1 hour ago
3
Elon Musk wants to build a chip factory so large it makes existing semiconductor fabs look like lemonade stands. Then he wants to launch most of what it produces into space.
The project is called Terafab, and it represents a collaboration between Tesla, SpaceX, and xAI to construct a vertically integrated semiconductor fabrication facility near Austin, Texas. The target: annual production capacity exceeding 1 terawatt of AI compute power, roughly 50 times the current global output.
Initial cost estimates sat between $20 billion and $25 billion, but more recent regulatory filings have revised the figure upward to approximately $119 billion.
What Terafab actually involves
The facility is planned to span approximately 100 million square feet. For context, that’s nearly four times the size of the Pentagon, which is already the world’s largest office building by floor area.
Terafab’s production will center on 2nm-class chips. Two distinct chip families are in the pipeline: “D3” chips designed specifically for space deployment, and AI5/AI6 chips intended for ground-based applications like Tesla’s full self-driving technology and xAI’s model training infrastructure.
Roughly 80% of Terafab’s output is earmarked for orbital deployment. The remaining 20% stays on Earth to serve Tesla’s autonomy and robotics ambitions alongside xAI’s compute-hungry operations.
Intel has been mentioned as a manufacturing partner, which would represent a significant strategic alignment for the chipmaker as it works to rebuild its foundry business.
The production targets are staggering: between 1 billion and 10 billion units per year, with the exact figure tied to Starship’s launch cadence.
The orbital compute thesis
The logic behind putting AI compute in space comes down to a constraint that most people never think about: power grids. Musk has argued that terrestrial electrical infrastructure simply cannot support the density of compute that advanced AI systems will demand. Solar energy in orbit, unfiltered by atmosphere and available nearly 24/7, offers a workaround.
SpaceX’s Starship is the linchpin of the entire operation. The plan calls for transporting around 10 million tons of payload to orbit each year to build out solar-powered AI compute infrastructure.
The D3 chips heading to orbit would power AI satellites, essentially creating a distributed computing network in space.
Why crypto and tech investors should pay attention
The $119 billion commitment, if realized, would represent one of the largest single infrastructure investments in technology history.
For the crypto sector specifically, the intersection of AI and decentralized compute has been one of the hottest narratives driving token valuations. Projects focused on distributed GPU networks and AI-oriented blockchain protocols have attracted billions in market capitalization on the premise that AI compute demand will outstrip centralized supply. A project like Terafab, which proposes to multiply global AI compute capacity by a factor of 50, could either validate that thesis by demonstrating insatiable demand, or undermine it by suggesting centralized mega-fabs will absorb the shortage.
There’s also the regulatory dimension. A project of this scale in Austin has already triggered significant community engagement around tax incentives and supplier negotiations.
The risk profile is equally notable. The gap between the initial $20-25 billion estimate and the revised $119 billion figure is the kind of cost escalation that has derailed megaprojects throughout history. Starship’s launch cadence remains unproven at the scale required, and manufacturing 2nm chips at the volumes described would push semiconductor fabrication into territory no company has navigated before.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) · 