店裡賣接頭賣了三十年，客人記得的永遠是某某牌馬桶、某某牌熱水器，沒人問過裡面那顆止逆閥是哪家做的。半導體這行也一樣。台積電（TSMC）三個字大概全台灣國小生都會唸，但聯電（UMC）出身、後來自己創業的黃崇仁，這個名字知道的人少得多。他辭世的消息傳出來，才有人想起，手上這支手機、家裡那台冷氣機控制板，說不定就有一顆晶片來自他一手拉起來的PSMC。
不是台積電體系,才是重點
台灣半導體故事講久了,容易講成一條線:台積電、日月光、聯發科,三個招牌撐起整個門面。PSMC偏偏不是這條線上的公司。它是少數不靠台積電體系長大、自己一路熬出規模的晶圓代工廠,规模上排進台灣第四。一個生態系如果只剩一家獨大,出貨排程、報價權、甚至地緣風險都集中在同一個節點,反而脆弱。有第二條、第三條血脈自己長出來,系統才耐震。PSMC存在的意義,某種程度上就是這個。
黃崇仁是台灣半導體第一代創業者之一那個年代,懂技術、懂資本、還願意自己扛風險開廠的人不多。這種人退場,留下的不只是一張董事長座位的空缺,是一整套只有創辦人腦子裡才有的判斷邏輯——跟哪家設備商談價、哪個客戶的訂單該不該接、景氣差的時候現金怎麼撐。這些東西不會寫在交接手冊裡。
交棒撞上AI浪頭,時機最壞
偏偏這次交班撞上AI晶片需求暴衝的檔口。訂單湧進來,產能、資本支出、技術路線的決策一個接一個,新的領導班子得在最吃緊的時候接手方向盤。水電行老闆最懂這種心情——師傅退休那天,剛好接到一張最難搞的訂單,新來的學徒手還在抖。晶圓廠當然比水電精密千百倍,但换班時最怕遇上旺季,這道理是一樣的。
交棒順不順,直接關係到PSMC能不能繼續在AI供應鏈裡卡住位置。卡不住,不是公司自己的損失而已,是台灣整個生態圈少了一根備用樑柱。半導體供應鏈這幾年被講得像神話,但神話底下其實是一堆不出名的中小型晶圓廠、封測廠、材料商,撐著台積電上不了新聞的那些訂單缺口。PSMC是這群「隱名英雄」裡規模最大的一個。
沒有名字的支柱,才是真正的地基
台灣半導體被講成護國神山群,鎂光燈永遠打在最高那一棟。但神山群能撐住,靠的是山腳下那些沒被拍到的岩層。黃崇仁這一代創業者最大的貢獻,或許不是做出多驚人的技術,而是證明了台積電體系之外還能長出一家能打的晶圓廠。這條路現在要換人走,能不能走穩,決定的不只是一家公司的股價,是整個生態圈還撐不撐得住多元。
店裡的接頭壞了可以換一顆一樣的。創辦人這種角色壞了,換上去的人得先花好幾年,才知道原來那顆閥門的壓力該調到多少。
— 巫德建
PSMC’s Founder Is Gone. The Chip Maker Nobody Names Still Holds the Chain Together
Nobody in a hardware store asks which factory made the check valve inside their toilet tank. The toilet’s brand gets remembered, never the part’s. Taiwan’s chip industry works the same way. TSMC’s name gets recognized everywhere. Far fewer names belong to Yao Chung-jen (黃崇仁), the man who built PSMC into Taiwan’s fourth-largest wafer foundry after coming up through UMC. His death is the reason phones and appliances are suddenly getting checked for what’s inside — and turning out to carry a chip from a company that went unnoticed until now.
Not Built Inside TSMC’s Orbit
The usual version of Taiwan’s semiconductor story runs through three names: TSMC, ASE, MediaTek. PSMC never grew inside that lineage. It’s one of the few wafer foundries that scaled up independently, outside the TSMC supply ecosystem, and still landed at fourth place nationally. That matters more than it sounds. A supply chain concentrated around a single dominant node — pricing power, scheduling, geopolitical exposure all funneled through one company — is a fragile one. A second and third lineage growing on their own terms makes the whole system harder to break. PSMC’s existence is, in some sense, exactly that kind of shock absorber.
Yao belonged to Taiwan’s first generation of semiconductor founders — people who understood both the engineering and the capital risk of opening a fab from scratch. Few of them are left. What disappears with a founder like this isn’t just a chairman’s seat. It’s a set of judgment calls that never made it into any handbook: which equipment vendor to trust on price, which client’s order to prioritize, how to hold cash steady through a downturn.
Succession Lands in the Middle of an AI Boom
The timing is rough. Orders for AI-related chips are surging, and decisions on capacity, capital spending, and technology roadmaps are piling up fast — exactly when a new leadership team has to take the wheel. Anyone who’s run a small shop knows the feeling: the veteran retires the same week the hardest order of the year comes in, and the new hire’s hands are still shaking. A fab is a different order of complexity, but the bad timing of a handover during peak season is the same problem, scaled up.
Whether PSMC’s transition goes smoothly decides whether it keeps its footing in the AI supply chain — and losing that footing wouldn’t just cost the company. It would strip Taiwan’s ecosystem of one of its backup pillars. The semiconductor supply chain gets talked about like a myth, but underneath the myth are dozens of foundries, packaging houses, and materials suppliers that TSMC’s headlines never mention. PSMC is the largest of that unnamed middle layer.
The Support Beam With No Name On It
Taiwan’s chip industry gets called a mountain range guarding the country, and the spotlight always lands on the tallest peak. What holds the range up is the bedrock nobody photographs. Yao’s real contribution may not have been a technical breakthrough — it was proving a competitive fab could grow outside TSMC’s orbit at all. Someone else has to walk that path now. Whether that path stays steady decides more than one company’s share price. It decides whether the ecosystem stays plural.
A broken valve gets swapped for an identical one. A founder doesn’t work that way — whoever takes over spends years just learning what pressure that valve was set to.
— 巫德建
延伸閱讀
https://justfly.idv.tw/s/41zK5M5