Will Semiconductor (OmniVision)

Phase A — Understand the business

Company Overview

What it is. Will Semiconductor — renamed OmniVision Integrated Circuits Group in mid-2025 — is a fabless designer of CMOS image sensors (CIS) and the world's #3 CIS supplier by revenue, behind Sony and Samsung. Its HK IPO prospectus put its 2024 global CIS revenue share at 13.7%.

Origin & the deal that made it. Founded 15 May 2007 by Yu Renrong as a semiconductor distributor and designer of discrete protection/power parts (TVS, MOSFET, Schottky, power-management ICs). IPO'd on the Shanghai main board in 2017. The company was transformed by its May 2019 acquisition of OmniVision Technologies (a US CIS pioneer, HQ Santa Clara, CA) plus Beijing OmniVision (Superpix) for ~$2.2bn / ~RMB 15.2bn. OmniVision had itself been taken private off Nasdaq in 2016 by a Chinese consortium (Hua Capital, CITIC Capital, Goldstone). That acquisition vaulted Will from a components distributor into a top-3 global CIS designer overnight.

How it makes money. Three reporting groups:

1. Image-sensor solutions (CIS) — the core. 74.65% of 2024 revenue (≈RMB 19.19bn) and 87.65% of 2024 gross profit. Sensors for smartphones, automotive (ADAS + cabin), security/surveillance, medical endoscopy, notebooks, and emerging AR/VR.
2. Touch & Display solutions — TDDI and display-driver ICs (DDIC).
3. Analog solutions — power-management, discrete protection devices, plus the legacy semiconductor-distribution business.

Groups 2–3 are ~25% of revenue but only ~12% of gross profit — i.e. low-margin ballast that dilutes the blended margin.

Customers & contract structure. End-market split within the image-sensor segment is roughly mobile 51% · automotive 31% · security 8% · medical 4% · emerging 6% (this reconciles: 51% × RMB 19.19bn ≈ RMB 9.8bn, matching the disclosed 2024 smartphone-CIS figure). Buyers are Chinese Android OEMs (Honor, Xiaomi, OPPO, vivo, Transsion), automotive Tier-1s/OEMs, and surveillance vendors. No take-or-pay; standard PO-based semiconductor supply. Customer names beyond the segment mix are n/a — not disclosed at line-item level.

Supply Chain

Upstream → OmniVision → end customer, named actors:

• Foundry (the chokepoint). OmniVision is fabless — it owns no wafer fab. TSMC is the primary foundry for its CIS logic and pixel wafers; the flagship 0.56µm pixel runs on a CIS-dedicated 28nm process + 22nm logic node at TSMC. TSMC produces >60% of all fabless-designed CIS wafers industry-wide — so OmniVision's leading-edge roadmap is structurally dependent on TSMC allocation and pricing. Historic TSMC–OmniVision packaging JV (VisEra) handles CIS-specific color-filter/microlens/wafer-level steps.
• Process / packaging. BSI (backside-illuminated) and stacked (wafer-bonded) CIS using OmniVision's PureCel Plus-S stacking and TheiaCel (LOFIC-based HDR) technologies. Back-end assembly/test via OSATs (specific partners n/a — not disclosed).
• Design centres. US (Santa Clara — inherited OmniVision R&D talent), China (Shanghai/Beijing).
• Downstream. Sensors → camera-module makers (Sunny Optical, O-Film, Truly, etc. — inferred, n/a — not disclosed at contract level) → smartphone OEMs / automotive Tier-1s / surveillance brands.

Single-source dependency: TSMC for leading-edge CIS is the critical node. A US–China constraint on OmniVision's TSMC access (given PRC ownership) is the sharpest supply-chain tail risk (see Lens 10/13).

Competitive Advantages (moats)

• Scale + top-3 incumbency. #3 globally at 13.7% CIS revenue share — a design-win franchise and process know-how (BSI/stacked pixel, sub-0.6µm) that Chinese challengers are still climbing toward. OmniVision led the pixel-shrink race with the world's smallest 0.56µm pixel (2022).
• Automotive design-in stickiness. Automotive CIS is qualified per-platform (AEC-Q100, multi-year design cycles, functional-safety). OmniVision has design wins with "nearly every major OEM" for ADAS + cabin monitoring and co-developed the OX08D20 capture scheme with Mobileye. Auto sockets, once won, are multi-year annuities — the most durable moat it has.
• US R&D + IP estate. Inherited OmniVision's Santa Clara engineering base and patent library; it won a non-infringement judgment vs VisionX Technologies (Nov 2024), i.e. it actively defends IP.
• Bargaining power — mixed. Weak over TSMC (needs TSMC more than TSMC needs it) and weak in flagship smartphone (Sony owns the Apple main-camera tier). Moderate over Chinese Android OEMs and automotive Tier-1s where it is a qualified top-2/3 alternative to Sony/onsemi.

Moat verdict: a real but mid-tier moat — process scale + automotive design-in are genuine; but it is the #3 in a market where #1 (Sony) has ~3–4x its share and owns its own fabs, and where domestic challengers attack from below on price.

Segments

Revenue by group (2024):

Within CIS, by end-market (approx.): mobile 51% · automotive 31% · security 8% · medical 4% · emerging 6%.

Trend & cause:

• Smartphone CIS ≈RMB 9.8bn in 2024, +26% YoY — recovery driven by high-end design wins (50MP+ flagship sensors at Chinese Android OEMs) after two lean years. Re-accelerating off a trough.
• Automotive — the structural grower. OmniVision claims ~26% of the global automotive CIS market by end-2025 (other sources cite ~18%; call it a clear top-3 auto position). Driver: rising camera-count per vehicle for ADAS/L2+ autonomy. Accelerating, secular.
• Security & medical — smaller, higher-margin niches (endoscopy is a durable annuity). Stable/mix-accretive.
• Touch/Display + Analog + distribution — flat, low-margin, dilutive.

Geography: predominantly China + Asia demand; n/a — precise geographic revenue split not sourced.

---

Phase B — Measure performance

Earnings Result

Five-year income statement (RMB millions):

The story the table tells: a classic semiconductor cycle. 2021 peak → 2022–23 inventory-glut trough (Android CIS oversupply; gross margin collapsed from 33.7% to 19.9% as prices were cut and inventory written down; net income fell ~88% then a further ~44%) → 2024–25 recovery on flagship-smartphone design wins + automotive. FY2025 (reported 2026-03-30): revenue RMB 28.86bn (+12.1%), net income RMB 4.05bn (+21.7%).

Critical nuance — the recovery is INCOMPLETE at the margin line. 2025 net income (RMB 4.05bn) is still below the 2021 peak (RMB 4.48bn) despite revenue being ~20% higher, because gross margin (29.94%) has not reclaimed the 33.74% peak. Operating leverage recovered; pricing power did not fully.

The freshest print is a warning. Q1 2026 (qtr ended 31 Mar 2026): revenue ≈RMB 6.41bn, −0.9% YoY (flat); net profit ≈RMB 0.50bn, down ~42% YoY from Q1-2025's RMB 0.87bn. (One feed shows ~RMB 0.62bn on a different basis; both are down sharply.) Management framed it as "in line, sequential improvement expected from Q2". Read plainly: the easy YoY comps are over, revenue growth stalled to flat, and profit went backwards — the first stumble of the recovery.

Balance sheet (latest): cash RMB 16.86bn, total debt RMB 10.07bn → net cash +RMB 6.79bn. Net-cash, well-funded, further reinforced by the HK$4.8bn ($603–616m) HK IPO in Jan 2026. No leverage stress.

Market reaction / what's priced in: A-share sits ~RMB 103 with a Strong-Buy street and PTs of RMB 131–166 — yet consensus just cut 2026 EPS from RMB 3.87 → RMB 3.27, i.e. below 2025's actual RMB 3.37. The market is paying a growth multiple for a year of flat-to-down EPS.

Earnings Calls (sentiment trend)

No transcripts in the research layer (transcripts/ empty; A-share filer). From public commentary:

• 2024 calls: euphoric recovery tone — "high-end smartphone breakthrough," "automotive share gains," margin-recovery narrative.
• Late-2025: confident but pivoting the story toward automotive + medical mix as the durable driver as mobile normalizes.
• Q1-2026 note: tone shifted to reassurance — "results in line," "sequential improvement from Q2," de-emphasizing the flat top line. The disappearance of the "another record quarter" cadence is itself the signal: management is now managing expectations down to a step-change, not extrapolating the 2024–25 rip. — treat as summary-derived, not verbatim.

Comps

Peer multiples — ** with source/date or n/a; nothing fabricated.**

All figures. Sony and Samsung are conglomerates where CIS is one division — a clean standalone CIS multiple is n/a; do not infer one. 5-yr average ROE for OmniVision ≈ ~10%; peers n/a. Dividend yield n/a (Chinese semis typically pay a token payout).

Read: OmniVision (35x) and SmartSens (36x) trade at near-identical multiples — the market is not discounting OmniVision's #3-incumbency vs the smaller domestic challenger; it is paying up for both. onsemi's headline P/E is distorted by trough auto/industrial earnings (fwd ~32x is the cleaner read). On forward P/E, OmniVision (~29x), SmartSens (~29x) and onsemi (~32x) are clustered — OmniVision is not cheap relative to peers, and priced for continued growth it just guided flat.

Stock-Price Catalysts (last ~5 years)

Reconstructed from the earnings path + known events (; A-share tick-by-tick history not fully retrievable):

• 2021 — melt-up: peak-cycle earnings (net income RMB 4.48bn), CIS shortage, A-share semi mania.
• 2022 — crash: Android inventory glut + China smartphone slump; net income −78%; the stock de-rated hard alongside the earnings collapse.
• 2023 — trough: gross margin bottomed at 19.9%, net income RMB 556m; sentiment washed out.
• 2024 — recovery rally: +498% net-income snap-back on flagship design wins → strong re-rating.
• 2025 — automotive + rename + record year: RMB 28.9bn revenue; renamed OmniVision Integrated Circuits Group.
• Jan 2026 — HK IPO catalyst: first A+H dual listing of 2026; H-shares priced HK$104.80, +16% debut to HK$121.80, raised ~$603m.
• Apr 2026 — soft Q1: flat revenue, −42% profit; consensus EPS cut.

Pattern: this name trades as a pure earnings-cycle proxy — it re-rates violently with the CIS/smartphone cycle and inventory correction, amplified by A-share retail momentum. The 2024–25 move was an earnings recovery, not a re-rating on new secular economics. What moves it: the smartphone-CIS cycle and margin direction — not any single customer.

---

Phase C — Judge people & books

Management

• Yu Renrong (虞仁荣) — founder & Chairman. Built Will from a 2007 distribution startup into the world's #3 CIS designer via the 2019 OmniVision deal — a genuinely impressive operator + dealmaker track record. Forbes fortune ~$5.8bn. One of China's most politically prominent chip figures (attended Xi Jinping's semiconductor symposium as an industry representative).
• Skin in the game — high. Holds ~31.3% (25.09% direct + 6.13% via Shaoxing Weihao). Strong founder control and alignment.
• Capital allocation: the OmniVision acquisition was transformational and value-creating (turned a components house into a top-3 franchise). Balance sheet kept net-cash through the trough. 70% of HK IPO proceeds → R&D — reinvestment-first, appropriate for a design leader. R&D ≈ >12% of revenue.
• Red flag to weigh: in 2024 Yu gifted RMB 4.9bn (~$688m) of company stock to the Eastern Institute of Technology (EIT) Ningbo — a university he founded. Laudable philanthropy, but a related-party transfer of this scale (the EIT foundation now holds ~4.21% ) is a governance item minorities should note — it moves a large block into a founder-affiliated entity.
• Archetype: founder-owner-operator with political capital — a strength for navigating China's chip-self-sufficiency drive, a concentration risk if succession or policy alignment shifts.

Forensic Red Flags

Forensic lens — every figure labeled; A-share accounting, no EDGAR.

1. Goodwill / intangibles from the 2019 deal — the biggest latent item. The ~$2.2bn / RMB 15.2bn OmniVision purchase loaded the balance sheet with material goodwill + acquired intangibles. Exact current carrying value n/a, but note: it was NOT impaired through the 2022–23 earnings collapse (net income fell to RMB 556m) — a period when an impairment test would have been most pressured. Carrying a large, un-impaired acquisition goodwill through a trough is a classic latent write-down risk; watch the annual impairment test.
2. Gross-margin volatility = inventory/pricing risk. GM swung 33.7% → 19.9% → 29.9% in four years. The 2022–23 collapse was inventory-writedown-driven. A fabless CIS maker into a consumer cycle carries recurring inventory-obsolescence exposure; the next downcycle repeats the pattern absent tighter demand visibility.
3. Margin-mix optics. Blended GM ~30% is dragged by the low-margin distribution/analog tail; image-sensor-only GM is higher (a content-mill cited "~34%" ). Do not read a blended 30% as the CIS economics — but also don't accept the un-sourced 34%. Segment-level GM disclosure would resolve it; n/a.
4. SBC / non-GAAP gap. 2024 non-GAAP net RMB 3.06bn vs GAAP RMB 3.32bn — non-GAAP was lower, i.e. no obvious SBC-flattering of adjusted numbers (reassuring).
5. Cash vs earnings: net-cash +RMB 6.79bn; no obvious cash-vs-earnings divergence surfaced, but full cash-flow statement n/a.

Regulatory findings (required sub-section).

• SEC (EDGAR): regulatory/regulatory-findings.md (fetched 2026-07-10) — 0 findings. OmniVision has no CIK; it is not an SEC filer, so no LR/AAER search is possible.
• China (CSRC/SSE) — the relevant regulator: targeted web search surfaced no CSRC investigation, administrative penalty, or securities-fraud finding against the listed entity. (CSRC actions in results named other companies — Futu, UP Fintech — not OmniVision.)
• US enforcement / litigation ((FTC OR DOJ OR FDA OR CFPB OR consent decree OR settlement OR fine OR penalty)): no material government enforcement action found. Relevant items: (a) CFIUS cleared the 2019 Beijing-OmniVision acquisition (Apr 2019) — a positive clearance, not an action; (b) OmniVision received US export licenses to supply Huawei during the export-control era — confirms exposure to, and navigation of, US–China controls; (c) patent-litigation WIN vs VisionX Technologies (non-infringement, Nov 2024) — routine IP defense; (d) data breach disclosed May 2024 following a 2023 Cactus ransomware attack on OmniVision Technologies — an operational/cyber incident, not a securities matter.
• 10-K Item 3: n/a — no 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes. exists (not an SEC filer).
• Net: No material securities-regulatory or accounting-enforcement findings across SEC EDGAR (0), CSRC web search, and US-agency web search as of 2026-07-10. The live risks are the un-impaired acquisition goodwill and US–China export-control exposure of the US-domiciled OmniVision subsidiary — structural risks, not enforcement facts.

---

Phase D — Project & stress-test

Forward Projection (FY2026–FY2028, RMB)

Built bottom-up from FY2025 actuals (rev RMB 28.86bn, net RMB 4.05bn, EPS RMB 3.37, ~1.26bn shares), the soft Q1-2026 print, and cut consensus (2026 rev RMB 30.8bn / EPS RMB 3.27). Outputs, arithmetic shown.

Base-case arithmetic (2026): rev RMB 28.86bn × 1.10 ≈ RMB 31.7bn; net margin ~13% → net ≈ RMB 4.1bn ÷ 1.26bn sh ≈ EPS RMB 3.3 — essentially flat on 2025. At RMB 103, that is ~31x forward on the base case — no margin of safety for a flat year.

Fair-value frame: a fair multiple for a top-3-but-cyclical CIS name printing flat EPS is ~25x → ~RMB 82–88 on base EPS. The street's RMB 131–166 PT implies re-rating and a bull earnings path both landing. Skew is unfavorable at RMB 103.

(No our model logged — unattended --watchlist mode; the base call is recorded here only.)

Bull vs Bear

Bull case. OmniVision is the only credible non-Sony/Samsung scale player in CIS, and it sits on the two best secular vectors in imaging: automotive (camera-count per car compounding with ADAS/L2+, multi-year sticky sockets, Mobileye co-development, ~top-3 auto share heading toward ~26%) and high-end mobile (50MP+ TheiaCel/LOFIC wins clawing share back inside the Chinese Android flagship tier). It is founder-controlled (31%), net-cash, R&D-first (>12% of sales), freshly capitalized by a well-received HK IPO, and a national champion inside China's chip-self-sufficiency policy tailwind. If auto mix structurally lifts blended GM back toward the 2021 peak, the earnings power is materially above today's run-rate.

Bear case (2–3 permanent-impairment risks). (1) It is structurally #3 in a market where #1 (Sony, ~45%) owns its fabs and the premium iPhone socket — OmniVision is a price-taker at the top and a share-defender at the bottom. (2) Domestic challengers (SmartSens +109% rev in 2024; GalaxyCore) attack from below at ~20% price discounts, capping the very GM recovery the multiple assumes. (3) US–China export-control exposure of the Santa-Clara-domiciled OmniVision subsidiary + TSMC-allocation dependence is an unhedged geopolitical tail. Expectations baked in: ~35x trailing / ~31x base-forward for a company that just guided flat revenue and −42% profit.

Pre-mortem (18 months out, thesis broke). It's early 2028. The 2024–25 recovery proved to be a cyclical bounce, not a re-rating — mobile rolled into a down-cycle, SmartSens/GalaxyCore took security + entry-auto share and forced CIS price cuts, blended GM slid back toward 26%, and consensus EPS never exceeded 2025's RMB 3.37 for two years. A modest goodwill write-down on the aging OmniVision intangibles clipped a quarter. The stock de-rated from 35x to ~20x on flat earnings — a ~40% drawdown — and the HK premium evaporated.

Contrarian view (what the market refuses to see): the tape treats this as a secular AI-sensing compounder; the five-year P&L says it is a high-quality cyclical whose margins have never returned to the last peak, now lapping impossible comps. The Strong-Buy consensus is anchored to the 2024–25 recovery slope, not the flat 2026 it just modeled.

Devil's Advocate (short-seller)

• What breaks the model: a mobile-CIS down-cycle (its 51%-of-CIS anchor) colliding with domestic price competition — the exact 2022–23 mechanism, and Q1-2026 (−42% profit) may be the leading edge.
• Revenue concentration: heavy on Chinese Android OEMs and China demand; a China smartphone air-pocket or a share loss at one flagship program hits hard. Auto is growing but still ~31% of CIS — not yet big enough to offset mobile.
• Why the moat is thinner than bulls think: OmniVision does not own fabs (Sony does) and lost the flagship-iPhone main-camera socket years ago; its "moat" leans on TSMC's process — a dependency, not a proprietary edge. SmartSens/GalaxyCore are closing the pixel/HDR gap.
• Most dangerous competitor bulls underestimate: SmartSens — +109% revenue in 2024, same 200MP/large-format roadmap, trading at the same multiple, taking security and pushing into auto at a discount.
• Capital-allocation flag: the RMB 4.9bn founder stock gift to his own university — generous, but a large related-party block movement.
• What must hold for RMB 103: blended GM holds ~30%+ AND auto compounds ~25%+ AND mobile doesn't roll — a three-variable bet, and one (mobile) just wobbled.
• −20–30% growth-disappointment scenario: if 2026 revenue is flat (Q1 already is) and GM slips to ~27%, EPS ~RMB 2.6 → at a de-rated 22x that's ~RMB 57, ~45% downside.
• Single permanently-impairing scenario: a US tightening that curtails OmniVision's TSMC leading-edge access or forces separation of the US subsidiary — low probability, high severity, and entirely outside management's control.

Management Questions (ordered by information value)

1. Q1-2026 net profit fell ~42% YoY on flat revenue — is that pure comp/normalization, or renewed price competition from SmartSens/GalaxyCore, and where does blended gross margin settle structurally (can it exceed the 2021 peak of ~34%)?
2. Automotive is ~31% of CIS and the growth engine — what is the design-win backlog and content-per-vehicle trajectory, and can auto mix alone lift blended GM even as mobile normalizes?
3. You lost the primary flagship-iPhone sensor slot — is there any realistic path back into Apple's main camera, and how dependent is the mobile recovery on Chinese Android OEMs specifically?
4. TSMC is your single leading-edge CIS foundry — how exposed are you to its allocation/pricing, and what is your second-source / China-foundry contingency given PRC-ownership constraints?
5. OmniVision Technologies is US-HQ'd (Santa Clara) under PRC control — how do you insulate US R&D talent, IP, and export licenses from further CFIUS/export-control action?
6. What is the current carrying value of goodwill + acquired intangibles from the 2019 deal, and the impairment-test sensitivity — why was none taken through the 2022–23 trough?
7. 70% of HK IPO proceeds are for R&D — which programs (TheiaCel/LOFIC, 3-layer stacking, 200MP, in-sensor compute) and at what ROI hurdle?
8. How do you defend security/surveillance and entry-auto against SmartSens/GalaxyCore's ~20% price discount — defend or cede?
9. What inventory / demand-visibility discipline has changed since the 2022–23 writedowns to prevent a repeat next downcycle?
10. Is medical/industrial CIS (endoscopy) a scalable high-margin annuity or a niche — target share of mix?
11. With ~RMB 6.8bn net cash + fresh IPO proceeds, what is the M&A appetite, buyback, and dividend policy now A+H listed?
12. How should minority holders read the RMB 4.9bn founder stock gift to a founder-affiliated university (now a ~4.2% holder)?
13. What is the strategy and timeline for AR/VR and robotics vision as the next volume driver beyond phone/auto?
14. Is the low-margin distribution/analog tail strategic, or a candidate for divestiture to lift blended margin?
15. What precise price/mix assumptions underpin "sequential improvement from Q2 2026"?

---