Tim Culpan, founder of the independent media Culpium, broke the news that TSMC is evaluating cooperation with Musk to assist him in the Terafab wafer fab project in Texas. In response to this rumor, Musk said, "It's only in the discussion stage, but maybe things will make progress," which aroused market attention. Terafab had originally approached Intel and planned to adopt the 14A advanced process, and the US government holds nearly 10% of Intel's shares. Now that Musk is still looking for TSMC, foreign media analysis highlights the cruel reality of advanced chip manufacturing. Moreover, TSMC Chairman Wei Zhejia said more than five months ago: "There are no shortcuts in the foundry field."
Musk's Tesla, SpaceX and xAI have rapidly expanding demand for artificial intelligence (AI) computing power. In order to control the supply of chips, Musk has promoted the massive Terafab project, hoping to establish semiconductor manufacturing capabilities covering logic chips, memory and advanced packaging.
Now Terafab may welcome a heavyweight partner. Musk recently confirmed on the social platform
Technology media "Tom's Hardware" reported that there are at least two plans A and B among the cooperation ideas currently on the table. "Plan A" has TSMC playing a leading role. TSMC will own and operate the new chip manufacturing plant, while SpaceX or Terafab may provide funding, commit to purchasing a certain number of chips, or both.
The report pointed out that this model is similar to TSMC’s joint venture layout in Japan and Germany. The partners provide funds and needs, while TSMC is responsible for the process technology and fab operation capabilities.
"Plan B" gives Musk greater control, with SpaceX holding a majority stake in Terafab and TSMC only holding a minority stake, but TSMC will still provide process technology and fab operation expertise. The report believes that if this structure is finally adopted, it will be an unprecedented cooperation model for TSMC.
However, Plans A and B are still in the discussion stage and have not yet been decided, and other ways of cooperation are not ruled out in the future. There is an obvious common point behind the two plans: No matter who ultimately owns the factory or invests more money, Terafab is looking at TSMC’s advanced manufacturing processes and fab operation capabilities accumulated over the years. This also highlights the reality that Musk must face when building his own chip supply chain.
According to report analysis, what Musk is really pursuing is not necessarily to build Terafab into another foundry company competing with TSMC, Intel or Samsung, but to ensure that Tesla, SpaceX and xAI can obtain sufficient and uninterrupted supply of customized chips. Therefore, as long as this goal can be achieved, Musk is likely to maintain considerable flexibility as to who owns the fab, who operates it, and even which process it uses.
It is worth mentioning that in April this year, after Intel announced that it would join the Terafab project, foreign investors directly asked Wei Zhejia at the TSMC legal meeting whether Tesla's construction of its own wafer fab and cooperation with Intel would have an impact on TSMC.
Wei Zhejia said at the time that both Tesla and Intel were TSMC customers, and Intel was a strong competitor. TSMC would never underestimate any competitor. However, he immediately emphasized that "there are no shortcuts in the foundry field." Technology leadership, excellent manufacturing, customer trust and service are all TSMC's competitive advantages.
At that time, Wei Zhejia also pointed out the time threshold for wafer manufacturing. It takes 2 to 3 years to build a new wafer fab, and another 1 to 2 years to ramp up production capacity. Therefore, there are no shortcuts not only for wafer foundry, but also for expansion of production capacity.
Now less than half a year later, Terafab, which originally planned to build its own huge chip production capacity and introduce Intel's 14A process, has begun to negotiate cooperation with TSMC, and even discussed that TSMC will provide the process and fab operation capabilities. This has also brought Wei Zhejia's original statement that "there is no shortcut" to attention again.
This also makes Intel's role nuanced. Musk only announced in April this year that Terafab plans to use Intel's 14A process. If it proceeds smoothly, Terafab is expected to become a very important heavyweight customer of 14A, which is of great significance to Intel, which is striving to revive its foundry business.
Moreover, the U.S. government itself is Intel’s largest shareholder. The Trump administration will spend US$8.9 billion to acquire approximately 9.9% of Intel's shares in 2025, making a direct bet on the domestic chip manufacturer. However, the government can invest funds to support Intel, but real customers must ultimately consider process technology and mass production capabilities.
The report pointed out that if TSMC eventually enters Terafab, takes charge of the fab operations and introduces its own process, the originally planned Intel 14A may lose its position; however, another possibility is that Terafab will still use Intel 14A to manufacture chips, and TSMC will provide advanced packaging and other services.
The report also pointed to another potential consideration, namely patent issues. Intel and TSMC have extensive cross-patent licensing arrangements, but Terafab itself does not have the same arrangement. If Terafab uses the process developed by Intel to produce its own chips, how to avoid related intellectual property risks in the future is also an issue that must be considered. Therefore, the report believes that Musk has reasons to keep TSMC within Terafab’s cooperation circle.
