Why Elon Musk Is Building His Own AI Chip Factory | TeraFab Explained

Kind: captions Language: en [music] >> This aerial view shows the footprint of a Texas [music] construction site designed to eclipse the Pentagon and Apple Park [music] combined. It is the future home of TerraFab, a semiconductor mega project [music] with a budget projected to reach up to 119 billion dollars. The massive facility [music] has a single clear objective to produce 1 terawatt of computing power every single year. That massive capacity is driven by internal demand. [music] In June 2026, Elon Musk merged his artificial intelligence laboratory xAI [music] with SpaceX. The resulting 1.7 trillion-dollar public entity requires an endless [music] supply of computing power to train advanced machine learning systems like the [music] Grok 4.6 model. Add to that the terrestrial demands of Tesla. The company requires advanced silicon to operate its full self-driving [music] software, its cybercab robotaxi program, and the millions of Optimus humanoid [music] robots it plans to manufacture. Combined, the Musk ecosystem devours silicon. Operating these aerospace, automotive, and AI projects requires [music] a volume of hardware that the global supply chain simply cannot produce. >> [music] >> The modern tech industry operates on a standard division of labor. Massive companies design their own chips in-house, but they outsource the physical manufacturing to dedicated foundries in Asia. Those legacy foundries like TSMC are currently operating at maximum capacity. >> [music] >> Building new nodes takes years and hundreds of billions of dollars, meaning these suppliers cannot expand fast [music] enough to meet sudden massive spikes in demand. This animated chart illustrates [music] the resulting supply deficit. The massive column represents Musk's total projected [music] silicon demand, while the tiny sliver at the bottom represents the 2% he estimates third-party foundries on Earth can actually supply to his companies. There is also [music] a geographic vulnerability. Relying entirely on Taiwan for physical hardware [music] creates an immense geopolitical risk for a multi-trillion-dollar aerospace [music] and automotive empire. If supply lines are severed, production stops. To secure the hardware needed for a spacefaring civilization and widespread robotics, Musk has to forge [music] the silicon himself. This flowchart illustrates Terafabs closed-loop [music] concept. Instead of shipping components across the globe, every stage of production, chip design, lithography, fabrication, memory [music] integration, advanced packaging, and testing will be consolidated under one [music] roof in Texas. The factory splits its focus into two distinct environments. The terrestrial [music] track will manufacture the AI5 and AI6 chips, which are heavily optimized [music] edge inference processors designed to run the neural networks inside Tesla cars and Optimus robots. The orbital track [music] is dedicated to the proprietary D3 chip. These processors are being built specifically to [music] power SpaceX's space-based AI data centers and satellite constellations. Operating silicon in the vacuum of space [music] requires specific physical engineering. The D3 chips must undergo radiation hardening to withstand high-energy cosmic rays and heavy ions, alongside high-temperature optimization [music] to survive in an environment with no air for cooling. Terafab operates as a dual environment forge, building the logic brains [music] tailored for both Earth and orbit. Designing a factory on a whiteboard is easy. Actually fabricating a sub-2-nanometer [music] chip is arguably the most complex manufacturing challenge on Earth. And Tesla has zero experience [music] running a commercial semiconductor foundry. In April 2026, the project gained the structural [music] support it needed. Intel officially joined the Terafab venture as a manufacturing partner. The alliance divides the labor based on specialized strengths. [music] Tesla and SpaceX provide the capital, the construction scale, and the guaranteed internal demand. Intel provides the process technology, [music] the equipment expertise, and the packaging capabilities. For Intel, this full-scale Grimes County plant serves as a massive scaling opportunity. [music] It guarantees an anchor customer for their bleeding-edge 14A, or 1.4 nanometer, process node right as it matures for commercial production. [music] By sharing the risk and the output, the companies turn a highly speculative ambition into a co-invested industrial powerhouse. Musk has a documented history of announcing timelines he fails to meet. Tesla's in-house Dojo supercomputers suffered years of delays and project resets, and the highly publicized million-mile battery never materialized at scale. >> [music] >> Terafab faces immense logistical hurdles before a single chip is printed. The facility requires an excess of $100 billion in funding >> [music] >> and massive amounts of natural gas energy just to power the manufacturing equipment. If [music] the facility reaches volume production, Tesla and SpaceX AI become fully self-sufficient. They decouple entirely from the Asian semiconductor bottleneck, [music] securing their own supply lines regardless of global trade disruptions. For Intel, delivering working silicon at this [music] scale validates its IDM 2.0 strategy, proving it can successfully operate as a foundry for the world's most demanding external customers. The artificial intelligence industry requires [music] physical infrastructure to process data. Software models are ultimately useless [music] without the massive server farms and edge processors required to run the math. Controlling autonomous vehicles, advanced robotics, and orbital infrastructure requires highly specialized [music] hardware. Designing the best silicon is only half the equation. You have [music] to own the factory that prints it.