In August 2026, TIME published its annual TIME100 AI list. Among the "Leaders" was a name most smartphone buyers have never heard: Mong-Song Liang (梁孟松), co-chief executive of Semiconductor Manufacturing International Corporation (中芯国际, SMIC), China's largest contract chipmaker. For a man who has not appeared at an SMIC earnings call since 2022, the recognition was a reminder that some of the most consequential work in the AI hardware race happens far from the spotlight.
Liang is 74, born in 1952 in Taiwan. He earned bachelor's and master's degrees from National Cheng Kung University (成功大学) and a PhD in electrical engineering and computer science at the University of California, Berkeley, where he studied under Chenming Hu (胡正明) — the inventor of the FinFET transistor structure that underpins every advanced chip from 14nm down to 2nm. That lineage matters: Liang did not just manage fabs, he understands the physics of making transistors smaller.
A career across three chipmaking powers
Liang's résumé reads like a map of the global semiconductor industry. He started at Advanced Micro Devices (超微, AMD) in the early 1980s, working on memory and logic process technology. He joined Taiwan Semiconductor Manufacturing Company (台积电, TSMC) in 1992 and spent 17 years there, becoming one of the company's most prolific inventors — credited with roughly 180 U.S. patents. He was part of the team that, in 2003, delivered TSMC's 130nm copper process ahead of IBM, a landmark win that helped turn TSMC from follower into leader. Inside TSMC, the core R&D group was nicknamed the "Six Knights (研发六骑士)."
In 2011 he moved to Samsung Electronics (三星), where as chief technology officer of the LSI division he pushed the Korean giant's process from 45nm to 14nm FinFET. Samsung's 14nm beat TSMC to market in early 2015 and won Apple and Qualcomm orders — a reordering of the industry that Liang directly shaped.
He joined SMIC in October 2017 as co-CEO alongside Zhao Haijun (赵海军), in a dual-leadership structure built specifically to give him command of advanced-process R&D.
What he actually changed at SMIC
When Liang arrived, SMIC's most advanced volume node was 28nm, several generations behind TSMC and Samsung, and its yields were unsteady. Under U.S. export controls that block sales of the most advanced lithography machines — specifically ASML's EUV scanners — SMIC cannot legally buy the tools needed for cutting-edge nodes below 7nm.
Liang's answer was to squeeze more out of older deep-ultraviolet (DUV, 深紫外) equipment through multi-patterning (多重曝光). That route let SMIC reach 14nm mass production in late 2019, with Huawei's HiSilicon (海思) Kirin 710A as an early customer, and later develop 7nm-class N+1 and N+2 processes. When Huawei surprised the market by releasing the Mate 60 Pro with a Kirin 9000S processor in August 2023, teardown firms including TechInsights concluded the chip was built on SMIC's 7nm-class process — the first time a mainstream Chinese phone shipped a domestically produced advanced node at scale.
The Wall Street Journal profiled Liang under the headline "Taiwanese Chip Wizard Drives China's Technology Ambitions," framing him as the engineer who took SMIC from a laggard to a company capable of 7nm-class output. TIME's 2026 citation builds on that: it notes SMIC's N+3 process now supplies Huawei's Kirin 9030 and the Ascend 950PR AI processor (昇腾950PR), positioning the foundry as a domestic alternative in AI accelerators. TIME also balanced the praise, noting N+3's transistor density still trails TSMC and Samsung, and that building advanced chips on older tools carries lower yields and higher cost.
Still at the center, quietly
Liang's low profile after 2022 sparked speculation that he had left. SMIC's latest filings put that to rest. He remains co-CEO, and in April 2026 the company granted him more than 80,000 restricted shares (RSUs) vesting through 2028. As of mid-2026 he held about 1.36 million SMIC shares, roughly 0.016% of the company. The equity package signals SMIC expects him to stay central to advanced-node development.
What readers can do now
- If you follow the AI hardware story, track SMIC's quarterly disclosures and independent teardowns (such as TechInsights) rather than rumor threads — the foundry no longer pre-announces node progress, so teardowns are the most reliable signal.
- For investors, separate "advanced node" hype from "mature node (成熟制程)" reality: SMIC's scale and margin come mostly from mature processes, while 7nm-class output is still capacity-constrained and costly.
- For students, Liang's path is a case study in process engineering as a career: the bottleneck in AI hardware is often not the model but the manufacturing know-how behind it.
Honest limitations
This article relies on SMIC's own filings, TIME's 2026 list, and Taiwan and Reuters-syndicated reporting (WSJ, China Times) on Liang's role. I have deliberately not used the precise yield percentages (for example, "28nm yield 60%→85%") that circulate in Chinese-language blogs, because those figures are not verifiable from primary sources. The exact transistor counts and density gaps of N+3 versus TSMC's nodes are not publicly disclosed by SMIC, so any numerical comparison is necessarily approximate. Liang himself rarely gives interviews, so much of his management style is reconstructed from secondhand accounts.
Key takeaways
- Mong-Song Liang is the process-engineering force behind SMIC's leap from 28nm to 7nm-class production under strict export controls.
- His inclusion on the 2026 TIME100 AI list reflects how central chip manufacturing has become to the global AI contest.
- The achievement is real but constrained: no EUV access means SMIC's advanced nodes trail TSMC on density and cost, and Liang's succession remains an open question at age 74.