News Commentary|Reuters: TSMC Proposes Joint Venture with Phaidon, Supermicro for Intel Foundry
Commentator: Mr. Lin Wei-Pu, Executive Vice President, Ji-Pu Industry Trend Research Institute
News.
According to an exclusive report by Reuters, TSMC has extended an invitation to Fidelity, Supermicro, Broadcom and Qualcomm to set up a joint venture to operate Intel's wafer foundry division. According to the source, TSMC will be responsible for the actual operation of this cooperative case, but the shareholding ratio will not exceed 50%, and must obtain the consent of the U.S. Trump administration.
When TSMC announced plans to invest an additional $100 billion in the U.S. and build five wafer fabs in the future, it already mentioned the joint venture to potential supporters, in response to Trump's goal of revitalizing Intel and boosting the U.S. advanced manufacturing industry. Negotiations are still at a preliminary stage, with final details yet to be negotiated.
Ji-Pu Viewpoint:
Conclusion: It's too hard to predict what the future holds for us at this point in time, so unless it happens, I'm going to think it's a story like Loretta's. I don't really believe Intel will be owned by an Asian. I don't really believe Intel will be owned by an Asian, as the CEOs of TSMC and Avago are all Asian-run companies. The current tendency is not to become, the following is my background / current situation / reasons for the discussion:
#. background:
1. As a representative of the once dominant semiconductor industry in the United States, Intel's collapse or closure under any president would be a major blemish, and the Trump administration, which now boasts Made in the USA, would not allow it to happen, so it leads me to believe that this should be a real thing in the works.
2. On the afternoon of March 3, U.S. time, TSMC Chairman Chieh-Chia Wei announced at the U.S. White House a $100 billion expansion of its investment in three advanced process fabs, two advanced packaging fabs, and a research and development center. The global industry, especially in Taiwan, had different reactions from both positive and negative points of view. Regardless of the real reason, with such a high investment and the R&D center, TSMC must be under tremendous pressure from its "customers".
#.
1. Intel is currently in the process of splitting up its factories, which is expected to be completed by 2026, to turn them into subsidiaries for the foundry business (hereinafter referred to as subsidiaries_I). In addition, Intel's advanced process is not as good as TSMC's, but it is the closest and has a better chance than the other competitor in the advanced process area.
2. I think Intel's current problems are
#. 由台積電作東找IC設計巨頭[輝達、超微、博通、蘋果(不確定) ]一起投資Intel,且由台積電經營的可能性小的原因如下:
1. 美國的白人主義,不會讓亞洲人入主甚至控制美國指標性企業(之前Avago 併購博通就被擋了好久,應該也被認為是個污點。另外,讓台積電做東,找資金投資美國代工廠來救Intel?除了完全讓美方沒面子外,讓台積電用技術去救Intel,那何苦台積還要宣布研發中心的設立[除非有我結論所提到的假設論述]?
2. 這些設計公司,投資子公司_I會有與客戶競爭的問題[AMD有x86的CPU & GPU;輝達有GPU,雖然目前Intel 的Gaudi處於更換架構後市場定位不明的情況;博通有矽光子與部分通訊晶片,但整體跟Intel目前產品重疊相對較少],未來他們真的敢下單在這裡?有些記者問我時會說:川普要救Intel 不會管這麼多。我同意,但如果是這樣,直接限制大家到台積電下單轉單給Intel 不是更快?何苦這麼麻煩?
3. 找這些IC巨頭一起投資Intel,但沒有客戶的問題還是沒解決(製程良率爬升不是投錢就搞定),因此這些IC設計巨頭勢必會面臨,投資子公司_I還要綁產品在這裡,陪你一起調整製程爬升良率等增加產品出貨時間?投資反而讓自己在市場競爭力下降,在我看來不太合理。
4. 在拆分Intel的設計與製造正在進行的當下,讓這些IC設計巨頭入主子公司_I,在產能無法擴充的情況下,分掉原本Intel 母公司x86 CPU的產能,拱手讓出主要命脈目前的優勢甚至霸權。以我看來母公司應該是有千百個不願意。
5. 假設,真的讓台積電入主技術後的Intel 還是原本的Intel? 呼應原因1,老白美國引以爲傲的技術,全部改成台積的,他們能忍受?另外,就過去經驗台積廠區之間同樣機台、同樣製程、同樣工程師轉移都需要不少時間。如果要現在談台積要用技術救Intel 不是一兩年就可以完成的,沒有這麼容易。以現在這時程推測,全部投資可能6月甚至更晚,實際執行台積技術入主可能要拖個半年以上。加上製程設備建置、調整等,可能真正開始生產兩、三年過去了,Intel 也可能已經餓死了
6. 要這些設計公司走回IDM的老路,去投資原本IDM拆分的工廠?重新做垂直整合?不知道是我認知淺薄還是美國當家的人沒有半導體專家?所以要求台積電這要做?[這情境,讓我想到周星馳行運一條龍電影裡面:吳孟達身為一個日薄西山的蛋塔店老闆,提議要原地轉型賣海鮮火鍋,並且可以去對面倒閉的海鮮火鍋店買她不要的鍋具爐具,真的有異曲同工之處。]
結論與可能的方案:一味的說不可能,在現在這個大川普時代反而有點中二或者說拒絕接受川普不可預測不按牌理出牌的特質。上面敘述都是以美國、IC設計巨頭或Intel的角度,成事的機率都不太可能。那台積電的角度,是否希望這件事情成呢?我認為,可能會與前面巨頭持相反態度。首先,依據背景2,台積有來自客戶大量投資美國的壓力。如果這件事情成了,我認為台積電因此可以拉高對外宣傳協助拯救Intel 的代工廠的態勢,讓新增加1000億美元的投資獲得舒緩或技術性延遲。其次,以過去經驗台積電也有投資過其他代工業者,像世界先進、中芯國際(和解協議獲得股權)等,最後台積電地位沒有因此受動搖仍然是市場的霸主,因為重點在於被投資的人競爭力,投資這些公司對台積電本身還可能做出相關市場區隔,降低反壟斷的風險。我相信台積電對自己不論在亞洲或世界的競爭力都相當有信心,被投資的晶圓代工公司在核心能力不夠強的情況下,商業競爭會讓他最後敗下陣來,即使有國家勢力的情力介入情況下。最後,如果這投資案真的成了,而且台積電也真的出了技術,往後幾年壓力就會有部分回到這些IC設計巨頭身上,對台積電來說不一定是絕對壞事。
綜觀目前現況,猜測川普政府在半導體產業想要的東西應該會是
1. 先進製程能在美國生產且有很大一個部分在美國生產。魏哲家董事長已經幫忙宣布,解決了這問題。
2. 讓Intel 活下去。
關於後者,以下是我認為可能的解決方法,以此來做為這篇新聞衍生出比較可能的劇本如下:
A. 解決Intel 核心問題:我認為核心問題解決方法會是,把一部分的客戶趕過去或讓給Intel,讓他在子公司_I下單[如現況2,蘋果會扮演非常重大的角色。而趕過去的方法有一些,之前的關稅可能是,未來有沒有其他的方法還要再觀察]
B. 讓Intel 的AI NPU或是ASIC 再接下來的AI Server 上能搶下一定的市佔(不是GPU,在GB300後的下一代Rubin, GPU 跟CPU 沒有綁在一起,這就有許多想像了),只要在未來兩年讓Intel 先進製程搶到一個客戶,他們就很有機會活下來。並且結合Chiplet 的封裝增加產品競爭力,讓他持續量產。
在這個半導體晶片升級為國家戰略物資的世代下,任何決定都不會有完美的,重點是要怎麼做才能保有自己最核心的競爭力,才是台積電董事長需要考慮的以及取捨的事情,我們相信台積電在歷經過去幾十年的歷練,最後一定能走出一條最好的道路。






