There was exactly one way to make money in KOSDAQ this year: of the top 20 gainers between January 2 and July 21, fifteen were AI and semiconductor infrastructure names — power, cooling, data centers, transformers. The things AI consumes. Meanwhile the company that gets subcontracted to actually design AI chips, Gaonchips (KOSDAQ 399720), spent most of that period near its 52-week low. As of July 24 it closed at 40,000 won, a market cap of about 474 billion won — less than half its 52-week high of 83,200.
The contrast is worth dissecting because Gaonchips has a foot in both worlds: theme and substance. Unlike the story-only small cap in my Nable case study, this is a company whose name actually appears on contracts in the AI silicon value chain. And yet the stock trades like a theme stock. July 10 was the proof.
What a design house actually does
Gaonchips is a semiconductor design house. It takes chip designs from fabless companies and implements them — back-end design — so they can actually be manufactured on Samsung Foundry’s process nodes, then manages the production ramp. It is the translator and general contractor between fabless and foundry.
Its position in that market is verifiable. It is an official Design Solution Partner (DSP) of Samsung Foundry and an official Arm Approved Design Partner (AADP). It is also cited as the first design house globally to officially win a 2nm-process project. As Samsung ramps 2nm from mobile in 2025 toward HPC and AI applications through 2027, every new customer Samsung lands needs someone to execute the design — and Gaonchips is first in line for that work.
The numbers went the opposite way from the story
| Metric | Value | Note |
|---|---|---|
| 2024 revenue · operating profit | 96.5bn · +3.5bn won | Record revenue |
| 2025 revenue · operating profit | 68.5bn · −16.7bn won | Revenue −29%, swing to loss |
| Q1 2026 | Revenue 20.6bn · operating loss −3.9bn won | Still loss-making |
| Share price (Jul 24) | 40,000 won | 52w range 34,950–83,200 |
| Market cap | ~474bn won | |
| P/B · debt ratio | ~8.5x · ~191% | As of May 2026 |
Through 2024, Gaonchips was that rare KOSDAQ combination: a growth company that made money. In 2025 the combination broke — revenue fell 29% and the operating line swung to a 16.7bn won loss. Understanding a design house’s cost structure tells you what kind of loss this is. The cost base is mostly design engineers, and you must hire them before projects are booked. When project revenue slips, the costs remain. So this loss supports two readings — “the business broke” or “they invested ahead of the order cycle” — and 2026–2027 mass-production revenue will decide which. The forecast that automotive customers’ new products enter volume production from 2026, bringing production royalties into the P&L, is that test.
The 191% debt ratio is the ugliest number on the balance sheet. This is not a debt-free company like Nable was. Order-driven businesses carry advances and lease liabilities, but a prolonged loss period makes that leverage genuinely expensive.
July 10: a thirty-minute rehearsal
On July 10, Gaonchips spiked 29.91% intraday to 55,600 won. There was no company disclosure. Theme money piled onto sector news about Samsung Foundry expectations. Two weeks later, on July 24, the stock sat at 40,000 won — essentially the entire spike returned.
The round trip that took Nable three months played out here in two weeks, compressed. Same lesson: nothing happened to the company; only the stock made the journey. The one difference is that in Gaonchips’ case, the space where “nothing” sits has actual, verifiable business events scheduled.
So — is it worth watching?
I don’t give buy/sell advice, and nothing in this piece is a recommendation. Here is what the public record supports.
Reasons to take it seriously. The dual official status of Samsung Foundry DSP plus Arm AADP; a 2nm project track record; profitability proven through 2024. As long as AI chip design demand grows, design-house workload grows structurally. And today’s price reflects the 2025 loss, not the story.
Reasons for skepticism. The loss is ongoing (−3.9bn won in Q1), and a 191% debt ratio puts a price on waiting. P/B of 8.5x means the stock is not remotely cheap on assets despite the “bottom” optics — even this price pre-pays much of the recovery. And as July 10 showed, the shareholder base is exposed to theme money, so fundamentals-free round trips can recur at any time, in either direction.
Checkpoints that actually change the picture. (1) Automotive projects’ mass-production revenue showing up as numbers in quarterly reports — a revenue line, not a forecast; (2) return to quarterly operating profit; (3) new 2nm wins announced as company disclosures. Until at least one lands, Gaonchips belongs — in my personal framework — on the watch list, not the buy list. Unlike Nable, though, this watch list comes with actual dates to check.
The recurring 2026 KOSDAQ pattern, once more: the market asks AI infrastructure no valuation questions and asks AI substance for earnings. If that asymmetry normalizes — brokerages expect KOSDAQ rotation after September — the first movers should be the companies whose names are on the contracts.
Nothing in this article is financial or investment advice. All figures come from public data as of July 24, 2026, may contain errors, and must be verified against primary disclosures before any decision. The author holds no position in any security discussed.
Sources
- Bosoop — Gaonchips (399720) price and financials snapshot
- Bullstory — Gaonchips intraday +29.91% and the semiconductor rally (2026.7.10)
- Orangeboard — Why Gaonchips surged: Samsung Foundry beneficiary?
- Wealthmoa — Gaonchips 2026 outlook: 2nm AI rebound after the swing to loss
- E-Today — Overlooked KOSDAQ looks to ‘AI data centers’ for its comeback