DEEP RESEARCH · QSI (큐에스아이)

QSI Deep Dive: A Compound-Semiconductor Platform Built on Four Growth Engines — LiDAR, AR/VR, InP Foundry & Micro-LED

How a single core competency — "compound-semiconductor IDM" — branches into LiDAR, metaverse, quantum computing, and next-gen display markets

Published: 2025-11-08 · Stock deep research · Original Naver blog post

0. Bottom Line First

• Unique weapon: Full EPI(MOCVD) → FAB → PKG verticalization — delivers customization, reliability, and extensibility together.
• Reference customer: Sole-source supplier of motion-sensing LDs to Valve (won after competitor reliability issues).
• Front expansion: 905nm 25–125W pulsed LDs, multijunction, 4-channel → long-range ADAS LiDAR.
• Platform leap: Proprietary 'INP-DHEMT-100' PDK already qualified for volume MMIC supply to a global Swedish cryo-LNA leader → next: 'INP-DHEMT-050' targeted for June 2026 (THz/6G, space & defense).
• Risks: CAPEX intensity; cash-flow, CB/BW balances, and R&D/CAPEX disclosure visibility is limited.

I. Company Profile — Evolving into a Compound-Semiconductor IDM

A. Core Identity — More Than a Component Maker

QSI is publicly known as an LD specialist, but its real edge is owning the full Design → Epitaxy → FAB → Package chain — an IDM model.

Interpretation: Most small-cap photonics players sit in fabless or OSAT. QSI runs the most capex-intensive, hardest-to-replicate EPI/FAB steps in-house — the platform that makes every adjacent move possible.

B. Strategic Implication of IDM — Reliability and Extensibility

In compound semiconductors, tiny epi-structure variations dictate wavelength, output power, and yield. End-to-end control means customer-specific epi tuning (wavelength, power, high-temp reliability) can be done fast — a decisive weapon for high-mix, low-volume production.

This reliability is what won the Valve partnership — when Valve required it, QSI moved its China site to Weihai near Valve's partner factory, cementing a real technology-partner relationship.

Crucially, III-V (GaAs, InP) process know-how and MOCVD assets accumulated for LDs are now being pivoted into compound-semi foundry and Micro-LED businesses.

C. Portfolio — Four Growth Axes

II. Cash Cow — Sole-Source AR/VR Sensor for Valve

A. Sole-Source Position at Valve

The clearest current growth engine is AR/VR. QSI is the sole supplier of motion-sensing LD sensors to Valve Corporation (operator of Steam, maker of Valve Index). Valve and HTC together (HTC Vive series) hold roughly 30% of the global VR device market.

Official fact: Cumulative shipments crossed 1 million units by October 2021, and H1 2021 related sales already hit 65% of the full-year 2020 revenue (Newspim, DataNet).

B. Technical Moat — Reliability Issues at Competitors Created the Opening

The sole-source win came from a competitor's reliability failures. AR/VR devices operate close to the user's eyes; sensor consistency and durability are non-negotiable. The most plausible failure modes (epi defects, FAB instability) require exactly the kind of full-process control QSI has — which is why it could clear Valve's QC bar. This is also a leading-quality reference point for the even more stringent automotive LiDAR market.

C. Market Tailwind — VCSEL / 3D Sensing

QSI's motion-sensing parts likely use VCSEL arrays. The global VCSEL market is forecast at roughly $1.4B–$3.0B in 2025, growing at a CAGR of 8.8%–21.3% to 2030 (P&S Intelligence).

III. Next Mainstream (1) — High-Power LiDAR LDs

A. LiDAR Technical Requirements (ToF · 905nm)

LiDAR is the autonomous car's "eye": laser pulses are fired, time-of-flight is measured, and a 3D point cloud is built. The wavelength market is split between 905nm and 1550nm — 905nm uses cheap Si detectors and is the core choice for ADAS L3–L5.

B. QSI's Portfolio — 905nm High-Power Pulsed LDs

QSI targets the high-performance, high-power end. Beyond standard 25/50/75W parts, it now offers 125W+ and 4-channel 125W products, all rated for 85°C operation (a baseline for AEC-Q automotive reliability).

Interpretation: 125W-class 905nm LDs signal entry into 200m+ long-range ADAS. Because 905nm faces eye-safety output limits, hitting long-range requires very high peak power within short pulse widths — direct proof of EPI/FAB design strength.

Legend: ○ (ES/PP/MP), * (development). Source: QSI Laser R&D page

C. Market Growth & Competition

The global LD market sits at about $8.58B–$10.15B in 2025 (Mordor, Precedence), heading to $13.7B–$29.7B by 2030–2035 at a CAGR of 9.9%–13.5%. Automotive LiDAR LDs grow faster at ~13.4% CAGR. Competition comes from full-stack giants like Lumentum and Coherent (formerly II-VI).

IV. Next Mainstream (2) — InP-HEMT Compound-Semi Foundry

A. The Most Disruptive Bet — InP Foundry

QSI is using its III-V IDM stack to launch a foundry service — not silicon (the TSMC/Samsung domain) but InP-based HEMT, an extraordinarily high-barrier process.

B. Why InP HEMT Wins — Ultra-High-Frequency, Low-Noise

InP HEMT delivers more than 2× the electron mobility of GaAs — the defining substrate for low-noise, ultra-high-frequency electronics.

Interpretation: Existing MOCVD/EPI/FAB assets get repositioned as a platform supplier for quantum/6G — markets with much higher moats and margins than LDs.

C. Tangible Milestones & Roadmap

Official fact: Proprietary 'INP-DHEMT-100' (100nm gate) PDK enabled MMIC supply and volume production qualification with a Swedish company holding a dominant position in the global cryogenic LNA market (Newsprime).

The next milestone is 'INP-DHEMT-050' (50nm gate) by June 2026 — the foundation for THz (6G) and space & defense customers. QSI also plans free MPW (Multi-Project Wafer) services and PDK releases to seed a domestic high-frequency semiconductor ecosystem.

V. Long-Term Option — Red Micro-LED Epi Platform

A. The Bottleneck — "Red Problem"

Micro-LED is the ultimate next-gen display: high brightness, low power, long lifetime. But unlike blue/green, InGaAlP-based red Micro-LEDs see external quantum efficiency (EQE) collapse below ~10μm chip size — the so-called "Red Problem", the biggest hurdle for high-resolution XR microdisplays.

B. QSI's Approach — High-Efficiency Epi Platform

QSI is not building finished panels; it is going after the most painful bottleneck — the epi wafer. The root cause sits in epi growth and sidewall losses, exactly the kind of work decades of red-LD (also phosphide-based) production has tuned its IDM stack for.

Official fact: Currently running a national R&D project targeting "EQE ≥20%, ~1×μm" for a standard platform-style red Micro-LED epi growth technology, with Samsung Display, LG Display, and Samsung Electronics participating as customer-side parties.

VI. Global Market & Competitive Landscape

A. Photonics Market Outlook

Ranges reflect different estimates across MRF, Precedence, Mordor, MMR, NextMSC, etc.

Note: VCSEL growth outpaces traditional LD growth — a tailwind directly under QSI's cash-cow line.

B. Competing With the Giants — Lumentum & Coherent

Interpretation: QSI does not try to win on economies of scale. Instead it picks ① customization-heavy niches (Valve), and ② emerging-platform markets the giants haven't seriously entered (InP foundry, red Micro-LED) — a "niche & new-market pioneer" strategy.

VII. Verdict & Strategic Recommendation

A. The Four Growth Drivers Compound Organically

The four engines are not independent — they branch out of one capability (compound-semi IDM):

Net: maximizing asset utilization — the MOCVD and process expertise needed for LDs is being stretched into much higher-value foundry and epi-wafer markets. Only an IDM can pull this off.

B. Risk — Capital Intensity & Disclosure Limits

• CAPEX risk: EPI/FAB operations demand large facility spending and ongoing R&D. All four roadmap items (125W LiDAR, 50nm InP PDK, Micro-LED ramp) require more investment.
• Limited financial visibility: The 2024 annual report shows summary numbers (operating profit ₩1.18B, net profit ₩3.25B), but cash-flow statements, R&D/CAPEX breakdowns, and CB/BW outstanding balances are not clearly available in the source materials.

C. Opportunity

D. Final Take

QSI is running a sophisticated playbook: prove reliability in a validated market (AR/VR), compete on performance in the next mainstream (LiDAR), and pre-empt future-platform markets (InP foundry, Micro-LED). Valuation hinges on ① InP foundry volume qualification and ② red Micro-LED epi platform success, not today's LD/sensor revenue. The right frame is "high-risk, high-return technology platform". Monitor R&D/CAPEX execution and cash flow closely to validate whether this excellent technology roadmap is financially executable.