Half a percentage point of that is an accounting artifact from a one-off telecom discount a year ago. Strip it out and inflation is running near 2.5%. The number that matters for anyone following memory is buried further down the release: the CPI component for computers rose 23.7% year over year, and the statistics agency attributes it to semiconductor prices.
This post separates the headline from the base effect, traces DRAM contract prices into Korean consumer prices, and follows the loop back: chip-export income lifting core inflation, the Bank of Korea hiking twice in a row, and what a stronger won and a 3%-plus policy rate do to the value chain that started the cycle.
KEY TAKEAWAYS
1. August CPI printed 3.1%, but roughly 0.6pp is the base effect of SK Telecom's 50% fee refund in August 2025. The statistics agency itself puts the underlying rate near 2.5%. Core (ex food and energy) at 3.4%, a 39-month high, carries the same distortion with a larger weight.
2. The computer CPI component has gone from +5.1% in January to +23.7% in August; durable goods accelerated to 4.1%. Statistics Korea names memory chip prices as the cause. Chipflation is now an official line item in the Korean inflation basket.
3. The Bank of Korea raised its base rate to 3.00% on August 27, its second consecutive hike, with a dot-plot median of 3.25%. Its stated logic: chip-export income flowing into domestic demand adds up to 0.2pp to core inflation. Today's CPI is the first data point before the October meeting.
What the 3.1% is made of
Statistics Korea (now the National Data Office) reported the August CPI at 120.05 (2020=100), up 3.1% year over year and 0.2% on the month. The path this year: 2.0, 2.0, 2.2, 2.6, 3.1, 3.2, 2.8, 3.1. May and June were fuel-driven after the Middle East war; August is a different animal.
The largest contributor is telecom: communication prices rose 16.6%, the most since the series began in 1985, with mobile fees up 26.7% and a 0.63pp contribution to the headline. In August 2025 SK Telecom refunded 50% of all subscribers' fees after a data breach; twelve months later that discount drops out of the base. The agency put the base effect at about 0.58pp and the underlying rate near 2.5%.
That matters for the core number. Core CPI ex food and energy rose 3.4%, the highest since May 2023, and many headlines ran with it. The agency's caveat: the core basket has fewer items, so the mobile-fee weight is proportionally larger. Core is more distorted than the headline. No ex-base core figure was published, so I will not invent one.
Outside telecom: fuel +14.2% (July 15.5%), contributing 0.54pp, with the petroleum price cap estimated to have shaved 0.5pp off the headline; farm products -2.6% on the year despite a 12.4% monthly jump in vegetables; industrial goods and services +3.7% each; durables up from 3.9% to 4.1%; living-cost index +3.2%, or +4.8% ex food; inflation expectations steady at 2.7%.
Chipflation has reached the CPI basket
The computer component, year over year, from January: 5.1, 10.8, 12.4, 19.4, 19.0, 22.2, 25.1, 23.7. July was a record; August eased but still runs ahead of fuel. Statistics Korea has explained it the same way in both briefings: semiconductor prices plus some new product launches. In June, storage devices were up 45.6%, the largest increase among all 458 items surveyed.
Upstream, the Bank of Korea's May PPI showed DRAM up 445.4% year over year, flash memory up 223.1%, notebooks up 30.7% and desktop PCs up 34.6%. The pass-through from memory to OEM list price to retail shelf is visible in official statistics at every stage.
On the supply side, TrendForce projected 3Q26 conventional DRAM contract prices up 13 to 18% quarter over quarter, NAND up 10 to 15% and mobile DRAM up 8 to 13% (July and August 2026 reports), driven by capacity moving to server and HBM at the expense of PC and consumer supply. It also flagged broad notebook retail price increases and consumer customers reaching their affordability limit.
Two readings for memory investors. DRAM increases reaching final consumers means supplier pricing power holds all the way down the chain. But the first dip in the computer component, 25.1% to 23.7%, points the same way as TrendForce's affordability comment: as consumers respond to price, suppliers push volume toward servers, which supports consumer DRAM pricing while squeezing set makers on volume.
The loop: chip income, core inflation, and the BOK
On August 28 the Bank of Korea published an issue note on demand-driven core inflation that is effectively the reasoning behind its hikes. Chip exports have improved the terms of trade, so real gross domestic income is growing well ahead of real GDP; if that purchasing power flows into spending, it becomes demand-side inflation.
The estimates: in a high-pressure regime a 1pp wider GDP gap lifts core by 0.1 to 0.4pp; if the income gain converts into consumption, add 0.05 to 0.2pp; a demand shock that fades to 0.2pp after three quarters in a low-pressure regime builds to 0.6pp after six quarters in a high-pressure one. Once core crosses roughly the mid-2% range, price co-movement across items intensifies sharply, especially in personal services.
The August forecast revision followed the same logic: core projections for 2026 and 2027 raised to 2.5%, 2026 growth lifted from 2.6% to 3.3% with half the upgrade attributed to semiconductors. August core at 3.4% cannot be compared cleanly to that threshold, but July core at 2.6% was already there. At today's inflation review the deputy governor said September's headline would fall as the base effect drops out, but the core uptrend would continue.
Rates and the won: how it comes back to the value chain
The BOK raised its base rate from 2.50% to 2.75% on July 16 and to 3.00% on August 27, the first back-to-back hikes at the start of a cycle; the six-month dot-plot median is 3.25%. Governor Shin Hyun Song said the board would look at CPI, the Q2 GDP revision, sentiment and income spillovers before October. Today's release is the first of those.
Three channels for the value chain. FX: the governor called the won in the upper 1,300s still weak and expects tightening to strengthen it. Memory exporters book dollars and report in won, so a stronger won shrinks reported revenue; for dollar-importing equipment and materials firms it lowers costs. Funding: a 3.00% base rate, or 3.25%, raises the cost of debt-funded capex, and mid-size suppliers and project-financed data center and power build-outs are more exposed than the large IDMs. Domestic demand: a 23.7% computer print means households are already paying for chipflation; higher rates hit PC and smartphone replacement demand first.
A quick illustration of the FX channel:
| Scenario | Avg KRW/USD | Revenue in won | Change |
|---|---|---|---|
| USD 10B revenue, current level | 1,380 | KRW 13.8T | - |
| Won 3.6% stronger | 1,330 | KRW 13.3T | -KRW 0.5T (-3.6%) |
| Won 7.2% stronger | 1,280 | KRW 12.8T | -KRW 1.0T (-7.2%) |
Illustrative calculation. Not actual company figures. While memory prices rise at a double-digit quarterly pace, FX is buried under the price effect; as increases narrow from 4Q26, it becomes more visible.
What I actually watch
| When | What | Why |
|---|---|---|
| Early October (September CPI) | Core inflation with the telecom base effect gone | Whether the mid-2% threshold is crossed without distortion is the input for the October decision |
| October MPC | Hike to 3.25% or hold; dot-plot shift | A third consecutive hike would make the FX and funding channels bite |
| Monthly CPI | Direction of the computer, storage and portable device components | Direct gauge of retail pass-through and consumer affordability limits |
| Sep-Oct TrendForce | 4Q26 DRAM and NAND contract price outlook | Whether consumer-segment increases narrow and the server/HBM reallocation continues |
Value chain read-through
| Segment | Direction | Basis |
|---|---|---|
| Memory IDMs (Samsung, SK hynix) | Pricing power confirmed; translation drag if won strengthens | Computer CPI +23.7%, PPI DRAM +445%; BOK wants a stronger won |
| PC and smartphone OEMs | Cost push plus softer domestic demand | TrendForce affordability limit; rate hikes squeeze purchasing power |
| Equipment and materials | Cheaper imports vs higher funding costs | FX and rate channels pull in opposite directions |
| Data center and power equipment | Higher project financing costs | 3.00% base rate, further hikes on the table |
| Telecom | Base effect disappears from September | 0.63pp contribution drops out, pulling headline down |
Risks to this view
- A re-escalation in the Middle East would push fuel's contribution back up, and the end of the petroleum price cap would add roughly 0.5pp in one step.
- If the September headline falls back into the 2% range as the telecom effect drops out, markets may price an end to tightening; the BOK is watching core, so reading the headline alone could mislead.
- If the July peak in the computer component marks the start of consumer demand destruction, consumer DRAM and NAND contract price increases could narrow faster than TrendForce's forecast.
- The core-distortion discussion here rests on the 0.58pp the agency provided for the headline. No official ex-base core figure exists.
The August print mattered less for its 3.1% than for two numbers inside it: the 0.6pp telecom illusion, and the 23.7% computer component that survives any adjustment. The loop from chip-export income to inflation, from inflation to rates and the won, and from there back to the chip value chain has now completed one visible turn. September CPI and the October MPC will show which way it turns next.
Sources: Statistics Korea (National Data Office), August 2026 CPI release and briefing (Sep 2, 2026); Bank of Korea inflation review (Sep 2, 2026), issue note on demand-driven core inflation (Aug 28, 2026), MPC statements (Jul 16, Aug 27, 2026), May 2026 PPI; TrendForce 3Q26 memory price forecast (Jul 3, 2026) and mobile DRAM contract price note (Aug 3, 2026). Everything here is from public sources.
Comments
Post a Comment