Terafab: The Most Audacious Bet in Semiconductor History

On 21 March 2026, standing in a decommissioned Austin power plant, Musk unveiled Terafab, a joint venture between Tesla, SpaceX and xAI to build a semiconductor mega-factory unlike anything the world has ever seen.

The goal? 1 terawatt of AI compute capacity per year. That’s more than every chip manufacturer on Earth produces combined, today and projected through 2030.

Here’s what makes this extraordinary:

• Vertically integrated — chip design, fabrication, memory, packaging and testing all under one roof. No more fragmented global supply chains.
• 2-nanometer technology, the most advanced process node in commercial production today.
• $20–25 billion investment, with Intel now officially on board as a manufacturing partner.
• Chips will power Tesla vehicles, Optimus robots, autonomous driving and eventually orbital AI data centres in space.

Musk’s reasoning is blunt: “We either build the Terafab, or we don’t have the chips, and we need the chips, so we build the Terafab.”

Sceptics point to Tesla’s Battery Day promises from 2020 that took years to materialise. Morgan Stanley estimates the true cost could hit $35–45 billion. This is not without risk.

But here’s the bigger strategic picture that businesses should absorb:

The AI compute bottleneck is now the defining constraint of our era. Every company planning AI-powered products, services or operations in the next decade will be affected by who controls semiconductor capacity and at what price.

Whether Terafab succeeds or stumbles, AI infrastructure is becoming as strategic as energy infrastructure.

📰 Sources:

Wikipedia · CBS News · TechCrunch · Engadget · Electrek