1 Terawatt Annual Computing Power! Tesla: Terafab to Be Built in Texas with Initial Investment of $16.8 Billion

On Thursday (the 6th), Tesla announced on the X platform that its ultra-large wafer fabrication facility, codenamed 'Terafab,' will be located in Grimes County, Texas, to meet the AI and aerospace chip demands of both Tesla and SpaceX.

The post stated: 'The chip demand from Tesla and SpaceX will far exceed the current and future global production capacity.'

Terafab is positioned by Tesla as the 'largest chip manufacturing facility in history,' aiming to produce over 1 terawatt (1 TW) of computing power annually.

Based on current AI accelerator power consumption levels, this equates to an annual delivery capacity of millions of H100-class chips, covering FSD training for autonomous driving, the brain of the Optimus robot, Starlink ground processing, and Starship guidance and control links.

Texas Governor Greg Abbott simultaneously released a statement indicating that the first-phase capital expenditure for Terafab exceeds $16.8 billion, creating 3,000 jobs. The Texas Enterprise Fund has allocated $30 million to SpaceX as part of a coordinated site development initiative.

Industry insiders interpret that Tesla is leveraging its internal cash flow from automotive and energy storage businesses to vertically integrate and bypass bottlenecks from foundries like TSMC and Samsung, as well as constraints in HBM supply. However, it will take at least 3–5 years from groundbreaking to achieving stable yields, so the short-term AI chip supply-demand landscape won't change significantly. Instead, this move is seen more as securing production capacity ahead of mass production of autonomous vehicles and robots in the 2030s.

Analysts point out that Terafab is still in the announcement phase, with no disclosure yet on process nodes, EUV sources, or joint venture partners. The $16.8 billion initial investment remains modest compared to Intel or TSMC's single-fab investments often exceeding $40 billion, suggesting an early-stage scale. Future success will depend on whether land, water, electricity, labor, and ASML equipment scheduling can be secured effectively.