SemiScope follow-up. The first OpenEdges note framed the company as Korea’s rare memory-subsystem IP platform with a long royalty J-curve. This second piece narrows the investable question: OpenEdges Technology is not primarily an “NPU stock.” It is a Samsung 4/5/8nm LPDDR5X/LPDDR6 memory-subsystem IP option, and the stock only deserves a larger position when license wins, revenue recognition, royalty mix and cost control start to confirm that option.

Related Reading

• SemiScope: OpenEdges Technology - Korea’s Memory Subsystem IP Platform With a Royalty J-Curve Option
• SemiScope: Three Korean Memory ATE & IP Names Re-Ranked - Why Exicon is the 2026 J-Curve Trade, Not the Afterthought

TL;DR

1. OpenEdges is better understood as a memory-subsystem IP company than as an NPU company. The core investment lens is LPDDR5X/LPDDR6 PHY, memory controller and NoC bundled for AI ASIC, edge AI, automotive and other data-intensive chips.
2. Samsung Foundry 4/5/8nm is the practical battleground. The stock is not a bet that OpenEdges defeats Synopsys or Cadence at the global 2/3nm frontier. It is a bet that Samsung’s mid-to-advanced mass-market nodes need credible, local, process-aware LPDDR IP.
3. The value chain is still early: porting, license, tape-out, silicon proof, production, royalty. OpenEdges has moved beyond pure R&D, but it is not yet a mature royalty platform.
4. TSS and the Japan unit strengthen the technology base. OpenEdges Square is a higher-risk option. TSS and Japan are 100% subsidiaries tied to DDR PHY and memory-controller depth. Square is more interesting but also more dilutive, because the parent owns roughly 65%.
5. Valuation already prices in a lot. Around KRW 530B market cap against roughly KRW 30.4B-31.8B of 2026F revenue implies about 16.7x-17.5x sales. This can work only if LPDDR6/5X wins, 2Q-3Q revenue ramp, cost discipline and royalty progress arrive.

Bottom line: OpenEdges remains a Watchlist / Pilot Hold candidate. A 0.3%-0.5% pilot position can be justified for investors who want exposure to Korea’s semiconductor IP layer. Scaling above 1% should wait for repeated LPDDR6/5X wins, visible revenue ramp, lower fixed-cost burn and better visibility on OpenEdges Square.

1. What This Note Adds to the First SemiScope OpenEdges Piece

The first SemiScope note made the broad case: OpenEdges Technology (394280 KQ) is Korea’s rare semiconductor IP platform, with memory controller, DDR PHY, NPU, NoC, UCIe and CXL-adjacent memory IP under one roof. That view still stands.

But the more useful follow-up is narrower.

The market often wants to label OpenEdges as an “AI semiconductor” or “NPU” stock. That is too loose. The company’s most investable asset is not the NPU by itself. It is the memory path around AI compute:

DRAM - DDR PHY - Memory Controller - NoC - NPU / CPU / Accelerator

For an AI ASIC or edge inference chip, compute is not the only bottleneck. Data movement can be just as important. If the memory interface is slow, power-hungry or unstable, the chip’s theoretical TOPS do not matter much. That is why LPDDR5X/LPDDR6 PHY and controller IP can become strategic.

The revised thesis is:

OpenEdges is a call option on Samsung 4/5/8nm AI and edge ASIC customers needing proven LPDDR5X/LPDDR6 memory-subsystem IP.

This wording matters. It keeps the upside but removes the exaggeration. The company is not yet a global IP champion. It is not yet a high-quality earnings compounder. It is an R&D-heavy IP vendor with a credible chance to become the domestic memory-subsystem layer for a part of Samsung Foundry’s customer base.

2. Correcting the Product Frame: NPU Is Not the Center

OpenEdges owns and develops several IP blocks:

The strongest version of the company is not “we sell an NPU.” It is “we reduce memory bottlenecks in AI chips by selling a bundled memory subsystem.”

That is also where switching cost appears. A standalone soft IP block can often be swapped. A PHY plus controller plus NoC combination that has been integrated, verified and timed inside a chip is much harder to replace. The customer’s tape-out schedule, verification plan, power budget and timing closure can all be affected.

This is why the stock should be separated from equipment names such as NeoSem and Exicon. Those companies can convert orders into revenue within roughly six to nine months. OpenEdges sells upstream design IP. Its payoff takes longer, but the royalty upside can be more durable if customer chips reach mass production.

3. The IP Value Chain: Do Not Confuse Porting With Royalties

The most important operating sequence is:

Porting - license win - customer tape-out - silicon proof - production - royalty

OpenEdges has progressed beyond “technology with no customer interest.” Its 2026 IR material points to more than 30 customers, 71 cumulative license contracts, more than 20 saleable IP products and a workforce of about 170 people, including roughly 149 R&D staff. That is not a paper-company profile.

But it is equally important not to jump from “porting” to “royalty platform” too quickly.

The royalty mix is still small. Earlier OpenEdges materials and sell-side work point to royalty revenue at only a tiny share of recent sales. One 2026 forecast set royalty revenue around KRW 1.7B against KRW 30.4B of revenue, or about 5.6%.

That 5%-6% range is progress from the 1H25 level, but it is not yet the ARM-style curve investors ultimately want. For this stock to rerate into a true quality IP compounder, I would want to see royalty mix moving toward 20% of quarterly revenue and still rising.

4. Why Samsung 4/5/8nm Matters

The phrase “Samsung 4/5/8nm” should not be read as a vague Samsung ecosystem story. It is specific.

The incorrect shortcut is:

OpenEdges has pre-ported LPDDR6 across Samsung 4-8nm.

The better version is:

OpenEdges is expanding coverage around Samsung 4nm LPDDR6 and Samsung 4/5/8nm high-speed LPDDR5X, targeting AI, edge, automotive and other ASIC customers that need better memory interfaces on Samsung Foundry mass-market nodes.

According to the 2026 IR roadmap cited in the source material, OpenEdges already provides Samsung 5nm LPDDR5X/LPDDR5/LPDDR4X/LPDDR4 coverage. Samsung 8nm LPDDR5X was shown as a 1H26 development item, while Samsung 4nm LPDDR5X 10.7Gbps and LPDDR6 14.4Gbps were shown as 2H26 development items.

This is the right battleground for OpenEdges.

The 2/3nm frontier is where global giants dominate. Synopsys, Cadence, Arm, Rambus and other interface-IP leaders have deeper field support, broader foundry relationships and more advanced-node references. It is not realistic to underwrite OpenEdges as a near-term global winner at that layer.

The more realistic opportunity is the mid-to-advanced mass-market:

• on-device AI;
• automotive AI chips;
• robot and industrial AI controllers;
• IoT and smart-home chips;
• security-camera and vision ASICs;
• defense and special-purpose chips;
• domestic AI accelerator projects;
• design-house-led turnkey ASIC programs.

These customers may not always need the most expensive 2/3nm process. They may care more about cost, power, memory bandwidth, local support, risk reduction and Samsung-process availability. That is where a domestic memory-subsystem IP vendor can matter.

The core question is:

Does the pool of Samsung 8nm-and-below customers needing LPDDR5X or LPDDR6 grow enough to create repeated license wins for OpenEdges?

If yes, the stock has a clear path to a higher-quality narrative. If no, the company remains a technically interesting but financially strained IP vendor.

5. Moat Inside the Samsung Value Chain

Samsung Foundry’s issue is not only process technology. It also needs a thicker ecosystem: IP, EDA, design houses, verification references and customer support. OpenEdges can help strengthen one layer of that stack.

The first moat layer is DDR PHY porting. PHY is not generic software. It is a hard IP block with analog and mixed-signal complexity. At 8.5-14.4Gbps speeds, signal integrity, jitter, noise, voltage/temperature variation, timing margin and training algorithms matter. A simulation-only IP block is not enough. Customers want silicon proof.

The second moat layer is PHY + controller + NoC bundling. AI ASIC customers do not simply need an interface pinout. They need a working data-movement system. If OpenEdges can sell the memory path as a bundle, it creates a higher switching cost than any single block can create alone.

The third moat layer is design-house channel leverage. If a design house repeatedly uses OpenEdges IP inside Samsung-turnkey projects, OpenEdges does not need to sell every customer one by one. The IP can become part of a repeatable package.

The moat is not absolute. It is local and process-specific. But local, process-specific moats can still create attractive public-market returns when the revenue base is small.

6. Subsidiaries: What Belongs to Shareholders?

The subsidiary structure matters because OpenEdges is not only one listed entity with one clean product line. The 2026 IR material breaks the group into the parent and key subsidiaries.

TSS should be viewed as a moat asset, not dilution. DDR PHY is the hardest part of the stack. If TSS deepens high-speed PHY capability, the economics accrue to the parent because it is wholly owned.

The Japan unit is similar. Memory-controller IP is less analog-heavy than PHY, but it sits close to customer architecture and product requirements. Japan also has relevant semiconductor engineering depth and customer relationships. Again, because this unit is 100% owned, the issue is cost, not value leakage.

OpenEdges Square is different.

Square is not merely an engineering branch. It was launched around online IP sales platform ambitions and CC NoC development. CC NoC, or cache-coherent network-on-chip, can become valuable if multicore AI ASICs require more sophisticated internal data movement and cache-coherency behavior.

The option is interesting. The ownership is less clean. If the parent owns about 65.4% of Square, then an OpenEdges shareholder’s look-through exposure to Square is only 65.4% of their parent-company stake.

Example:

1.0% parent stake x 65.4% Square ownership = 0.654% look-through Square exposure

That is not automatically bad. External capital can reduce the parent’s cash burden and accelerate development. But it means Square’s success does not accrue 100% to listed shareholders.

I would treat Square as a call option with two questions:

1. Can CC NoC become a real commercial product rather than a development story?
2. Can Square’s cost and financing needs stay controlled enough that the option does not weaken the parent income statement?

7. Dilution and Funding: The 2026 CPS Issue

OpenEdges raised roughly KRW 20.0B in March 2026 through a third-party allotment of convertible preferred shares. The key figures in the source material are:

The dilution math is:

1,145,278 / (26,276,655 + 1,145,278) = 4.18%

This is real dilution. It should not be ignored. The softer points are that the issuance was at a premium to the reference price and the cash extends R&D runway.

The risk is refixing. If the share price falls and the conversion price adjusts down toward the floor, effective dilution can increase. For a company still funding R&D ahead of mature royalties, the cost of capital remains part of the thesis.

8. Valuation: The Stock Already Prices a Better Future

Using the source-market snapshot around Apr. 29, 2026, OpenEdges traded near KRW 20,150 with market capitalization around KRW 530.9B. Against 2026F revenue assumptions of roughly KRW 30.4B-31.8B, the sales multiple is high.

This is not a cheap cyclical small-cap. It is an expensive option on a better future business model.

The market is already underwriting several things:

• 2026 revenue moves into the KRW 30B+ range;
• the company approaches breakeven in 2H26 or 2027;
• LPDDR6/5X wins become repeatable;
• Samsung 4/5/8nm customer activity improves;
• OpenEdges Square contributes option value rather than only cost;
• royalty revenue gradually becomes visible.

That is a demanding setup. The right conclusion is not “avoid.” The right conclusion is “do not scale until the company proves enough of the roadmap.”

9. Investment Framework

Idea 1 - Samsung 4/5/8nm LPDDR memory-subsystem IP

Idea 2 - TSS and DDR PHY as the real moat

Idea 3 - OpenEdges Square and CC NoC

10. Positioning: Watchlist / Pilot Hold

For now, I would put OpenEdges in the Watchlist / Pilot Hold bucket.

The stock is investable as a small exploratory position because the upside path is clear and the addressable strategic layer is real. But the current valuation is too high for blind scaling. The company needs to show that its technology roadmap can turn into repeatable contracts and recognized revenue.

Conditions to increase exposure

I would consider moving from a 0.3%-0.5% pilot position toward 0.5%-0.7% if at least two of the following occur:

1. additional LPDDR6/5X license wins;
2. 2Q26 or 3Q26 quarterly revenue above KRW 7.0B;
3. quarterly fixed costs move closer to the KRW 7.0B-8.0B range;
4. Samsung 4nm LPDDR6 tape-out or silicon proof is disclosed;
5. design-house channel wins become more visible;
6. OpenEdges Square shows cost control and a credible CC NoC launch path;
7. the share price holds KRW 20,000 and recovers through the KRW 21,500-22,000 range on real news.

I would only consider a 1%+ position after three or more of those milestones are visible.

Catalysts

Invalidation

11. Final Note

OpenEdges is a good technology option. It is not yet a good earnings company.

The cleanest investment thesis is not “Korea’s NPU winner.” It is LPDDR5X/LPDDR6 PHY and controller IP for Samsung 4/5/8nm AI and edge ASIC customers. That is narrower, but also stronger. It identifies the real bottleneck, the real customer pool and the real reason OpenEdges can matter despite global IP giants.

TSS and the Japan unit strengthen the core IP base. OpenEdges Square adds a separate CC NoC and IP-platform option, but that option is only partly owned by the listed parent and must be monitored for cost discipline. The 2026 CPS financing gives the company runway, but also adds dilution risk.

At KRW 500B+ market capitalization and roughly 17x 2026F sales, the stock already assumes a better future. The job now is not to admire the technology. The job is to verify the conversion chain: LPDDR6/5X wins, tape-out, silicon proof, revenue recognition, breakeven and eventually royalties.

My conclusion is unchanged but sharper: Watchlist / Pilot Hold. Scale only after the numbers start confirming the IP story.

Source Notes

• Company IR Book: OpenEdges Technology 2026.03 IR Book via KIND, including product roadmap, customer count, license history, employee mix and subsidiary structure.
• OpenEdges official release on OpenEdges Square launch: OpenEdges Technology launches subsidiary OpenEdges Square.
• The Bell coverage on OpenEdges and Square: OpenEdges Technology and Square synergy story.
• DigitalToday coverage on OpenEdges Square capital increase and ownership change: OpenEdges Technology participates in OpenEdges Square capital increase.
• Design-Reuse coverage on LPDDR6/5X commercialization: OpenEdges advances commercialization of LPDDR6/5X Memory Subsystem IP.
• Market snapshot used for valuation reference: Maeil Business stock page for OpenEdges Technology.

Disclaimer: For research and information purposes only. Not investment advice. Names cited are for analytical illustration; readers should perform their own due diligence and consult licensed advisors before any investment decision.