Why Elon Is Building the Largest Factory in History

Kind: captions Language: en There is a hard limit on everything Elon Musk is building. He's known about it for years. He just announced the only solution that exists. Last week, Tesla and SpaceX announced they will jointly build a semiconductor manufacturing facility in Grimes County, Texas, outside Houston. Terafab. Initial investment, $16.8 billion. Plan plan size, over 100 million square feet. 3,000 jobs minimum. Most coverage treated this as a construction announcement or a job story. It is neither. Here is how Musk described the logic behind it in his own words. We either build the Terafab or we don't have the chips, and we need the chips, so we build the Terafab. Musk has stated publicly that current worldwide semiconductor manufacturing capacity would fulfill just 2% of his company's projected future requirements. 2%. Not a supply chain inconvenience, not a temporary shortage, a structural ceiling. To understand what that means, consider what his companies are building toward. 10 million Optimus robots annually. Hundreds of Starship flights per year. An orbital constellation of AI data centers. Autonomous vehicles operating without human oversight. Neuralink devices interfacing biological minds with digital systems. Every single one of those products requires chips. Specialized, high-performance, manufactured at enormous scale. The chips that currently exist from TSMC, Samsung, and Nvidia are extraordinary. Musk acknowledged that explicitly at the Terafab announcement. But gratitude doesn't solve the math. If current global chip output satisfies 2% of what his companies will eventually need, and those companies are already building toward that future at full speed, then external chip supply is not a constraint that can managed around. It is a wall. Terafab is the door. Here is the detail that makes Terafab unlike any semiconductor facility ever built. It will manufacture two completely different chips for two completely different environments. The first, the AI5, powers Tesla's vehicles and Optimus robots on the ground. Inference chips optimized for real-world physical navigation, computer vision, and autonomous decision-making in Earth's atmosphere. 25% of Terafab's total output goes to this chip. The second, the D3, is built for somewhere no commercial semiconductor has been manufactured for at scale before. Space. The D3 is engineered specifically for SpaceX's orbital AI data centers, designed to withstand extreme temperature fluctuations, radiation exposure, and the conditions of low Earth orbit. 75% of Terafab's output is allocated to this chip. Three quarters of the largest chip factory on Earth is being built to manufacture hardware that will operate in space. Most semiconductor coverage focuses on the terrestrial competition between TSMC, Samsung, and Intel. Terafab is not competing in that market. It is creating an entirely new one. Now to the physical reality of what is being built. Giga Texas. Tesla's existing factory in Austin is already one of the largest buildings on Earth by floor space. It covers approximately 10 million square feet and is so vast that employees use electric scooters to move between sections. Terafab is planned at over 100 million square feet. 10 times the size of Giga Texas. Musk described it as the largest and most valuable building on Earth by far. If it reaches planned scale, it will be the largest enclosed structure ever built by a private company in human history. Morgan Stanley projected that even under an aggressive construction schedule, initial chip output from Terafab would not begin until mid-2028 at the earliest. Full-scale manufacturing, 2029 to 2030. Those timelines matter because the Optimus production targets, the orbital compute expansion, and the Starship launch cadence all depend on Terafab delivering on schedule. Any delay in chip output delays everything downstream. This is not a factory being built for today's needs. It is being built for a production scale that does not yet exist and may not for 3 to 4 years. That is either extraordinary foresight or extraordinary risk. Possibly both. The D3 chip deserves its own moment because almost nobody covering Terafab has explained what it actually is. Standard semiconductors fail in space. Cosmic radiation flips bits, corrupting memory, and causing computational errors that on the ground would be recoverable and in orbit are catastrophic. Standard cooling systems that rely on moving air or liquid across a heat sink don't function in a vacuum. Chips built for terrestrial data centers are not designed for the thermal cycling that occurs as a satellite moves between direct sunlight and Earth's shadow every 90 minutes. The D3 is engineered from the ground up for these conditions. Radiation-hardened. Designed for radiative cooling, the only heat dissipation mechanism that works in vacuum. Built to operate across temperature swings that would destroy standard commercial chips. And 75% of Terafab's output goes to this chip. SpaceX is not building orbital data centers as a distant experiment. It is building Terafab to manufacture the specific hardware those data centers require at a scale that suggests the orbital compute vision is already an engineering program with a chip factory under construction. Pull all of this together because the timing was not coincidental. SpaceX reported its first earnings as a public company. Five revenue streams generating simultaneously. The $1 trillion target pulled forward. The financial constraint that forced sequential funding has structurally loosened. Days later, Terafab was announced. The hardware constraint, the 2% ceiling on chip supply, now has a $16.8 billion answer breaking ground in Texas. Revenue without chips is a plan that stalls. Chips without revenue to fund them is a factory that doesn't get built. This week, both constraints moved in the same direction in the same week for the first time. If Terafab delivers first chips by mid-2028 and Optimus production scales as Tesla has guided, the convergence point for robots, orbital compute, and autonomous systems arrives in the 2029 to 2031 window. The same window SpaceX's Mars transfer opportunities open. Watch the construction milestones at Grimes County. Watch the first chip tape-out date. Watch whether D3 chips reach SpaceX satellites before 2030. Those three data points will tell you more about whether this plan is real than anything else. In Grimes County, Texas, on land that was empty last week, the largest building on Earth is about to be constructed. Two chips, one for robots on the ground, one for data centers in space. Musk said it himself, "We either build the Terafab or we don't have the chips." They're building the Terafab. This is gated. Subscribe. The next door opens soon. >> Mhm.