The most elegant business model in semiconductors is not making chips. Arm became the standard of the world’s mobile silicon on licensing and royalties for design assets (IP) alone. Korea has a listed company that aspires to that model: OpenEdges Technology (KOSDAQ 394280), a developer of memory-subsystem and NPU (neural processing unit) IP.
In April 2026 the company produced evidence that the model is more than talk: it won its first license deal for an interface IP that simultaneously supports the LPDDR6 and LPDDR5X memory standards — a first for a Korean company, in a technology only a handful of firms worldwide have commercialized. Locking down the interface for a next-generation memory standard opens the classic structure: royalties every time a chip built on that standard ships.
The stock, however, shows the market didn’t hold onto that news for long. July 24 close: 9,800 won — a whisker above the 52-week low of 9,460, and less than half the high of 22,700. Market cap: about 258 billion won. While AI infrastructure swept KOSDAQ’s top gainers in the first half, the company selling foundational on-device AI technology was printing new lows.
What OpenEdges actually sells
OpenEdges’ products sit on two axes. One is memory subsystem IP — the data pathway between a chip and its memory (controller, PHY, interconnect). Since the bottleneck in AI computation is usually memory bandwidth rather than compute, the value of this pathway rises as on-device AI grows. The other is NPU IP — blueprints for dedicated cores that run AI inference on edge devices.
Parts of the track record are verifiable. In 2023, its memory subsystem IP was adopted for Renesas’ next-generation MPU platform in Japan, and the same year it secured ISO 26262 certification, the global automotive functional-safety standard. Adoption by a global tier-one semiconductor company is the strongest quality proof an IP vendor can hold. There is also research commentary that it is in talks to supply NPU IP to Korean defense companies for AI chip localization — a potent story where sovereign AI meets defense, but one that must be read exactly as written: still at the word “talks.”
The numbers — what the IP model costs in time
| Metric | Value | Note |
|---|---|---|
| Share price (Jul 24) | 9,800 won | 52w range 9,460–22,700, bottom 2.6% |
| Market cap | ~258bn won | |
| 2025 revenue | ~18bn won (+17.7%) | Street estimate |
| 2025 operating loss | ~−27.5bn won | Loss larger than revenue |
| 2026 estimate (Korea Investment & Securities) | Revenue 29.7bn (+64.8%) · loss −4.8bn | ”A very different 2026” |
| P/B · BPS | ~14.4x · 683 won | Thin equity |
| Funding | Oct 2024 third-party placement, 60bn won |
The tension in this table compresses to one line: revenue of 18 billion won against an operating loss of 27.5 billion. The defense: IP businesses burn R&D payroll long before revenue arrives — Arm took years to turn profitable. The rebuttal: if losses exceed revenue years after listing, that is a problem of scale, not model. Korea Investment & Securities forecasts 2026 revenue of 29.7bn won (+65%) with the loss narrowing sharply to 4.8bn. On that path, 2027 breakeven comes into view; off it, the next question is funding.
Funding can’t be avoided. In October 2024 the company raised 60 billion won in a third-party placement. At a cash burn in the low-to-mid 20-billions per year, that granary has a lifespan, and a P/B of 14x (BPS of 683 won) means any next raise becomes, once again, a shareholder-value question. A 52-week low is most naturally read as the market doing precisely this arithmetic.
So — what price for the IP dream?
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. Verified engineering in the form of Korea’s first commercial LPDDR6 interface IP win; global references in Renesas adoption and ISO 26262; three tailwinds — on-device AI, sovereign AI, defense localization; and revenue that is actually growing (+17.7%, with +65% forecast). Once an IP model reaches orbit, royalties bend the margin curve non-linearly.
Reasons for skepticism. Operating losses exceeding revenue; thin equity (BPS 683 won) and the P/B illusion it creates; raise risk; and the fact that the NPU story’s key items remain at the “in talks” stage. Being near the low is not a valuation argument — for loss-makers, new lows are often just the path.
Checkpoints that actually change the picture. (1) Follow-on license deals after LPDDR6 — does the first win become a series; (2) the appearance of a royalty revenue line — the distinction between one-off licenses and recurring royalties is this entire business; (3) defense/automotive NPU “talks” converting into disclosed contracts; (4) a slowdown in quarterly cash burn. In my personal framework, OpenEdges stays on the watch list until (1) or (3) shows up as a disclosure.
If Nable was a story that substance never caught up to, OpenEdges is the reverse: real technology that hasn’t yet convinced the income statement. KOSDAQ in 2026 is punishing both the same way — but the punishments should end differently. The former needs the theme to come back. The latter needs the numbers to show up.
Nothing in this article is financial or investment advice. All figures come from public data and brokerage estimates 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
- ZDNet Korea — OpenEdges wins LPDDR6-based interface IP license deal (2026.4.9)
- Korea Investment & Securities — OpenEdges (394280): A very different 2026
- Bosoop — OpenEdges (394280) price and financials snapshot
- AI Fortunate — Why the stock doubled: OpenEdges flips the board with LPDDR6
- E-Today — Overlooked KOSDAQ looks to ‘AI data centers’ for its comeback