QRT (405100) Stock Outlook 2026: AI Chip Testing Demand vs. Customer Concentration

QRT: AI Chip Testing Demand or Concentration Trap?

Here is the honest take. QRT is one of the few small caps where the “picks and shovels for AI” label actually fits, because every new generation of memory and logic has to be proven reliable before anyone ships it. But a stock this size, with revenue tied to a few giant customers, can look brilliant and fragile in the same quarter. My read is that the business is good, the price has already run, and the right move is to size small and let earnings earn your trust.

QRT tests semiconductors for durability and takes failed chips apart to explain what went wrong. Spun out of SK hynix, it inherited deep memory-failure know-how. That heritage is also the catch: a company with a famous parent tends to depend on a short list of buyers.

Market data snapshot (2026-10-16 as of · Toss Securities Open API) Price 16,890 KRW · Market cap KRW 0.21T · 52-week range 8,060 KRW~24,200 KRW (52-week position 55%) · Return 1M +28.0% / 3M +57.9% / 6M -20.9% / 1Y +0.2% · 20-day avg volume 60,672 · Foreign net buy (20d) -33,741 sh · Institutional net buy (20d) -4,012 sh · Foreign ownership 4.1% · Short-sale ratio (5d) 3.22% · Trading alerts none _{Returns use adjusted daily closes; market cap = price × shares outstanding.}

Start with where the price sits. At 16,890 KRW, the stock is a little past the midpoint of a 52-week range that runs from 8,060 to 24,200. That is neither a bargain-bin low nor a blow-off high. The return line is more telling: up 57.9% in three months, yet down 20.9% over six and flat over a year. Anyone who bought a year ago is roughly where they started, after a round trip. Anyone who bought in the last quarter is sitting on a big gain. That split matters, because those two groups behave differently when the stock bounces.

---

What does QRT actually do?

Nothing leaves a chip factory for a customer’s product until there is data showing it will last. That data comes from reliability testing. When something fails during those tests, or in the field, failure analysis finds the root cause by cutting the chip open and tracing the fault. QRT sells both services as an outside specialist.

Service	What happens	Revenue character
Reliability testing	Heat, humidity and voltage stress to compress years of wear into weeks	Project-driven, peaks around new-product qualification
Failure analysis	Physical and electrical tracing of a dead chip back to its cause	Event-driven, spikes when yield or field problems appear
Soft-error evaluation	Checking resilience against radiation-induced glitches	Required for high-reliability uses such as memory and automotive
Standards-based testing	JEDEC and automotive AEC-Q100 style test programs	Recurring, formulaic volume

Notice what this is not. It is not a business that sells a finished product to end consumers. It rides on its customers’ development and ramp cycles. When customers are busy launching chips, QRT’s labs fill up. When they pause, utilization sags.

The cost structure explains the volatility. Chambers and analysis tools are depreciated whether or not they are busy, engineers are paid either way, and power and consumables run regardless. So a full lab converts revenue to profit efficiently, and a half-empty lab does the opposite. Operating leverage swings both directions. If you only look at the margin from a strong quarter, you will overestimate what the business earns on average.

---

Why would AI and HBM make reliability testing more valuable?

A modern AI accelerator pairs a GPU with several stacks of high-bandwidth memory, and each stack is a tower of DRAM dies connected vertically. More layers, more connection points, more ways to fail. If one layer in a tower is bad, the whole stack can be scrapped, so chipmakers test more and analyze more when a problem appears. AI servers also run hot and hard for long stretches, which makes durability a bigger selling point than it is for a phone chip used a few hours a day.

That is the bull case in a paragraph: each product generation adds qualification work, and tougher failures need more forensic effort.

Two caveats keep me from getting carried away. First, large chipmakers already run their own reliability labs. External specialists often catch overflow and specialized work, so growth in industry demand does not automatically become growth in outsourced revenue. Second, being in the right industry is not the same as capturing the demand. Equipment, customer approvals and trained analysts decide how much QRT can actually take on. For the broader AI investing frame, see our AI stocks investment guide for 2026.

---

How dangerous is the customer concentration?

This is the question the whole thesis hangs on. The annual report lists major customers and their share of revenue. I am deliberately not quoting a percentage here, because it changes and I will not print a figure I have not verified. Open the latest filing and read it yourself.

Why it matters structurally: the buyer universe for reliability work is small. A few memory makers, a few foundries and some large fabless companies generate the big volumes. When one of them also happens to be the former parent, bargaining power tilts. A price cut request, a decision to build more internal capacity, or a delayed product launch can each hit revenue and utilization at once. Because fixed costs dominate, a modest drop in utilization hits profit much harder than it hits sales.

There is a flip side. A big anchor customer provides steady volume while it develops new products, and years of working through the same specifications creates familiarity that a newcomer cannot copy overnight. The problem is that this advantage shows up in good times and the risk shows up in bad ones. So I care less about concentration as a static number and more about its direction. Is the share of non-anchor customers rising, quarter by quarter? That one line tells me more than any headline growth rate.

If you want to see how another Korean small cap with a customer-linked business model looks through the same lens, our write-up on SFA Engineering asks a similar question.

---

Who competes with QRT, and is there a real moat?

Very few listed companies do exactly this, so the peer set stretches to adjacent businesses.

Group	Examples	Relationship to QRT
Korean back-end test	Doosan Tesna, Nepes, Hana Micron	Wafer and package testing is their core, partial overlap, different model
Global testing and certification	Intertek, SGS, Eurofins	Strong in general testing, semiconductor-specific depth varies
Taiwan failure analysis	MA-tek and peers	Strong near the foundry ecosystem
Customer in-house labs	Large chipmakers’ own teams	The most realistic competitor and the biggest customer at once

That last row is the nature of the industry. The toughest rival is not an outside firm. It is the customer’s own lab.

If QRT has a moat, I see three layers. One is memory failure experience: years of accumulated fault patterns and analysis technique do not get replicated in a few quarters. Two is customer trust: test results become the basis for production approval, so customers rarely swap labs casually, and consistency of data is itself an asset. Three is independence: a third-party report carries weight that an internal one sometimes does not, especially when a dispute or an external standard is involved.

But be precise about what a moat gives you here. It gives a barrier to entry, not pricing power. With few customers, price is negotiated, and a narrow buyer base tends to push it down. A company can be hard to displace and still earn modest margins.

---

Why did the stock swing from -20.9% to +57.9%?

Look at the flow data next to the return data. Over the last 20 trading days, foreigners were net sellers of 33,741 shares and institutions net sellers of 4,012 shares. With a 20-day average volume of 60,672 shares, the entire foreign net selling over that span is a bit more than half of a typical single day’s trading. That is not heavy distribution. Foreign ownership is only 4.1%, so this was never a foreign-driven name to begin with.

My inference is that the recent surge was retail-led, with big institutions neither chasing nor dumping. At a market cap of roughly KRW 0.21 trillion, a small group of aggressive buyers can move this stock, and the same group can leave quickly. A 5-day short-sale ratio of 3.22% is not alarming, but it shows some traders are betting on a pullback after the run. No trading alerts are flagged, so the exchange is not warning about anything unusual.

Position adds one more consideration. Sitting above the midpoint of the 52-week range, the price is still below where it was six months ago. Holders who bought then are underwater, and a bounce toward that area tends to attract sellers who just want their money back. Expect supply on the way up.

---

What can go wrong?

Concentration. Covered above. One customer’s decision becomes your quarter.

In-sourcing. Ironically, when AI memory volumes surge, big manufacturers may build more of their own test capacity, leaving less to outsource.

Capital intensity. New standards and new packaging require new chambers and analysis tools. If revenue does not arrive, depreciation eats profit.

Thin liquidity. With average volume around 60 thousand shares a day, size is hard to move in or out without paying for it.

Priced-in optimism. A 57.9% three-month gain means a lot of the verification story is already in the quote. A miss would hurt more than a beat would help.

Theme beta. When the AI semiconductor theme cools, small related names get sold regardless of their own numbers. Small caps trade on narrative first, a pattern also visible in our look at Binex.

---

How would a US investor approach it? Three practical scenarios

Scenario 1: Do not chase, scale in

If I wanted exposure, I would start with a third of my intended position and add only if the next quarters show the major-customer share easing or utilization improving. If results disappoint, I would not average down on hope. I would also set the exit condition in advance, something like a decisive break below the middle of the 52-week range on rising volume, so the decision is not made in the heat of a drawdown.

Scenario 2: Taxes and account type

For a US taxpayer, gains are reported in dollars, with short-term gains taxed at ordinary rates and long-term gains at the preferential rates after holding for more than a year. Losses can offset gains, and the wash-sale rule can disallow a loss if you rebuy within the window, which matters for a volatile small cap. Because the stock trades in won, your dollar result blends the share price move with the won-dollar move. Foreign account reporting and Korean-side tax treatment depend on how and where you hold it, so check with a tax professional. Our capital gains tax guide is a useful companion read on how gains get taxed.

Scenario 3: Currency drag or tailwind

You are buying won. If the won weakens against the dollar while you hold, your dollar return shrinks even if the share price is flat. If it strengthens, it adds. For a position that is already volatile, currency can double the noise. Decide whether you want to carry that risk deliberately, or keep the position small enough that it does not matter. Also compare the cost of converting dollars and the broker’s foreign-market fees, since on a small position those frictions are a real percentage of your return.

For a different flavor of Korean small cap with its own risks, you can also read our Miwon Specialty Chemical outlook.

---

Metrics to watch each quarter

1. Revenue share of the largest customers. The most important line. Falling dependence is the best news this stock can deliver.
2. Operating margin trend. Rising revenue with falling margin means price pressure or depreciation drag.
3. Capacity additions versus utilization. Look at the gap between buying equipment and seeing the revenue arrive.
4. New-service revenue. Any separate mention of HBM, automotive or advanced-packaging testing is a sign the thesis is becoming numbers.
5. Investor flow. Right now it is retail-heavy. A shift to net buying from foreigners or institutions would add credibility.

Until several of those point the same way, I keep the position small.

---

Keep reading

• SFA Engineering stock outlook 2026
• Binex stock outlook 2026
• Miwon Specialty Chemical stock outlook 2026
• AI stocks investment guide 2026

---

This article is an informational opinion and is not a recommendation to buy or sell any security. Investing involves risk, including loss of principal. Business conditions and outlooks described here reflect the time of writing, so check current filings and consult a licensed professional before making decisions. Tax comments are general and not tax advice.