Speaking at an event in Austin on Saturday, Musk announced that the project, TerraFab, will launch as an “advanced technology fab” aimed at producing two types of chips used by Tesla and newly merged SpaceX and XAI.
One of those chips, Musk said, will be optimized for “edge and inference,” which will be used primarily in Tesla’s autonomous driving ventures and the Optimus humanoid robot. The second would be a high-powered chip that would be used in space, on the way to Musk’s long-touted goals of transforming humanity into a “galactic civilization.” At the center of that mission right now is Musk’s space-based data center plans, which are also fueling the company’s IPO plans for later this year.
Musk has been talking about the project for some time, which he called on Saturday “the most epic chip manufacturing exercise ever in history,” but this event was its official launch.
Although he is “very grateful” to existing chip suppliers such as Samsung, TSMC and others, he said the chip giants are not “comfortable expanding” at the maximum rate he would like.
“We need chips. So we’re going to build TerraFab,” Musk said after sharing his goal of producing a terawatt of output per year through TerraFab to meet AI’s growing power demands.
“We will have all the equipment we need to make any type of chip,” Musk said. “We’re really going to push the limits of physics and computation, and we’re going to try a lot of wild and crazy things.”
Ultimately, Musk, who is known for his tendency to overpromise, may actually need to raise at least some limits as the plan is very ambitious and likely to be incredibly expensive.
The company already plans to spend more than $20 billion this year, without even taking into account the cost of TerraFab. This is almost double the amount spent in 2024, the most expensive year so far. Musk has been teasing these plans for some time, and a UBS analyst estimated earlier this year that his vision for TerraFab could cost the company as much as $300 billion.
Chip factories are also very difficult to build and typically take longer to build in the US than in Taiwan. Leave alone fully operationalizing the factory, it may take a few years to even acquire the machinery. Musk’s plan is particularly ambitious and will require deep expertise, which neither Musk nor the US chip industry currently possesses.
One of the few domestic chip plant construction projects in the country in nearly a decade is being driven by highly experienced chip giant TSMC, and even that project is reportedly plagued with numerous issues and an eye-watering financial commitment, having been deemed “one of the most expensive ventures on earth” by The New York Times.
<a href