SMIC founder Richard Chang has pushed back on the idea that semiconductor success should be measured only by whether a company reaches 3nm or 2nm manufacturing. In recent interview remarks, Chang argued that the industry’s fixation on leading-edge nodes can obscure other areas where chipmakers may create value.
The point is notable because much of the current semiconductor conversation is shaped by advanced-node manufacturing. TSMC’s position in the market is closely tied to its ability to build leading-edge chips at scale for major customers such as Apple and NVIDIA. Those relationships are supported by process stability, high-volume manufacturing, and the kind of production reliability that only a small number of foundries can offer.
Chang, however, framed the issue differently. In his view, treating 2nm or 3nm as the only meaningful benchmark is too narrow. He described that mindset as a misconception and argued that semiconductor companies can make important progress by focusing on specific technical gaps, especially in markets that do not require the newest lithography.
Why Mature Nodes Still Matter
According to Chang’s remarks as reported, advanced manufacturing represents only a portion of the total chip market, while mature processes still support a broad range of demand. That distinction matters for buyers, suppliers, and policymakers because not every chip needs to be built on the most advanced node available.
Automotive electronics, industrial control systems, power management chips, sensors, display drivers, connectivity parts, and embedded processors often depend more on cost, supply consistency, qualification cycles, and long-term availability than on headline process density. In those areas, a foundry does not necessarily need to compete directly with the most advanced TSMC or Samsung processes to be commercially relevant.
Understanding Semiconductors
Readers who want more context on wafer fabrication, chip design, packaging, and semiconductor terminology may find this technical primer useful alongside the article’s discussion of mature-node strategy.
As an Amazon Associate I earn from qualifying purchases.
Chang’s broader argument is that a company can contribute meaningfully by becoming excellent in a particular niche. Rather than trying to cover every part of the semiconductor supply chain at once, he suggested that companies should identify bottlenecks and focus resources where a specific weakness can be solved.
That approach is especially relevant for China’s domestic semiconductor industry, where some advanced manufacturing options remain constrained. SMIC has faced limits tied to access to the most advanced chipmaking equipment, including EUV lithography systems used by leading foundries for newer process nodes. The company has continued working with DUV-based equipment, but claims about its exact technical ceiling and future roadmap should be treated carefully unless confirmed by the company itself.
A Different Path for AI Hardware
Chang also pointed to AI hardware as an area where the discussion may be too concentrated on large-scale cloud computing. The current AI chip market is heavily associated with data center accelerators, high-bandwidth memory, and leading-edge manufacturing capacity. That segment attracts enormous investment because training and serving large AI models can require expensive, power-hungry hardware.
His comments suggest that there may also be room for more scenario-specific AI hardware, though the scale of that opportunity remains difficult to verify from the interview alone. Edge devices, industrial systems, smart cameras, vehicles, appliances, and specialized embedded products may all require different design priorities than cloud AI accelerators. In many of those cases, efficiency, integration, reliability, and cost can matter as much as raw compute density.
Chip War
For readers following the China, Taiwan, and advanced-manufacturing angle, this book offers a broader history of how semiconductor supply chains became strategically important.
As an Amazon Associate I earn from qualifying purchases.
For startups, Chang’s message was also more cautious than the usual race-to-scale narrative. Rather than trying to challenge dominant AI hardware suppliers head-on, he appeared to favor narrower applications where a company can solve a practical problem. That does not guarantee commercial success, but it reflects a more targeted strategy than spending heavily to compete in the most crowded part of the market.
What This Means for Chip Buyers
For customers evaluating semiconductor supply, the takeaway is not that advanced nodes are unimportant. They remain essential for flagship smartphones, high-end GPUs, AI accelerators, and other performance-sensitive products. The point is that the chip industry is larger than that top layer.
A mature-node foundry can still be strategically useful if it offers dependable capacity, competitive pricing, stable yields, and expertise in the right product category. Buyers with long product lifecycles may also value process maturity because redesigning around a newer node is not always worth the cost or risk.
Chang’s comments therefore read less like a dismissal of advanced manufacturing and more like an argument for prioritization. The 2nm and 3nm race will continue to shape the high end of the market, but many commercial opportunities sit elsewhere. For SMIC and other manufacturers operating under technical or supply constraints, those overlooked segments may be where focused progress is most practical.