We like ‘em because their angle towards news is similar to ours — without the usual bias, concise, without the extra fluff:

Remember when the news was about what happened, not how to feel about it? 1440's Daily Digest is bringing that back. Every morning, they sift through 100+ sources to deliver a concise, unbiased briefing — no pundits, no paywalls, no politics. Just the facts, all in five minutes. For free.

Read the newsletter trusted by 4.5 million fact-seekers.

If you’d like to advertise to thousands of Tesla geeks in our Tesla Space newsletters with a relevant offer, do reach out via [email protected]!

By the way, this deep dive today is jointly written with Simon Alvarez, whom you might recognize as the main writer from Teslarati. It so happens that Simon has now joined me (Jaan) at my EVwire news publication full-time instead. And this means you will start seeing some more of his insights around Tesla Space, too! Anyway, to the story…

Elon Musk has formally announced the Terafab Project, a joint effort from Tesla, SpaceX, and xAI designed to scale AI compute far beyond current global capacity.

In short, it’s about chips. The ones going to the humanoids, cars, and to the AI satellites. Mostly the latter.

If you want to check out the full livestream yourself, you can find it here. (jump to 21st minute, and it lasts another 28 minutes)

At its core, the Terafab is a vertically integrated chip manufacturing facility.

Its first phase will begin with an advanced technology fab in Austin, Texas, that would build lithography masks, produce chips, test them, and iterate designs in-house.

That’s what this building shown on the livestream was about.

Keep in mind that this is not the 1 TW/Year factory — something that a lot of people assumed at first. We’ll come to that building later on.

During the project’s launch, Elon Musk explained the importance of having a chip factory that can iterate on the go:

A key reason for building the Terafab is the growing gap between AI compute demand and global chip supply.

According to Musk, all existing chip fabs combined currently produce only a fraction of what would be required for Tesla’s long-term AI goals.

Tesla is already a massive client of the world’s top chipmakers like TSMC and Samsung, but Musk stated that “there’s a maximum rate at which they’re comfortable expanding, but that rate is much less than we would like.”

The Terafab is being built to produce two primary types of chips: edge inference chips for Tesla’s products, such as Optimus and the Cybercab, and high-performance compute chips for large-scale AI workloads.

Edge chips will power Tesla’s Full Self-Driving systems and Optimus humanoid robots. Musk suggested that humanoid robot production could eventually reach 1 billion to 10 billion units annually, or about 10 to 100 times the current global vehicle output.

At the same time, the Terafab will also develop and produce chips optimized for space environments, where power, radiation, and thermal constraints differ significantly from those of Earth-based systems.

Musk outlined plans for AI satellites, with early “mini” systems operating at around 100 kW and future versions scaling into the megawatt range.

Musk stated that SpaceX has the potential to deliver 10 million tons to orbit in a year through Starship, while Tesla is building up to produce a terawatt of solar panels. The “missing ingredient,” then, would be the compute needed to achieve the terawatt scale.

That’s where the Terafab comes in.

But achieving a terawatt of compute is just a step. During the closing parts of the Terafab launch event, Musk teased an electromagnetic mass driver on the moon that could launch satellites directly into space.

This, Musk said, would be a way to achieve petawatt-scale computing.

Musk framed the Terafab as part of a broader push toward scaling AI, energy, and robotics, which are key components of what he has described as a path to “amazing abundance.”

The project mirrors Tesla’s decision from 2014 to build Gigafactory 1, which was created to secure battery supply and accelerate electric vehicle production at scale.

When Gigafactory 1, now called Giga Nevada, was announced, a dedicated factory for EV batteries didn’t sound so sensible. At the time, Tesla was only producing mid-volume vehicles like the Model S and Model X.

What critics failed to see then was that Tesla was already building for its future. The Model 3 needed Gigafactory 1 to be feasible, and the company believed that it would dominate the EV sector and break into the mass market vehicle segment. It did, and the rest is history.

In a similar way, the Terafab is designed to secure AI chip supply for Tesla, SpaceX, and xAI’s future products and infrastructure.

If Elon Musk’s AI bet pays off like his EV battery bet with Gigafactory 1, then the Terafab could go down as one of the most important investments of Tesla, SpaceX, and xAI combined.

According to Elon, Tesla’s planned Terafab could span around 100 million square feet.

This would make it the largest manufacturing facility ever conceived and far larger than any existing factory or major tech campus today.

The Terafab’s actual size estimate emerged after initial observations suggested that the facility would be a ~2 million square foot building near Gigafactory Texas — which was actually the image on Advanced Technology Fab instead.

Prior to the Terafab launch event, Tesla and SpaceX watcher Joe Tegtmeyer observed that a large piece of land near the Giga Texas main building was being prepared for construction. Speculations then suggested that the area would be the site of the Terafactory.

Sure enough, a render of a large facility in the exact area was shown during Elon Musk’s speech at the Terafab announcement. But in a reply on X, Musk clarified that the building featured in his presentation is only a small advanced technology fab used for chip design iteration.

“No, that’s just the little advanced technology fab… We couldn’t possibly fit the Terafab on the Giga Texas campus,” Musk wrote.

And in a post on X, industry observer Phil Beisel provided a rough framework for the scale of the Terafactory.

Based on Tesla’s stated goal of reaching one terawatt of AI compute, and assuming individual chips operate at around 250 watts, Beisel estimated that a terawatt would require something on the order of four billion chips per year.

Producing that volume using current semiconductor manufacturing approaches would require thousands of acres of land and potentially hundreds of millions of square feet of production space.

Beisel suggested that with significant efficiency improvements, a facility on the order of 100 million square feet would be a reasonable estimate.

Musk responded positively to Beisel’s estimate, writing, “Yeah, 100M sq ft is the right order of magnitude.”

A factory with 100 million square feet of manufacturing space is unprecedented, and it makes all existing factories today, including Tesla’s massive Gigafactories, look small in comparison. 

For context, Gigafactory Texas, already one of the largest buildings in the world, spans about 10 million square feet. The Terafab would be roughly ten times larger than Giga Texas.

To compare, Apple Park, Apple’s headquarters in Cupertino, has roughly 2.8 million square feet of office space. Microsoft’s Redmond campus spans a little over 8 million square feet across dozens of buildings. 

At 100 million square feet, the Terafab would be roughly 35 times larger than Apple’s HQ and about 12 times larger than Microsoft’s Redmond campus.

For a more fun comparison, the Terafab will be about three times the size of the City of London.

The insane size of the Terafab aside, it’s important to remember that the factory is not being designed around today’s manufacturing scale, but around future demand for AI chips. Think Gigafactory 1 before the Tesla Model 3.

The Terafab Project will push the limits of semiconductor manufacturing, but Elon Musk has outlined how Tesla, SpaceX, and xAI plan to achieve the initiative’s extremely ambitious targets.

The discussion began after industry observers pointed out that the Terafab combines mask production, fabrication, testing, and packaging into a single system.

These processes are typically separated across different facilities, and integrating them under one roof is considered highly complex.

One engineer on X described the concept as “unicorn territory,” noting that simply achieving this level of integration would already be a major step beyond current semiconductor practices.

Still, advanced chip manufacturing is one of the most complex engineering fields, with the engineer noting that the Terafab could very well be more complicated than actual rocket science.

Musk acknowledged the Terafab’s challenge in a reply, but suggested that other projects undertaken by his companies may be even more difficult. This is one of the rare Elon’s lengthy replies:

“Given that several companies make advanced chips, but no companies have ever made fully reusable rockets or achieved SpaceX scale, I think Starship is harder, but we shall see.”

Elon’s comments about the strategies to achieve the Terafab’s targets are classic Tesla. Focusing on simplicity is a typical trademark of Tesla, as shown in its development and use of Gigacasts and a software-heavy driver interface. 

Among all of Elon Musk’s companies, Tesla is also the most experienced when it comes to scale of operations, with the company having produced the world’s best-selling vehicle by volume, the Model Y, for the past three years using only four factories across the United States, Germany, and China.

We actually don’t have that much room in this email anymore (it literally is limited how much we can push to you in one email), so let’s speedrun through some other news without sections this time:

I’ll end today with a book recommendation.

This one isn’t affiliated with us in any way, but comes from a real quality source that put it together, Eric Jorgenson. Some of you likely read one of his best-sellers (and one of the most-highlighted books on Kindle), on Naval Ravikant. Well, now Eric took five years of his life and published The Book of Elon.

There’s even a blurb from Mr. Beast. Oh, and he got the approval from Elon on the idea to launch this.

The fun part is that if you act fast, you can get the book for free. That’s because David Senra, the founder of one of the top podcasts out there called Founders and something I religiously listen to, just made an episode with Eric (this video, I highly recommend it too), and is giving away 1,000 copies of this book. I

f you fit within that thousand (still live as of now), you’ll only pay shipping.

So I can vouch, for once, that this website isn’t a scam and can actually get you the Book of Elon, physical copy, for free: elonbookgiveaway.com. I ordered one.

Or get it right from the founder for the regular price here.

Let me know if you got it, and later if you liked it. Want to see the impact of this here.

PS! If you enjoy our work, you should join us as a Tesla Space Insider.

You’ll get Inside knowledge, Insider Alerts when something breaking happens, and a bunch of other perks.

Upgrade to Tesla Space Insider