Terafab: Elon Musk's plan to build 358 chip fabs could cost up to $5 trillion and reshape the global semiconductor industry

A semiconductor project valued at up to $5 trillion would consume more than 70% of the US annual budget and require the construction of as many as 358 fabrication plants. This is not a science-fiction scenario, but the figures analysts cite when evaluating Elon Musk's Terafab ambition. In the context of the global AI race becoming increasingly intense, if realized, the plan could redefine the entire semiconductor industry. However, the question remains: is this a bold strategic move or an ambition beyond practical limits? According to Bernstein's analysis, to realize a target of producing AI chips consuming 1 terawatt (1 TW) of electricity per year, the Terafab project would require an unprecedented, massive production ecosystem in the history of semiconductors. Specifically, to reach this capacity, the system would need to process annually: 22.4 million GPU Rubin Ultra wafers, 2.716 million Vera CPU wafers, and 15.824 million HBM4E wafers. In total, this output would require between 142 and 358 semiconductor plants operating simultaneously. These figures exceed current norms. By comparison, TSMC, the world's largest chipmaker, produced about 15,023 million wafers (300 mm equivalent) in 2025, with around 50 module factories built over more than two decades. Musk's ambition, therefore, is not merely to build a few new plants, but to create a 'super ecosystem' for chip production on a scale that would exceed the current industry. Notably, the initial $20 billion investment for Terafab, reportedly to produce logic and memory under one roof, is actually only enough to build a single 7nm fab. Meanwhile, the long-term goal is to produce millions, even billions, of AI chips. The $5 trillion figure Bernstein cites even matches Sam Altman's earlier ambition to build a global network of chip factories, a plan that ultimately did not come to fruition. Terafab’s ambition to dominate the semiconductor industry. The final question is Musk's true objective. Is he seeking to build a chip manufacturing system large enough to surpass the combined size of TSMC, Intel, and Samsung to serve companies like Tesla, SpaceX, or xAI? At present, the answer remains open. However, if Terafab becomes a reality, it could be a watershed in global technology supply chains where power shifts from software or data to hardware manufacturing at an unprecedented scale. If it fails, it could be one of the most expensive bets in tech history. Source: Tom's Hardware