SMIC founder backs mature nodes
- SMIC founder Richard Chang said on May 9 that China’s chip push is too fixated on 2 nm and 3 nm, missing bigger mature-node markets. - Chang’s key claim was numerical: advanced nodes are under 20% of chip demand by product count, while 80% sits in mature processes. - The timing matters because mature-node capacity is tightening again, with 8-inch foundry utilization projected near 90% in 2026.
Semiconductors are having one of those moments where the loudest conversation is not the most useful one. Everybody talks about 2 nm and 3 nm. But Richard Chang — the founder of SMIC, China’s biggest contract chipmaker — just made the opposite case. In an interview published May 9, he said the obsession with bleeding-edge nodes is a category error, because most real chip demand still lives in mature processes and specialty technologies. ### Who said what? Chang, also known as Zhang Rujing, told Chinese tech outlet Kechuangban Daily that “only” chasing 3 nm and 2 nm is the wrong way to think about semiconductor competition. His core point was simple: by product volume, advanced nodes account for less than 20% of the market, while more than 80% of demand comes from mature nodes and specialty processes. He framed that as the practical opening for Chinese companies — not full-spectrum dominance, but targeted wins in niches still controlled by foreign suppliers. (news.qq.com) ### What counts as a mature node? Basically, not every chip needs the newest transistor geometry. Power chips, display drivers, analog parts, RF components, industrial controllers, automotive MCUs, sensor interfaces, and a lot of embedded logic often run on older process nodes or on specialty variants that optimize voltage, reliability, radio performance, or non-volatile memory instead of raw density. That is why foundries like SMIC still emphasize 8-inch and 12-inch manufacturing across broad process families rather than a single race to the smallest number. (news.qq.com) ### Why is Chang pushing this now? Because the market is lining up behind the argument. TrendForce said this week that average 8-inch utilization among the world’s top 10 foundries is projected to approach 90% in 2026 and stay above 80% through the first half of 2027. The driver is not smartphone application processors at 2 nm. It is power-related demand, AI hardware spillover, and capacity being redirected inside the mature-node world itself. (smics.com) That makes Chang’s point feel less like ideology and more like a read on where money is actually being made. ### Why does this matter for China specifically? The catch is that “mature” does not mean easy. These businesses still depend on process control, yields, materials, equipment, packaging, and application-specific know-how. Chang’s argument was that China does not need to win every layer at once. He said second- and third-tier cities should avoid copying first-tier cities by pouring money into giant wafer-fab projects with long payback periods. (trendforce.com) Instead, they should build clusters around one narrow strength and become excellent there. ### What kind of niches is he talking about? He pointed to niche markets that are small in volume but strategically valuable — things like chips for medical ultrasound, radiation-hardened aerospace parts, industrial servo drives, and silicon-based capacitor technologies. In the same interview, he highlighted a local 3D silicon capacitor project as an example of a specialty path that could substitute for incumbent components and even overlap with advanced-packaging use cases in wearables, communications, vehicles, and distributed AI hardware. (news.qq.com) ### Where does AI fit in? Not mainly in giant training clusters — at least not for most startups. Chang split AI into two camps: big centralized compute, which needs national-scale backing or huge capital, and edge or distributed AI, which he said represents the larger practical opportunity for ordinary companies. That fits the mature-node thesis. Edge AI devices still need lots of power management, sensing, connectivity, control, and packaging tricks — the kind of work that often sits outside the leading edge. (news.qq.com) ### Does this mean advanced nodes do not matter? No. They matter a lot for top-end compute, flagship mobile chips, and national prestige. But Chang’s point is that people confuse “most glamorous” with “most of the market.” Even SMIC’s own recent results underline the value of scale and utilization across its broader manufacturing base — 2025 revenue rose to $9.33 billion, monthly capacity topped 1 million 8-inch-equivalent wafers, and utilization reached 93.5%. (news.qq.com) That is a foundry business built on breadth, not just frontier bragging rights. ### Bottom line? Chang is not saying the leading edge is fake. He is saying the industry conversation is distorted. If 80% of demand really sits in mature and specialty processes, then the smarter play is not to imitate TSMC at every node. It is to own the boring, sticky, high-value parts of the stack that everybody still needs. (news.qq.com) (smics.com)
