Design-Only vs Full Fab: Which Semiconductor Business Model for China?
When entering China’s semiconductor market, one of the first strategic decisions a company faces is choosing between a design-only (fabless) model and a fully integrated (IDM or pure-play foundry) manufacturing model. This choice fundamentally shapes capital requirements, risk profile, regulatory exposure, talent needs, and competitive positioning. While in global markets the fabless model has become dominant for most chip companies (representing over 50% of semiconductor revenue in 2024), China’s unique regulatory environment, incentive structure, and ecosystem maturity create different dynamics that make this decision far more nuanced.
This comprehensive comparison examines both business models across ten strategic dimensions, helping foreign semiconductor companies determine which approach aligns with their capabilities and objectives in China’s market.
1. Defining the Models
1.1 Design-Only (Fabless) Model
A fabless semiconductor company focuses exclusively on chip design, architecture, and verification, outsourcing all manufacturing to third-party foundries. In China, the fabless model is the most common structure for foreign IC design houses and domestic chip startups alike. Key characteristics include:
- Capital-light: Initial investment typically RMB 10–50 million for a mid-sized design team
- IP-centric: Core value resides in design IP, architecture, and software stacks
- Foundry-dependent: Manufacturing relationships essential; capacity allocation and process access are critical dependencies
- Fast time-to-market: Typical product cycle 12–24 months from concept to tape-out
- Flexible scaling: Can scale up or down without fixed manufacturing asset commitments
1.2 Full Fab (IDM/Foundry) Model
A full fab model involves owning and operating semiconductor fabrication facilities. This includes both the Integrated Device Manufacturer (IDM) model (design + manufacture + sell own chips) and the pure-play foundry model (manufacture chips designed by others). Key characteristics include:
- Capital-intensive: A leading-edge fab requires RMB 10–30 billion+ in investment; mature-node fabs RMB 2–10 billion
- Asset-heavy: Significant fixed assets; depreciation dominates financial statements
- Process-dependent: Competitive advantage rooted in manufacturing process technology and yield management
- Long development time: 3–5 years from ground-breaking to commercial production
- Economies of scale: Unit costs decrease dramatically with volume; minimum efficient scale is large
2. Comparative Analysis
| Dimension | Design-Only (Fabless) | Full Fab (IDM/Foundry) |
|---|---|---|
| Capital Requirement | RMB 10–50M startup; working capital manageable | RMB 2–30B+; capital markets/fund access essential |
| Regulatory Approval Complexity | Low–Moderate: WFOE generally open; minimal sector-specific licensing | Very High: National security review, MIIT certification, EIA, hazardous chemicals, multiple permits |
| IP Protection Risk | Moderate: IP is the product; enhanced protective measures needed | High: Process technology leaked through equipment suppliers, engineers, and collaborative partners |
| Talent Requirements | 50–200 highly specialized IC design engineers; competitive recruitment | 500–5,000+ staff across 30+ disciplines; broader but shallower talent pool |
| Incentive Eligibility | Limited to design enterprise incentives; R&D super deduction | Maximum incentives: 10-year CIT exemption, capex subsidies, free land, utility discounts |
| US/Export Control Exposure | Moderate: EDA access, certain design tools may be restricted | Very High: Equipment, chemicals, and process technologies face severe US/EU export restrictions |
| Exit Flexibility | High: Can be acquired by larger players (IPO or trade sale); no asset encumbrance | Low: Asset-heavy structure limits buyer pool; complex separation |
| Revenue Scalability | Low marginal cost per additional design; high-margin at scale | Volume-dependent; high fixed costs need high utilization (80%+) for profitability |
| Foundry Capacity Risk | High dependency on foundry partners; capacity allocation risk during shortages | Internal capacity control; no allocation disputes—but lower utilization = higher unit cost |
| Time to Revenue | 12–24 months from team formation to first product revenue | 36–60 months from funding to first wafer revenue |
3. China-Specific Factors That Change the Equation
3.1 Foundry Access and Capacity Allocation
In global markets, fabless companies have multiple foundry options with broadly comparable terms. China’s foundry landscape is different:
- SMIC dominance: SMIC controls approximately 60% of China’s advanced foundry capacity (55nm and below). Capacity allocation is not purely commercial—the government has influence over which design companies get priority access.
- Capacity shortages: During semiconductor upcycles (2021–2023, potential 2025–2026), foundry capacity is rationed, and foreign fabless companies may find themselves deprioritized relative to domestic champions.
- Process technology access: The most advanced nodes available in China (SMIC’s N+1/N+2, roughly equivalent to 7nm) have limited capacity and are preferentially allocated to “national team” companies.
- Alternative foundries: Hua Hong (MCU, power, memory specialty), Shanghai Advanced (power management), and Wuhan Xinxin (memory) offer differentiated but more limited process options.
Implication for fabless companies: Foundry access in China is not purely a commercial negotiation. Fabless companies must invest in government relationships and ecosystem partnerships alongside their foundry commercial agreements. Some foreign fabless companies are forming strategic alliances with domestic Chinese IC design houses in part to improve their foundry access position.
3.2 The Incentive Asymmetry
China’s semiconductor incentive system is heavily skewed toward manufacturing. The tax benefits available to a qualifying sub-28nm fab (10-year full CIT exemption, effectively saving RMB 500 million–2 billion over a decade) far exceed what is available to even the most successful fabless design house (reduced 10% CIT rate). This creates a strong financial incentive to consider the fab model—but only if your company can bear the capital commitment and regulatory complexity.
3.3 The “Make in China” Policy Push
China’s industrial policy increasingly favors companies that manufacture within China. The Manufacturing 2025 initiative and subsequent policies encourage—and in some cases require—domestic manufacturing for chips used in critical infrastructure, government procurement, and national security applications. For a fabless company whose chips are designed in China but manufactured abroad (e.g., at TSMC or Samsung), access to these growing market segments may be restricted or require special approvals.
For foreign companies, this creates a strategic tension: manufacturing abroad keeps production in familiar regulatory jurisdictions but limits addressable market in China. Manufacturing in China opens the full market but exposes the company to China’s regulatory oversight and potential technology transfer requirements.
3.4 Export Control and Sanctions Exposure
US and EU export controls on semiconductor manufacturing equipment and technologies have fundamentally altered the risk calculus for full fab operations in China:
- Equipment access: Advanced lithography (EUV), certain etch/deposition tools, and metrology equipment are subject to US and Dutch export restrictions for China-based fabs. A full fab in China cannot access the world’s most advanced manufacturing tools.
- Foundry customer restrictions: Companies that own or operate fabs in China may find their access to non-Chinese customers restricted due to supply chain security concerns raised by those customers’ home governments.
- Technology “look-through”: US export controls increasingly apply a “foreign direct product rule” that restricts products made with US technology, even if manufactured outside the US. A China fab using any US-origin equipment faces downstream restrictions on selling to certain customers.
Implication for fabless companies: Fabless companies manufacturing at Chinese foundries face downstream export control risks. If your chips are manufactured at SMIC, your ability to export those chips to US or allied-country customers may be restricted under US Entity List rules. Many foreign fabless companies maintain dual-source strategies—manufacturing their lead products at TSMC or Samsung for global markets, and only shifting volume to Chinese foundries for the China domestic market.
4. Decision Matrix: Which Model for Your Business?
| Business Profile | Recommended Model | Rationale |
|---|---|---|
| AI accelerator chip company | Fabless | Need advanced process nodes (7nm/5nm) unavailable in China; global customer base requires TSMC/Samsung manufacturing |
| Automotive MCU company | Fabless (with foundry partnership) | Mature-node (65nm–28nm) available at SMIC/Hua Hong; automotive qualification requires stable foundry partnership; WFOE structure protects IP |
| Power management IC company | Fabless or IDM-lite | Specialty processes (BCD, SOI) available at multiple foundries; captive fab only justified at very high volume (>500M units/year) |
| Memories (DRAM/NAND) | Full Fab (IDM) | Memory economics demand full vertical integration; China’s memory champion YMTC demonstrates state support model |
| MEMS/sensors | Fabless (preferred) or IDM-lite | Specialty MEMS foundries available in China; captive fab only for proprietary processes |
| RF/analog/specialty analog | IDM-lite or Fabless | Analog differentiation often in process technology; some captive capability may be justified but can use specialty foundries |
| High-volume commodity chips (MCUs, PMICs, drivers) | Fabless with China foundry | Volume justifies dedicated foundry capacity agreements but not captive fab; competitive cost structure |
| Defense/sensitive applications | Full Fab (IDM or JV) | Supply chain security requirements; government mandates domestic manufacturing; but foreign companies should carefully evaluate risks |
| Startup (seed/Series A) | Fabless | Capital-light; focus on design innovation; foundry partnerships scalable; exit optionality preserved |
| Established IDM entering China | Fabless China entity + offshore fab | Design center in China accesses talent and market; manufacturing remains offshore for IP protection and export control flexibility |
5. The Hybrid Models That Work Best in China
Increasingly, the most successful semiconductor business models in China are hybrids that blend elements of both design-only and full fab approaches:
5.1 “China Design Center” Model (Most Common for Foreign Companies)
The foreign parent maintains global manufacturing operations (often a mix of captive fabs and foundry relationships) while establishing a design-only subsidiary in China. This model:
- Accesses China’s IC design talent pool (particularly in Shanghai and Shenzhen)
- Keeps core manufacturing IP outside China’s legal jurisdiction
- Allows ad-hoc shifting of China-designed products to Chinese foundries when market access requires it
- Preserves maximum strategic flexibility and regulatory optionality
5.2 “Foundry Collaboration” Model
A foreign design house establishes a deep strategic partnership with a Chinese foundry (typically SMIC or Hua Hong) that includes:
- Dedicated capacity allocation agreement (2–5 year term)
- Joint process development for optimized design-manufacturing co-optimization (DTCO)
- Revenue-sharing on products that use jointly developed process modifications
- This provides many benefits of vertical integration without the capital commitment
5.3 “Captive Fab for Specialty Only” Model
For companies whose competitive advantage depends on specialized process technology (e.g., BCD power management, high-voltage processes, specialized analog), building a small, dedicated fab for that specific specialty process while using foundries for standard CMOS products can be optimal. China offers support for specialty fabs at lower costs than advanced node fabs, and the incentive-to-investment ratio is often favorable.
6. Risk Assessment by Model
| Risk Category | Design-Only | Full Fab |
|---|---|---|
| Capital loss risk | Low (limited fixed assets) | Very high (sunk equipment costs; thin secondary market) |
| IP theft risk | Moderate (design files and databases targetable) | High (process recipes, tool parameters, entire manufacturing methodology) |
| Export control disruption | Moderate (EDA tool access risk; foundry allocation risk) | Very high (equipment access, spare parts, chemicals, customer restrictions) |
| Regulatory change risk | Low-Moderate (less direct regulatory exposure) | High (multiple agency oversight; policy-reversal risk) |
| Talent retention risk | Moderate (competition for top designers intense) | High (broader workforce; higher turnover in manufacturing roles) |
| Market cyclicality exposure | Moderate (variable costs; can scale down) | High (fixed costs continue regardless of utilization) |
| Exit risk | Low (clean exit; multiple buyer options) | Very high (few buyers; complex separation; asset impairment) |
7. Financial Comparison: 10-Year Total Cost of Ownership
| Metric (RMB) | Fabless (Design WFOE) | Full Fab (Mature Node, 65nm) | Full Fab (Advanced, 28nm) |
|---|---|---|---|
| Initial Investment | 30M | 3B | 20B |
| Annual Operating Cost (Year 3+) | 50M–100M | 1.5B–2B | 5B–8B |
| 10-Year Pre-Tax Profit (at target revenue) | 500M–2B | 3B–8B | 10B–30B |
| 10-Year Tax Savings (Incentives) | 50M–200M | 750M–2B | 2.5B–7.5B |
| 10-Year Net Profit After Tax | 450M–1.8B | 3.75B–10B | 12.5B–37.5B |
| Return on Investment (10-year IRR) | 25–45% | 8–15% | 5–12% |
| Breakeven Year | Year 2–3 | Year 5–7 | Year 7–10 |
Note: Above figures are illustrative ranges based on typical market conditions. Actual results vary significantly based on product market fit, execution quality, and market conditions.
8. Strategic Recommendation
A full fab model should only be pursued in specific circumstances: (1) Your core competitive advantage IS your manufacturing process technology; (2) you have RMB 3 billion+ in committed funding; (3) you have a clear path to 80%+ fab utilization within 24 months; (4) you have secured all necessary equipment, chemical, and materials supply chains despite export control constraints; and (5) you have a strategic Chinese partner (typically an SOE) with government relationships and political cover.
The hybrid model—China design center + offshore manufacturing—is increasingly the “default” for sophisticated foreign semiconductor companies. It captures the best of both worlds: China’s design talent and market access without the capital intensity, regulatory complexity, and IP exposure of onshore manufacturing.
9. Practical Next Steps
- Model selection: Use the decision matrix above to determine which model aligns with your specific business profile. Document your assumptions and test them against current regulatory conditions.
- If fabless: Identify 2–3 Chinese foundry partners; begin process qualification; evaluate EDA tool compliance with Chinese regulations.
- If full fab: Engage Chinese legal counsel for feasibility assessment; begin partner search (JV structure likely required); prepare national security review application; secure equipment supply chain commitments.
- If hybrid: Establish design center WFOE; negotiate capacity agreement with Chinese or Taiwanese foundry; document IP protection strategy for China-based design activities.
- Risk mitigation: Regardless of model selected, develop a comprehensive China risk management framework addressing IP protection, export control compliance, talent retention, and regulatory change scenarios.
