Elon Musk's $25B 'Terafab' chip mega-project

Running the world’s biggest EV company, a dominant rocket business, and a top-tier AI lab apparently isn’t enough. Elon Musk now wants to build his own silicon from scratch.

Over the weekend in downtown Austin, Texas, Musk unveiled plans for "Terafab," an ambitious chip fabrication plant to be jointly operated by Tesla and SpaceX, with AI startup xAI in the loop. Musk framed the project as the most epic chip-building exercise in history, claiming the goal is to produce roughly one terawatt of AI computing power annually—around 50 times the world’s current global compute output.

Vertical integration meets orbital servers

Musk’s stated motivation was straightforward: external chipmakers are moving too slow to keep up with his robotics and AI ambitions. He reportedly put it bluntly: we either build the Terafab or we do not have the chips, so we build it.

Reports place the project's ambition at around $25 billion, with the plant expected near Tesla’s Austin headquarters and Gigafactory. The facility is planned to start small with advanced tech testing before scaling up, taking on logic, memory, packaging, and testing under one roof—a degree of vertical integration Musk claimed exists nowhere else in the world.

Two chip variants are in the works: one destined for Tesla vehicles and Optimus humanoid robots, and a space-grade chip designed for solar-powered AI satellite clusters launched aboard Starship. Musk argued that ground-based data centers face local pushback, predicting orbital compute will undercut terrestrial costs within two to three years. Terafab targets 100 to 200 gigawatts of annual compute capacity on Earth, alongside a full terawatt in space.

Big sci-fi talk, zero timelines

Musk pitched Terafab as a foundational step toward a galactic civilization and a post-scarcity future. Yet, as multiple reports highlighted, the grand announcement lacked one crucial detail: an actual timeline. Musk offered no launch dates for when the fab might break ground, come online, or reach those astronomical production targets.

Industry watchers and reporters quickly pointed out the reality check. Building a leading-edge fab from scratch requires tens of billions of dollars, highly specialized lithography gear, and years of yield optimization. Musk has no formal semiconductor manufacturing background, alongside a long track record of overpromising on aggressive product timelines.

Why it matters

Terafab looks like the logical extreme of Musk’s obsession with vertical control, but we think anyone in AI infrastructure should treat this with heavy skepticism for now. There is nothing to test or sign up for, and chip manufacturing is an entirely different beast from building rockets or electric cars. It relies on a hyper-specialized global supply chain, and going from zero to leading-edge packaging and memory in-house has humbled seasoned chip giants before.

Still, the strategic thesis explains where the industry's pain points are headed. Data centers on Earth are running out of cheap power, cooling, and grid bandwidth. Musk’s idea of pairing space-grade chips with Starship launches to bypass ground power politics is a clever conceptual flex, even if the physics and economics of thermal dissipation and orbital radiation hardening remain brutal hurdles.

If this project ever materializes, it would signal a dramatic break from relying on TSMC and commercial foundries. But until Tesla and SpaceX demonstrate a working pilot line with viable yields, Terafab looks less like an imminent chip threat and more like a visionary pitch to keep markets excited about an autonomous, space-based compute future.