US gov't pushing TSMC and Intel to create joint venture in the US: Report

A Wall Street analyst's speculation suggests that the U.S. government is pushing Intel and TSMC to form a joint chip production venture, which would be owned by the two companies and operated by TSMC.

According to a Wall Street Journal report, the U.S. government wants TSMC to apply its experience with high-volume manufacturing using EUV lithography to Intel's process technologies. Another rumor cited by the Financial Times points to TSMC's accelerated Arizona Fab 21 complex expansion.

A weird rumor

"There are discussions from the Asia supply chain that the U.S. government will get involved in potentially the following: TSMC would send engineers to Intel's 3nm/2nm fab, applying the company's know-how to ensure that the fab and subsequent manufacturing projects from Intel become viable," Tristan Gerra, an analyst with Baird, wrote in a note sent to clients (via @WallStEngine). "The fab could be spun off into a new entity jointly owned by TSMC and Intel, and run by TSMC. The new entity would receive U.S. Chip Act funding."

Although Tristan Gerra admits there are no confirmations of the rumors and that the project could take a long time to materialize, the strategy apparently appears logical to analysts. If implemented, Intel's financial pressures would be eased, and it could shift its focus toward chip architecture and platform solutions. Also, Intel would retain its manufacturing capacity, aligning with the strategy set by former chief executive Pat Gelsinger. Additionally, an operational and competitive fabrication facility could draw major chip designers seeking a stable and geographically secure manufacturing alternative, which would ensure the utilization of the fab.

The rumor is odd, raising both business and technology concerns. Let’s start with the technology — two ways for semiconductor producers to “share know-how” would be by helping each other tune equipment for a particular process technology (or technologies, as Intel 3-capable fabs are likely also capable of 18A), or porting an established production node to a new fab.

Technology concerns

TSMC engineers are unfamiliar with Intel’s EUV-based fabrication processes and process recipes, so their ability to tune anything will be limited, or take months. Meanwhile, Intel’s 3nm node is in mass production, and the 18A manufacturing technology will enter high-volume manufacturing a few months later. Based on what Intel says about 18A, this process does not need further tuning, especially not adjustments made by external engineers.

Porting TSMC’s fabrication technologies to an Intel fab in the U.S. is challenging. Even though Intel and TSMC fabs might use EUV lithography, each has unique tool setups, fab configurations, and vendor-specific modifications. For example, ASML’s lithography machines typically feature custom calibrations and/or additional modules that suit the chipmakers' respective processes. Also, Intel and TSMC have different etch recipes, deposition sequences, and step-by-step controls, which usually affect tools.

Fabrication processes depend on carefully vetted materials with strict purity and composition requirements. Even small changes in photoresist or etch chemistries can introduce defects or cause process drift. Moreover, TSMC’s approved chemical suppliers and their formulations differ from Intel’s, and it is not guaranteed that the raw materials used by TSMC can effectively be used on Intel’s tools.

Even minor variations in temperature controls, gas flows, or photomask handling can cause significant drift in line widths, transistor characteristics, and yields, making porting complex, expensive, or financially unfeasible.

Business concerns

On the business side of the issue, TSMC has no interest in helping Intel improve its EUV fabrication technologies as Intel is a rival. The company also does not need to form a joint venture with Intel, as it could impact its gross margins. The only motivation for TSMC to work with Intel would be to avoid potential tariffs introduced by the Trump administration, which could require TSMC to produce advanced chips on its N3 (3nm-class) and N2 (2nm-class) process technologies in the U.S.

However, Intel is unlikely to have enough EUV production capacity in America to meet demand for its products and products from TSMC’s customers. To that end, accelerating the adoption of N3 and N2 in TSMC's own Arizona fabs, as reported by the Financial Times citing industry analysts, makes far more business sense.

For Intel, working with TSMC also seems impractical. It is unlikely that TSMC engineers can help significantly improve Intel 3nm and 18A process technologies. However, porting TSMC nodes to fabs in the U.S. means that Intel’s fabs in Ireland and Israel would be the only facility to use Intel’s nodes.

Yet, it is unlikely that these sites will be financially viable in the long run, as process technologies cost billions to develop, and the more chips a company produces using a single node, the better. For obvious reasons, TSMC is hardly interested in porting its advanced nodes to Intel facilities in Ireland and Israel, even as part of a joint venture, as this could create overcapacity for its fabrication technologies.

Thoughts

Ongoing geopolitical turmoil, coupled with Intel's financial and execution struggles, have generated various rumors surrounding the blue giant. In this case, significant technological and business hurdles — ranging from differences in tooling process recipes at Intel and TSMC to TSMC's lack of incentive to aid a competitor — cast doubt on the feasibility of such a partnership.