TSMC Tried a 24-Hour R&D Schedule. It Lasted Six Months.

October 30 — Y.J. Mii, TSMC’s senior vice-president and co-chief operating officer, told a group of university students that extending working hours does not effectively raise R&D efficiency.

He recalled that when TSMC was behind on process technology and anxious to catch international rivals, it tried a 24-hour R&D programme called “Night Owl,” putting R&D engineers on overnight development work in the hope that longer operating hours would produce a technical breakthrough. The programme was declared a failure after only about six months.

According to the Economic Daily News, the thinking inside TSMC at the time was that if its R&D teams worked around the clock while US competitors worked eight-hour days, the company could build a time advantage and catch up.

In practice, the idea did not deliver the expected results. Mii concluded that R&D depends on thinking, judgement and innovation — it is fundamentally “brain work, not labour” — and that simply adding hours does not translate into higher R&D efficiency.

The failure also pushed TSMC to rethink how R&D and manufacturing work together. The old process — R&D finishes developing a process, then hands it to the fab — ran into trouble; over time the company built a “One Team” culture in which R&D and manufacturing work towards shared goals and share responsibility for the outcome.

Mii gave an example: when a new process moves from development to mass production, ramp-up goes more smoothly if the fab is involved in development early and the handover overlap is extended. And when a process problem appears on the line, several teams analysing the root cause together is faster than one department investigating alone.