A single post on X on Saturday morning — “Terafab Project launches in 7 days” — was enough to send the semiconductor world scrambling for context. Tesla CEO Elon Musk confirmed the company’s Terafab project to make artificial intelligence chips will launch in seven days. The announcement follows months of construction activity at Tesla’s Gigafactory Texas campus and a January earnings call on which the chip supply problem was laid out in unusual candour.
Tesla’s Terafab is a $25 billion domestically built semiconductor fabrication plant designed to produce 100 to 200 billion custom AI and memory chips per year.
The motivation is not abstract. Tesla’s Full Self-Driving software, its Cybercab robotaxi programme, and its Optimus humanoid robot line all run on purpose-built AI silicon. Musk’s projections for Optimus alone — which envisions millions of units annually — require chip volumes that no external supplier has committed to delivering on Tesla’s timeline.
Musk has been consistent on why third-party suppliers won’t cut it. On the January 2026 earnings call he said: “In order to remove the probable constraint… we are going to have to build a Tesla TeraFab.” Separately he elaborated: “It’s like giga but way bigger. I can’t see any other way to get to the volume of chips that we’re looking for. So I think we’re probably going to have to build a gigantic chip fab. It’s got to be done.”
A note of precision is warranted about what “launch” means here. Semiconductor fabs of this scale take years to construct and qualify. What the announcement more likely signals is one of the following: a formal project reveal with location and timeline details, a groundbreaking ceremony, a public disclosure of facility specifications, or the start of construction on the first phase.
Tesla’s fifth-generation AI chip, AI5, is expected in small-batch production in 2026 with volume production following in 2027 — suggesting that Terafab’s first practical output is still more than a year away from mattering to Tesla’s supply chain.
The broader context is a strategic shift that mirrors what Amazon did with AWS infrastructure and what Apple did with its silicon transition: removing dependency on a third party at the moment that third party becomes the primary constraint on the company’s most important products.
If Terafab delivers even a fraction of its stated ambition, Tesla transforms from a chip customer into a chip producer — with consequences for its cost structure in autonomous vehicles, robotics, and whatever xAI requires next. March 21 will provide the first real details.