One City. Three Companies. A $20 Billion Bet That Could Reshape the Tech World
Image Credit: Leonardo AI
News Summary
- Elon Musk officially launched TERAFAB on March 22, 2026, at Austin's Seaholm Power Plant, a joint venture between Tesla, SpaceX, and xAI.
- The project carries a $20–25 billion price tag and targets 1 terawatt of AI computing power per year, roughly 100–200 billion chips annually.
- TERAFAB will consolidate chip design, fabrication, memory, packaging, and testing all under one roof, a first in semiconductor history.
- Two chip types planned: one for Tesla vehicles and Optimus robots, and a high-power variant for SpaceX's orbital AI satellites.
- Tesla has already started hiring for TERAFAB, posting roles in semiconductor infrastructure just days after the announcement.
Elon Musk just stood inside a defunct power plant in Austin, Texas, laser beams shooting into the night sky, and announced something that even his biggest fans called the wildest bet in American tech history. A chip factory so large, so ambitious, it would need to outpace the entire global semiconductor industry just to meet its own companies' needs. The question isn't whether you find Musk annoying or brilliant. The question is: what if he actually pulls this off?
Table of Contents
What Is TERAFAB, Exactly?
Most chip companies do one or two things well. They either design chips or they manufacture them. Very few do both. Nobody does all of it: design, lithography, fabrication, memory, packaging, and testing under a single roof.
That's exactly what TERAFAB aims to be. Musk described it from the Seaholm Power Plant stage as the most epic chip-building exercise in history by far. According to KUT Public Radio, his goal is to produce more than 1 terawatt, that's 1 trillion watts, of AI computing power per year.
To understand the scale: Tom's Hardware reports the current global output of AI compute sits at roughly 20 gigawatts per year. Musk wants to build something that dwarfs what every chip factory on Earth produces today. Combined. In a single facility.
The facility targets 2-nanometer process technology, the most advanced chip node entering commercial production right now. TSMC has spent decades and hundreds of billions of dollars reaching this point. Musk wants to get there too, and fast. Whether that reads as ambition or delusion depends on which expert you ask.
Three Companies, One Massive Goal
This isn't just a Tesla project. It's not just a SpaceX project. TERAFAB is a three-way joint venture between Tesla, SpaceX, and xAI, and that combination matters more than most headlines are letting on.
xAI, Musk's AI startup, merged with SpaceX in an all-stock deal in February 2026, a move many analysts saw coming long before it happened. Now all three companies sit under the same strategic umbrella, and TERAFAB is the physical infrastructure that ties them together.
Investing.com notes that xAI is expected to consume the vast majority of the chips TERAFAB produces, since its AI workloads, particularly for the Grok AI model, demand enormous compute. Tesla confirmed a $2 billion investment in xAI and has already integrated Grok into its vehicle lineup.
"We either build the TeraFab, or we don't have the chips, and we need the chips, so we build the TeraFab." — Elon Musk, March 22, 2026, Seaholm Power Plant, Austin
That logic is blunt. Some would call it circular. But it reflects a very real supply problem: current chip suppliers TSMC, Samsung, and Micron simply aren't scaling fast enough to meet Musk's projected demand. Tesla said on X that TERAFAB chips would require "more than all chip manufacturers in the world combined can provide today."
The scale of AI infrastructure investment in 2026 has been staggering across the industry. Musk's companies aren't the only ones scrambling for silicon. They're just the ones deciding to build their own foundry about it.
Two Chips, Two Worlds: Earth and Space
TERAFAB doesn't plan to make one chip. It plans to make two very different ones, and that distinction is where things get genuinely interesting.
Chip One: The Ground Chip
The first is an inference chip built for edge computing, Tesla's Full Self-Driving system, robotaxi fleets, and the Optimus humanoid robot. Teslarati reports Tesla's upcoming AI5 chip will deliver 40x–50x more compute performance than AI4, with 9x more memory. Small-batch production is expected in 2026, with volume production targeting mid-2027.
It's worth noting that Tesla has already signed a deal with Samsung to manufacture the AI6 chip in U.S.-based fabs. TERAFAB represents what comes after that point at which Tesla no longer relies on any outside manufacturer at all. As Not a Tesla App explains, TERAFAB is the endgame; it would allow Tesla to iterate on future chip designs entirely at its own pace.
Tesla currently relies on Samsung to manufacture its AI4 chips. AI5 will be made jointly by TSMC and Samsung. TERAFAB is the long-term plan to cut both out entirely.
Chip Two: The Space Chip
The second chip type is far more unusual. Engadget reports it's designed specifically for orbital AI satellites, a network of space-based data centers powered by solar energy.
Musk made a striking claim: 80% of TERAFAB's compute output is aimed at space applications. He argues that solar irradiance in space is roughly 5x greater than on Earth's surface, and heat dissipation in a vacuum makes running massive AI workloads in orbit thermally viable.
He's already filed an application with the Federal Communications Commission to launch a million satellites to create what he calls an orbital data center. The FCC filing is real. Whether a million satellites follow is another matter entirely.
Musk also placed TERAFAB inside a far grander vision. At the event, he said: "We want to be a civilization that expands to the galaxy... with a city on the moon, cities on Mars, populate the solar system." This ambition connects directly to the broader AI chip arms race playing out across the tech industry, where every major player is trying to control its own silicon destiny.
Why Austin? The Texas Advantage
The location is no accident. KXAN Austin reports that TERAFAB will be built on the North Campus of Giga Texas, the Tesla manufacturing campus already based in eastern Travis County.
In a Sunday post on X after the launch event, Musk clarified an important detail: the Austin facility will focus specifically on chip design. The main TERAFAB fabrication facility would require thousands of acres, and multiple locations across the U.S. are still being evaluated. Austin is the starting point, not the whole story.
Texas brings three advantages that matter at this scale: large available land, business-friendly tax policy, and a fast-growing tech workforce. Drone footage captured months before the announcement already showed massive site preparation north of the existing Giga Texas factory. Texas Governor Greg Abbott was in the audience that night. He later posted: "Thanks, Elon. Extraordinary event tonight. Your vision is powerful, and we are proud of all you do in Texas."
Austin is quietly becoming one of America's most important tech cities. TERAFAB even partially makes that status official.
TERAFAB and America's Chip War
There's a bigger picture here that goes beyond Musk and his companies. The U.S. has been in an active push to reduce dependence on foreign semiconductor manufacturing, particularly from Taiwan, which produces over 90% of the world's most advanced chips.
The CHIPS and Science Act, signed in 2022, allocated over $52 billion to rebuild domestic semiconductor manufacturing. Intel, TSMC, and Samsung have all announced or begun U.S. fabs with government support. TERAFAB now adds a private-sector heavyweight to that effort with no government subsidies announced, at least not yet.
NVIDIA's own strategic pivots in 2026 signal just how fast the semiconductor landscape is shifting. Every major tech company now understands that whoever controls chip production controls the future of AI, and right now, that future still runs primarily through Taiwan.
TERAFAB, if built at full capacity, could scale to roughly 70% of TSMC's current global output, according to Teslarati's analysis. That number deserves a moment of quiet reflection. And maybe a glass of water.
The Scale Problem Nobody Is Talking About
Here's where we need to be honest with you. The announcement is real. The ambition is real. But the engineering challenge is almost incomprehensibly large.
Electrek points out that a single 2nm fabrication plant costs roughly $28 billion and takes about 38 months to build under optimal U.S. conditions. TSMC spent $165 billion over decades building its six-fab Arizona operation. Geographic precision matters too; advanced lithography is so sensitive that vibrations from a nearby highway can ruin an entire production run.
Tesla has zero semiconductor manufacturing experience. That's not a criticism, it's a fact. And Musk gave no construction timeline at the event. Bloomberg noted he "has a history of over-promising on goals and timelines."
Then there's the money. EVXL reports that Tesla's CFO confirmed the $20–25B TERAFAB cost is not yet included in Tesla's 2026 capex plan, which already exceeds $20 billion on its own.
Tesla is planning a $20B+ project that isn't in its budget yet, in an industry it has no experience in, with no announced completion date. You don't need an MBA to find that part interesting.
Some analysts also believe TERAFAB is partly tied to SpaceX's upcoming IPO, a pre-listing narrative play as much as a genuine engineering commitment. The timing, they argue, is too convenient to ignore.
It's Already Moving, Tesla Is Hiring Now
Here's the detail that separates this announcement from a vague press event: Tesla started hiring for TERAFAB within days.
Not a Tesla App reports that a job listing for a Technical Program Manager in Infrastructure Semiconductors, based in Austin, appeared almost immediately after the launch event. That's a very specific hire, not a generalist role, but exactly the kind of specialist you'd recruit to manage the early construction and planning phase of a semiconductor facility.
For a decade, Tesla has been a leader in vertical integration, making its own batteries, its own software, its own motors. But chips remained one of the few critical components it had to outsource. As Tesla pivots from carmaker to AI and robotics company, that dependency has become a serious liability.
The revenue model powering Musk's expanding empire depends increasingly on AI compute, and the companies supplying that compute are not moving fast enough for his plans. TERAFAB is the logical conclusion of that frustration.
What the Skeptics Are Saying
Musk has a complicated relationship with his own timelines. He promised a $40,000 Cybertruck, which it launched at nearly double the price. He said Tesla would have fully autonomous driving by 2019. His 4680 battery program, announced with fanfare in 2020, has significantly underdelivered five years later.
Critics on Electrek draw a direct parallel: this is Tesla's Battery Day on steroids. And they're not wrong to ask the question. The broader reset happening across big tech in 2026 means investor patience for moonshot timelines is thinner than it used to be.
On the other hand, reusable rockets were also called impossible. And SpaceX is now the most flight-proven rocket company on Earth. Musk does eventually deliver the timeline just tends to stretch, sometimes by years.
The chip supply argument isn't economically absurd either. If Tesla, SpaceX, and xAI's combined AI compute needs grow at the rates Musk projects, owning a fab starts to make real strategic sense, regardless of how hard it is to build one. The geopolitical stakes of AI compute infrastructure are no longer abstract. Governments and companies worldwide now understand that computing is power, quite literally.
What Happens Next
In the near term, Tesla, SpaceX, and xAI will continue purchasing chips from TSMC, Samsung, and Micron. Musk made that clear. TERAFAB is a long-term play, not a tomorrow solution.
The Austin North Campus focused on chip design, is expected to begin work by the end of 2026. EVXL estimates a functioning 2nm fab producing chips in volume is realistically a 2030–2031 milestone at the earliest. The main fabrication site, requiring thousands of acres, remains unannounced.
SpaceX's IPO is expected later in 2026; it will be critical. It's the most likely funding vehicle for the orbital AI satellite network that TERAFAB's space chips are designed to power. The pace at which AI is embedding itself into every industry only increases the urgency of what Musk is trying to build.
Meanwhile, the rest of the tech world is watching. If TERAFAB succeeds even partially, it signals something profound: the era of depending on foreign chip supply chains for America's most critical AI infrastructure may have a real expiration date.
And if it doesn't? Well, at least Austin got a spectacular light show out of it.