A person described as familiar with TSMC's internal affairs has publicly dismissed Elon Musk's Terafab semiconductor initiative, telling Wccftech that the entrepreneur "would be better off sending rockets to Mars than productively making chips through Terafab."

The blunt assessment lands as Musk pushes harder to bring Tesla's chip supply chain in-house and reduce the company's reliance on Taiwan Semiconductor Manufacturing Co. (TSMC), the world's largest contract chipmaker.

What is Terafab?

Terafab is Musk's umbrella label for a planned in-house chip operation intended to serve Tesla's autonomous-driving hardware and AI compute workloads. The project, as Musk has publicly described it, would span advanced logic production and advanced packaging — the two areas where TSMC holds its strongest competitive moat.

The Wccftech source offered no specific technical objections, cost figures, or timeline. The single substantive claim is the sweeping verdict: that producing chips at Terafab would prove less tractable than Musk's other flagship engineering effort — SpaceX's interplanetary rocket program.

Why does the Mars-rocket comparison sting?

The analogy is pointed. SpaceX has spent roughly two decades iterating Falcon 9, building Starship, and reaching commercial reusability. A leading-edge fab demands comparable patience across:

• Cleanroom construction
• EUV tool installation
• Process recipe development
• Yield ramp
• Customer qualification

Each step is a multi-year program on its own.

How has Tesla sourced chips so far?

Tesla has already cycled through several generations of custom silicon. HW3, deployed from 2019, used an in-house design built by Samsung on 14nm. HW4, introduced in 2023, shifted to a 7nm TSMC node. The Dojo training system, unveiled in 2021, runs on a custom D1 chip at 7nm. Terafab would mark Tesla's first move to fab ownership.

Why is building a fab so hard?

Building a competitive wafer fab from a greenfield site ranks among the most capital-intensive industrial projects on earth. Industry consensus places the cost of a leading-edge fab at $20 billion or more before a single wafer ships at yield. Construction alone can run two to three years.

TSMC took more than a decade after its 1987 founding to reach technological parity with established IDMs. Samsung, Intel, GlobalFoundries, and SMIC — all with comparable or longer foundry histories — currently operate one or more process generations behind TSMC's leading edge.

The gap compounds over time. Yield learning curves, customer trust, and refined process recipes mature over decades. Without them, a new fab produces wafers, but at cost and quality levels the open market does not reward.

Why does Musk want a fab?

Tesla's autonomy stack depends on advanced silicon. The Full Self-Driving (FSD) computer and the Dojo training system have both relied on TSMC's leading-edge nodes. Musk has framed Terafab as a hedge against foundry capacity constraints and a path to lower per-unit silicon cost, in line with Tesla's vertical-integration strategy across batteries, motors, and software.

The Wccftech source did not address those strategic motivations directly. Their criticism focused on the manufacturing reality: that no well-funded entrant has closed the gap to TSMC's leading edge inside a decade.

What comes next for Terafab?

The Wccftech report does not name the source, does not specify a Terafab timeline, and does not disclose a target process node, wafer size, or capacity figure. Musk has not publicly responded in the available record.

If Terafab follows the path of every prior foundry entrant, the next milestones will be capital commitment, ASML tool orders, and a site announcement — each a multi-billion-dollar, multi-year step. Whether Musk can compress that timeline, or whether the Wccftech source's verdict proves accurate, will become visible only as Terafab either ships wafer or fails to.