TSMC has built its semiconductor dominance on a relentless push into cutting-edge manufacturing, moving quickly from breakthrough to stable mass production and turning that technical edge into long-running partnerships with some of the world’s biggest companies, including Apple and NVIDIA. Hitting huge volumes at 3nm today and preparing for 2nm next is often treated as the ultimate measure of success in the chip industry.
But SMIC founder Richard Chang says that thinking is flawed.
In a recent interview, Chang argued that the semiconductor race isn’t defined solely by who reaches the most advanced node first. In his view, treating 3nm or 2nm as the only real finish line is a misconception, because those advanced processes represent a surprisingly small slice of the overall chip market.
Chang’s key point is that leading-edge manufacturing accounts for less than 20 percent of total demand, while more than 80 percent comes from mature process technologies. These older, widely used nodes power massive portions of everyday electronics, industrial systems, automotive applications, connectivity hardware, and countless embedded devices. That’s why, instead of chasing the same crowded path as every other chipmaker, he believes the smarter strategy is “excellence in a specific niche.”
According to Chang, niche markets that have historically been dominated by overseas suppliers could be ideal targets for domestic semiconductor companies. The opportunity, he suggests, is not in trying to do everything at once, but in prioritizing the right problem and solving it decisively—especially in areas where there’s a clear bottleneck holding the broader industry back.
He also cautioned against “homogeneous competition,” where too many companies pile into the same popular segments and burn resources trying to copy the same playbook. A more pragmatic path, he said, is to focus on missing links—specialized, underserved sub-markets where focused effort can lead to meaningful breakthroughs and real industry value.
Chang extended this thinking to artificial intelligence as well. While investment and attention have heavily centered on cloud computing and data center AI, he sees distributed AI as a major area being overlooked. That gap points to unmet demand for scenario-based hardware—AI designed around specific real-world use cases rather than one-size-fits-all infrastructure.
For AI startups, his advice is especially pointed: don’t try to outspend the biggest players by burning through millions in funding in a head-on fight. Instead, pursue scenario-based applications where differentiated hardware and targeted solutions can win without requiring endless capital.
This perspective also fits SMIC’s current reality. China’s leading chipmaker remains constrained by limited access to the most advanced EUV lithography equipment required for producing chips at 5nm and below. Relying on older DUV tools has effectively kept SMIC around the 7nm class of manufacturing for now. But Chang’s comments suggest that mature nodes aren’t simply a limitation—they may be a strategic advantage when paired with the right niche-market focus.
If the market truly is dominated by mature-process demand, then a chipmaker that prioritizes specialization, stability, and solving specific bottlenecks could still carve out a powerful position, even without leading-edge node bragging rights.
