Design vs. Manufacturing: The Technical Boundaries of Advanced Semiconductor Chips
The competition in advanced process chips is often oversimplified as a race to build the smallest transistor. However, from an engineering and industrial perspective, bringing a chip from concept to mass production involves multiple critical stages, including architecture design, physical implementation, wafer manufacturing, packaging, testing, and system verification. This article explores the division of labor between chip design and manufacturing, explains the true meaning behind process nodes, and analyzes the distinct challenges of each phase. It also examines the impact of export controls and offers practical insights for technical teams to manage supply chain risks and compliance in their hardware and software architectures.
The competition in advanced process chips is often simplified to “who can build the smallest transistor.” However, from an engineering and industry perspective, taking a chip from concept to mass production requires multiple stages: architecture design, physical implementation, wafer manufacturing, packaging, testing, and system verification. Each stage has unique capabilities and thresholds. Understanding these differences is essential for interpreting industry news and evaluating technological roadmaps.
1. The Division of Design and Manufacturing
The modern semiconductor industry widely adopts a division-of-labor model. Design companies are responsible for selecting architectures, planning circuit logic, completing layout, and performing functional and timing verification. The manufacturing phase is handled by foundries, which use processes like lithography, etching, and deposition to transform design files into physical structures. Design companies can either develop core modules in-house or purchase general-purpose IP licenses, such as processor cores, GPUs, and baseband modules.
This division leads to two key conclusions. First, companies with design teams do not necessarily need to build their own fabs. Second, passing simulation and verification only proves that the logic and layout are theoretically correct; it does not guarantee stable mass production. Yield, unit cost, and lead time ultimately determine whether a product can reach the market at scale. Therefore, when evaluating a company’s chip capabilities, design, foundry, and mass production must be assessed as separate dimensions.
2. What Process Nodes Actually Mean
Names like “3nm” and “5nm” originally referred to critical dimensions like gate length. As technology evolved, they became general industry terms for process generations. The naming standards vary among manufacturers and do not directly correspond to the same physical dimensions. Therefore, when assessing the advancement of a process, it is better to refer to comparable metrics like transistor density, performance per watt, yield, and production scale, rather than just the node name.
Transistor density is the most common quantification method, calculated as follows:
Density = Total number of transistors ÷ Chip areaExample: Total transistors = 19 billion = 19 × 10^9Chip area = 109 square millimetersDensity ≈ 1.74 × 10^8 per square millimeter (about 174 million per square millimeter)
The values above are solely for demonstrating the calculation method. The actual density of a product is also influenced by factors such as cell library design, the number of routing layers, the proportion of on-chip memory, and functional module division. Under the same process, densities can vary significantly across different designs, so a single number cannot determine the superiority of a process.
3. Design Capability Does Not Equal Mass Production Capability
The difficulty of designing advanced chips lies primarily in complexity management. As the number of transistors increases, timing closure becomes more difficult, power networks require finer planning, and thermal management becomes stricter. Verification work often occupies the majority of the project cycle, with a significant increase in simulation test cases and regression testing scales.
On the manufacturing side, challenges are concentrated on the process window. Advanced nodes require high-end equipment such as Extreme Ultraviolet (EUV) lithography. With numerous process steps, any deviation in a single step can lower yield. Tape-out results from the same layout on different production lines can vary greatly in cost and delivery. In other words, what a design team can accomplish may not necessarily be realized on a specific production line at an acceptable cost.
4. The Mechanism of Export Controls
Export controls typically do not target the design of a specific chip itself, but rather the technologies, equipment, software, and entities that could be used for sensitive applications. Common control measures include the following three categories:
- Restricting specific entities by adding them to a list, limiting their access to controlled items or technologies. This directly affects procurement and partnerships; even if a company has design capabilities, it may be unable to secure foundry services.
- Imposing licensing requirements on manufacturing equipment, materials, and Electronic Design Automation (EDA) tools related to advanced processes. This affects the upstream supply chain and limits the pace of process expansion across the entire industry.
- Requiring foundries to review end-users and end-uses before accepting orders. This increases compliance costs and lead times, making project scheduling more uncertain.
These measures have different impact paths. Restriction lists primarily change who can collaborate with whom, while technology and equipment controls affect the technology acquisition capabilities of the entire industry. Consequently, the same design may yield entirely different supply results under different policy frameworks. When analyzing policy impacts, it is crucial to first determine whether the clause targets entities, technologies, or equipment, and then evaluate its actual effect on specific products.
5. Implications for Enterprises and Developers
For technical teams, supply chain risks should be included in architecture reviews rather than being discovered just before mass production. Consider the following approaches:
- Incorporate process nodes, foundry channels, and backup plans into the project risk register. Every critical component should specify a primary supplier, alternative solutions, and switching conditions, subject to regular reviews.
- Maintain at least one alternative design or packaging solution for critical components. If a certain component becomes unavailable, the system should still operate in a degraded or alternative mode.
- Implement hardware abstraction at the software level by encapsulating chip-related drivers, instruction optimizations, and low-level interfaces into independent modules. This ensures that migration costs remain acceptable even if the platform changes.
- Factor compliance review cycles into the mass production timeline. Export-related audits may take weeks or longer; ignoring this step will directly lead to delivery delays.
6. Clarifying Common Misconceptions
Misconception 1: Strong design capability equals mastery of advanced manufacturing.
Fact: Design and manufacturing belong to different industry segments, and their capabilities should be evaluated separately. A company can lead in design while relying on external foundries for production.
Misconception 2: A smaller process node number always means a stronger product.
Fact: The node is only one process metric. System-level performance also depends on architecture choices, memory bandwidth, software optimization, and cooling. A chip with an older process but a rational architecture may perform better in specific scenarios.
Misconception 3: Policy restrictions only target the product itself.
Fact: Export controls may target entities, technologies, equipment, or end-uses. Evaluating their impact requires reviewing the scope of specific clauses rather than relying solely on media summaries.
