A century ago, Ford built the River Rouge complex because its supply chain had become the bottleneck. Raw materials came in. Cars came out. Terafab is Elon Musk’s attempt to build that model for AI chips. Tesla is building a research fab at Giga Texas: make a chip, test it, change the design, repeat. SpaceX will build the production facility in Grimes County. Texas says phase one alone represents more than $16.8B of investment and 3,000 jobs. Terafab’s unusual feature is how much it wants to put in one place. Tesla job descriptions list logic, memory, advanced packaging, test and lithography-mask production inside the same facility. Those are normally separate industries, separate suppliers and separate queues. That matters because every new Tesla vehicle, Optimus robot, Starlink satellite and xAI cluster is becoming a customer for specialized silicon. Buying chips from a foundry works when designs change every few years. AI hardware changes with models, workloads and power constraints. A company that wants to redesign chips continuously needs a faster learning loop than a normal supply chain provides. “Over one terawatt of compute per year” is a production target: Annual chip output capacity, not the factory’s electricity bill. Ford used vertical integration to make cars cheaper. Terafab uses it to make AI hardware iterate faster.