Summary
For much of Korea’s AI cycle, the investment case has been concentrated in Samsung Electronics and SK hynix.
That concentration made sense.
AI infrastructure spending first translated into stronger HBM demand, tighter advanced-memory capacity and rapidly improving economics for Korea’s two memory leaders.
The next question is whether that demand is beginning to create a second earnings layer further down the semiconductor value chain.
Korea’s semiconductor exports reached USD 34.12 billion during September 1–20, up 259.4% year over year and the highest level recorded for the period.
At the same time, SK hynix has approved KRW 35.2 trillion for its Yongin Y2 DRAM fab and KRW 19.1 trillion for its Cheongju M17 NAND fab. Samsung Electronics expects HBM to absorb nearly 30% of industry DRAM wafer capacity in 2027, compared with around 20% currently.
Those developments do not automatically produce equipment earnings.
But they create the conditions for additional process equipment, advanced packaging and testing demand.
The central investment question is whether Korea’s AI trade is evolving from a memory-led earnings cycle into a broader semiconductor investment cycle — and which suppliers can turn that transition into actual orders and profits.
Key Takeaways
- ✓ Semiconductor exports reached USD 34.12 billion during September 1–20, rising 259.4% year over year and accounting for nearly half of Korea’s total exports during the period.
- ✓ SK hynix has committed KRW 54.3 trillion across two new memory fabs, but the exposure is broader than HBM alone: Yongin Y2 is a next-generation DRAM base, while Cheongju M17 is a NAND production base.
- ✓ Samsung Electronics expects HBM to consume nearly 30% of industry DRAM wafer capacity in 2027, increasing the capital required to support both HBM and conventional memory output.
- ✓ Equipment-market evidence is beginning to appear, but unevenly. Wonik IPS ended 2Q26 with KRW 634.6 billion of order backlog according to NH Investment & Securities, even as its reported quarterly revenue and operating profit remained below year-earlier levels.
- ✓ The next phase of Korea’s AI trade will be determined by orders, backlog and revenue recognition — not by semiconductor-equipment share prices alone.
What Happened
Korea entered the final part of September with exceptionally strong semiconductor export momentum.
Total exports reached USD 71.4 billion during September 1–20, up 78.3% year over year and setting a record for the period.
Semiconductor exports reached USD 34.12 billion, up 259.4% year over year.
Chips accounted for approximately 48% of total exports during the period.
Meanwhile, SK hynix approved two major domestic fab investments in August.
The company plans to invest KRW 35.2 trillion in Yongin Y2, a DRAM production base designed to manufacture HBM and other next-generation DRAM products.
Construction is scheduled to begin in July 2027, with the first cleanroom targeted to open in June 2029.
SK hynix will separately invest KRW 19.1 trillion in Cheongju M17.
M17 is designed as a NAND production base, with its first cleanroom scheduled for December 2028. The company has linked the investment to growing enterprise SSD and AI-service storage demand.
Samsung Electronics has also highlighted increasing capacity pressure from HBM.
A company executive said HBM is expected to account for nearly 30% of the DRAM industry’s wafer capacity in 2027, compared with around 20% currently.
Because HBM and standard DRAM compete for underlying wafer capacity, increasing HBM production can reduce the capacity available for conventional DRAM unless additional production resources are added.
Market participation has started to broaden as well.
On September 29, institutional net buying was concentrated across several Korean semiconductor-equipment and component companies, including HPSP, Wonik IPS, Jusung Engineering, EO Technics, Reno Industrial, PSK, Eugene Technology and ISC.
One trading session does not establish a semiconductor-equipment cycle.
It does show that investors are increasingly looking beyond the memory producers themselves.
Why It Matters
The first phase of Korea’s AI semiconductor cycle was relatively direct.
AI Compute Demand
→ HBM Demand
→ Memory Pricing
→ Memory Earnings
That transmission is already visible.
The next phase requires another step.
AI Memory Demand
→ Capacity Pressure
→ Fab & Process Investment
→ Equipment Orders
→ Advanced Packaging
→ Testing
→ Supplier Earnings
This distinction matters.
Equipment companies do not benefit simply because the price of HBM rises.
They benefit when semiconductor manufacturers decide that existing production infrastructure is insufficient, when new manufacturing processes require different equipment, or when chip complexity increases the value of packaging and testing.
CAPEX is therefore the bridge between AI-memory demand and equipment earnings.
And that bridge takes time to cross.
A fab investment decision is not revenue.
A purchase order is not yet recognized earnings.
Backlog is not cash flow.
The investment case becomes stronger as each stage is converted into the next.
The interesting signal is not that semiconductor-equipment stocks are rising.
It is that the conditions required for earnings to spread beyond memory are becoming increasingly visible.
Semiconductor exports are exceptionally strong.
HBM is consuming a growing share of DRAM production capacity.
SK hynix has committed capital to additional DRAM and NAND production infrastructure.
And some equipment suppliers are beginning to show stronger forward order visibility.
Wonik IPS illustrates both the opportunity and the remaining gap.
According to NH Investment & Securities, its order backlog reached KRW 634.6 billion at the end of 2Q26, approximately 220% higher than a year earlier.
Yet its actual second-quarter revenue was KRW 216.5 billion, down 10.6% year over year, while operating profit fell 49.6% to KRW 18.4 billion.
That divergence is important.
The market is increasingly pricing the earnings that could emerge from the next CAPEX cycle.
The reported earnings have not fully arrived yet.
That makes backlog, equipment move-in schedules and customer investment execution more important than daily stock-price momentum.
Korea’s AI trade may be broadening.
The next test is whether earnings breadth follows market breadth.