Musk Unveils $5T--$13T Terafab Chip Plan Targeting 1 Terawatt Compute

Elon Musk signaled a potentially aggressive expansion into semiconductor manufacturing over the weekend, outlining plans that could position Tesla TSLA, alongside xAI and SpaceX, closer to building large-scale chip production capacity. Speaking in Austin, Musk described the Terafab concept as an effort to materially expand computing supply for artificial intelligence, robotics and space applications, areas he indicated may require far more silicon than is currently available. Estimates referenced by analysts suggest the project could involve $5 trillion to $13 trillion in capital spending, supporting hundreds of fabrication plants and targeting up to one terawatt of annual compute capacity, a scale that may rival or exceed today's global semiconductor footprint.

The backdrop to Musk's proposal reflects intensifying demand across the AI ecosystem, with hyperscalers such as Amazon, Alphabet and Microsoft expected to allocate roughly $650 billion this year toward data center infrastructure. That level of spending is already contributing to tighter supply in memory and beginning to extend into AI accelerators, reinforcing concerns that chip production may not be keeping pace with projected demand. Musk indicated he has encouraged suppliers including Taiwan Semiconductor Manufacturing TSM, Samsung Electronics (SSNLF) and Micron Technology MU to expand capacity more rapidly, even suggesting he would absorb all available output, though leading foundries continue to emphasize a disciplined approach to scaling production.

Analysts remain cautious on execution, pointing to the structural and economic challenges associated with advanced chip fabrication. Building cutting-edge facilities requires tens of billions of dollars per site and carries risks tied to rapid technological obsolescence, while the industry has largely shifted toward outsourcing models that concentrate volume through players like TSMC. Musk's approach, which appears to lean toward a more vertically integrated model, could introduce additional complexity and cost pressures. Even so, his remarks highlight a broader industry tension, where accelerating AI demand may be testing the limits of existing semiconductor capacity and potentially reshaping how supply is built over time.