Thread by @dnystedt

Ex-TSMC R&D boss Shang-yi Chiang made headlines last week over a Computer History Museum interview. Chiang was vital to TSMC's success. When he joined TSMC in 1997, it was a technology laggard. When he retired in 2013, it was an industry leader. $TSM #semiconductors
Thread 1/25

2/25 Chiang: Employees, the key to TSMC's success
“0.13 micron (130-nanometers, in year 2000) was a very important node for TSMC because we became the first one in the world to deliver manufacturing wafer of copper and low-K. I remember going, "Wow! This company did something."

3/25 The key TSMC success
Other chipmakers asked Chiang, "We all take 2-years to develop one generation, how come you guys can do it in 1 or 1 ½ year?"

Chiang: "because my R&D engineer works three shifts (24-hours a day) and you only work one shift (8-hours)…"

4/25 “So, I firmly believe this is one of the really important reasons why TSMC succeeded.” In the semiconductor industry, equipment depreciation costs are so high, you really want to run your equipment 24-hours a day, he said. But many company’s don’t. $TSM

5/25 When he was in the US (HP, TI, others) if equipment broke down, they waited until morning to fix it.

“But (at TSMC) if at two o'clock in the morning, we just called the equipment engineer... He won't complain. And his wife won't complain. And that's the way it is.”

6/25 He was asked, ‘How do you make R&D engineers work nights?’

Chiang joked, "In Taiwan, we all have to serve the military." And everyone does guard duty at night. “And so, all my engineers have been through that!”

7/25 Interviewer: What’s the real answer?
Chiang: “Asians are more hungry, because we had a tougher life. So, to make money is more important to us. People are willing to sacrifice their private life in order to have financial security.”

8/25 But Chiang added, …”TSMC doesn't have that spirit anymore.” The older generation had commitment that young ones today don’t share, he said, “and the young ones didn't work as hard anymore.” $TSM

9/25 How TSMC beat Intel in chip manufacturing
Chiang’s biggest career regret: “We didn't catch up with Intel in my career.”

But he set TSMC on a firm path to success by working consistently, methodically, and avoiding mistakes. $TSM $INTC

10/25 “…we didn't really do anything special, anything great. But we didn't make any major mistake,” he said, adding early on, UMC was a strong rival, but at the 0.13 micron node, UMC made mistakes, and Intel made some mistakes, but “TSMC didn’t make any big mistakes.”

11/25 Over time, TSMC overtook Intel.
Chiang: “I really respect Intel… In every generation, they are willing to take risk ... And in many key areas, for example, like high-K metal gate, strained engineering, FinFET, etc., it was always Intel" first.

12/25 Winning Apple (TSMC's biggest customer)
TSMC won over Apple with advanced chip packaging technology, Chiang says. He talked with then-chairman, Morris Chang, in 2009, and stressed the need to work on Advanced Package Technology (NOT wire-bonding). $TSM $AAPL

13/25 TSMC was seeking new ways to keep Moore's Law moving forward.
Progress in silicon wafers had been significant under Moore's law, he said, "but if you look at the package in a PC board"…it had not changed in nearly 20-years. It had become a bottleneck.

14/25 The focus in chip packaging for years had been cost cutting, not R&D, and TSMC saw the package had become a bottleneck, he said. In graphics chips, for example, it created problems with memory access and power.

15/25 Chiang put 400 engineers and US$100 million in equipment to work on a solution.
“So, a year later, we develop this technology. So I just simply … want to replace PC board with silicon. It’s called silicon interposer.” $TSM

16/25 The results seemed great, but Chiang found it a hard sell. It cost too much. He’d had Nvidia in mind as a customer, but they said no. Xilinx used it in a limited way, but not as Chiang envisioned.
“In my mind, my good innovation was not used in a good way,” he said.

17/25 Xilinx codenamed the packaging technology, CoWoS (Chip-on-Wafer-on-Substrate).
“It’s a funny name for TSMC’s silicon interposer,” Chiang said, but his real problem was a lack of sales. Xilinx "ordered 50 wafers a month, so I became a joke in the company.”

18/25 The hurdle was that it cost 7 cents per sq. millimeter, when clients would only pay 1 cent. So he told his engineering team to cut the cost, even if that meant sacrificing some performance. The result, InFO (Integrated Fan-Out) wafer-level packaging. $TSM

19/25 TSMC's 2nd generation package technology, InFO, met the price point/performance and sold like hotcakes, he said.
“So that one word saved my life and the InFO was why Apple was hooked by TSMC.” $TSM $AAPL

20/25 Advanced packaging has become an even hotter topic in the semiconductor industry these days, as a way to push computing speeds forward. And in particular, for chiplets.

21/25 Joining China’s biggest chipmaker, SMIC
Chiang joined SMIC in December, 2016 as an independent director and left in mid-2019. Many ex-TSMC engineers have gone to SMIC.
“It was a mistake,” he says, calling it a "foolish" thing to do. $SMICY

22/25 Chiang was invited to join SMIC in late 2016 by then-CEO TY Chiu, a former TSMC colleague and also from the same hometown in China. Chiu had been criticized for falling behind in technology, so he asked Chiang for help.

23/25 At first, Chiang demurred, as SMIC was a TSMC rival. But then they agreed on a board position, which would mean Chiang would only visit SMIC once a quarter. Out of respect for his former employer, Chiang sought approval from Morris Chang before joining.

24/25 SMIC had to call a special board meeting to appoint Chiang as a new director, which made big news in Taiwan. “I didn’t expect that,” he said. “Before that, I had a pretty good image in Taiwan. That really hurt my image a lot.”

25/25 Born in China in 1946, Chiang was raised in Taiwan. In 1969, he moved to the United States to attend Princeton, then Stanford. He worked at Texas Instruments and Hewlett-Packard before joining TSMC. He resides in the U.S. today. $TXN $HPQ

The full interview can be found here: archive.computerhistory.org/resources/acce…