01What TerraFab Is and How It Works
TerraFab $25 billion chip factory is Elon Musk's plan to build the largest chip factory in human history. Announced March 21st at the old Seaholm power plant in Austin, Texas, it is a joint venture between Tesla, SpaceX, and xAI. The goal is to produce 1 terawatt of computing power per year—that's 1 trillion watts of AI processing.
To understand the scale, every chip factory on earth combined currently produces only 2% of what TerraFab targets. That means Musk is trying to build something 50 times bigger than everything that exists today. The factory will consolidate every stage of chip production under one roof: design, fabrication, memory production, and packaging.
No company has ever attempted this at this scale.
It will target two nanometer process technology, the most advanced chip process entering commercial production. Think of it like building a restaurant that grows its own food, cooks every dish, packages it, and delivers it—all in one building. That's vertical integration, and Musk has done it before with Tesla's batteries and SpaceX's rockets.
02Main Features and Planned Chip Types
TerraFab will produce two main types of chips. The first is for Tesla's self-driving cars, the CyberCab robo-taxi program, and Optimus humanoid robots. These chips will power autonomous vehicles and humanoid robots.
The second chip type is called D3, built specifically for space. It will go into satellites and orbital data centers. Musk's plan is to send 80% of TerraFab's computing power into space, keeping only 20% on earth.
He believes that within two to three years, running AI in orbit will be cheaper per watt than running it on the ground. The reasons are no heat problems, constant solar power, and no land costs.
- 1 terawatt of AI compute per year target
- Two nanometer process technology
- Vertical integration: design, fabrication, memory, packaging all under one roof
- Two chip types: one for Tesla vehicles and robots, D3 for space
- 80% of compute destined for orbital data centers
SpaceX has already filed with the SEC to launch 1 million data satellites into low earth orbit. Each satellite would carry AI chips. This is a massive infrastructure bet that goes beyond earth-based computing.
03How TerraFab Will Get Started: Steps and Timeline
TerraFab is still in the announcement phase, so setup steps are based on the projected plan shared in the episode. Here is the expected path:
- Secure the site: The factory will be built at the old Seaholm power plant in Austin, Texas.
- Finalize funding: The $20 to $25 billion cost is not yet in Tesla's capital expenditure plan, which already exceeds $20 billion per year. Total spending could exceed $40 billion in a single year.
- Begin construction: Musk gave no firm timeline at the launch event. Morgan Stanley projects initial chip output no earlier than mid-2028, even under aggressive build-out.
- Produce small batch chips: Small batch production is targeted for late 2026.
- Scale to volume production: Volume production is projected for 2027.
- Reach full scale: The 1 terawatt vision is years away, if it happens at all.
Keep in mind that Musk's companies have never fabricated a chip before. They buy chips, not build them. This is a huge leap from their current expertise.
04How TerraFab Would Be Used: Walking Through Examples
If TerraFab works, the chips would flow into several use cases shown in the episode. For Tesla, the first chip type would go into CyberCab robo-taxis, enabling full self-driving without human oversight. The same chip would power Optimus humanoid robots, allowing them to operate autonomously in factories and homes.
The D3 chips for space would be installed on satellites. SpaceX plans to launch 1 million data satellites. These would form an orbital data center network, running AI models in space.
AI could process data in orbit and send results back to earth, bypassing ground-based server farms.
For everyday AI tools like Claude, ChatGPT, and Grok, TerraFab chips would provide more computing power. More chips mean more compute. That translates into faster model training, cheaper access, and more capable AI assistants.
An agency owner could automate client reports with less delay. A freelancer could run multiple AI projects simultaneously. A creator could offload research, scheduling, and outreach to AI that runs on abundant chips.
05Results, Comparisons, and Skepticism in the Episode
The episode shows a big gap between promise and reality. At full scale, TerraFab's production would equal roughly 70% of TSMC's entire global output—from a single facility run by companies that have never made a chip. For context, TSMC has 50 years of experience.
Patrick Moorhead of Moor Insights and Strategy said, 'Saturday was a presentation, not a fab.' This means the announcement was just talk, not a working factory.
On the positive side, Musk's track record includes pulling off things people said were impossible: Tesla making cars at scale, SpaceX landing rockets. But he also overpromised on Battery Day 2020, and those results still haven't arrived. So the question is not whether TerraFab hits every target, but what happens if even 20% works.
- TSMC's global output compared: TerraFab full scale = 70% of TSMC's total
- Morgan Stanley analyst: Initial output not until mid-2028
- Musk's own timeline: Small batch late 2026, volume 2027, full scale years away
- Competitors: NVIDIA, TSMC, Samsung, Intel are also building at this level
The episode makes clear that chip supply is already constrained. Musk warned at Tesla's Q4 2025 earnings call that chip supply from TSMC, Samsung, and Micron would hit a ceiling within three to four years.
06Tips, Limits, and Mistakes to Avoid With TerraFab
The biggest limit is that TerraFab does not yet exist. Musk has never fabricated a chip. Promising to outproduce the world's most advanced chip maker is an enormous bet.
Skeptics have a point. Do not assume the factory will deliver on time or at all. The episode compares it to Battery Day—promises that still haven't arrived.
Another limit: the $20 to $25 billion cost is not in Tesla's capex plan. That means funding is not secured. The company would need to spend over $40 billion in a single year on top of everything else Tesla is doing.
That is a massive financial strain.
Mistake to avoid: waiting for TerraFab to exist before using AI. The episode stresses that AI tools are already powerful enough to 10x your output today. You don't need new chips to start building AI workflows.
The people who wait will fall behind.
Tip: use the current chip shortage as a reason to learn AI now. The window to get ahead is closing. Build AI skills while tools are still accessible.
If compute gets cheaper later, you'll be positioned to scale fast.
07Who TerraFab Is For and Why It Matters
TerraFab matters to anyone who uses AI tools, because chip supply directly affects the cost and speed of those tools. If you run a business, agency, or creative studio, the outcome of this bet will shape your competitive landscape. More chips mean cheaper, faster AI, which lets you automate more tasks, handle more clients, and deliver faster results.
For creators and freelancers, the episode says you don't need TerraFab chips to get started. The tools today are already powerful enough to multiply your output. The smart move is to build AI workflows now, before the next wave of compute makes the gap even bigger between early adopters and latecomers.
For managers and team leads, your job will look fundamentally different in 18 months. The computers and models are already here. Workflows are being built right now by people who are not waiting.
The episode puts it plainly: the question is not whether AI changes things, but whether you are on the right side of that change. TerraFab is a $25 billion bet that AI compute is the most important resource for the next decade.