Terafab 21 March 2026

[MKWVentures]

Your theories might be correct, but they don't make sense in the context of what Musk has been talking about. He's been calling it a "terafab" because he thinks his volumes will be very high. He must be thinking hundreds of millions of chips to be in high volume territory. In fact, his rhetoric reminds me of Sam Altman's from a couple of years ago. If you're very high volume, as a terafab would be if it earned the name, cost is a big factor, because even a small cost adder on hundreds of millions of chips multiplies out quickly.

Co-location makes great sense if you're a control freak, which we all know Musk is, but now he has to ramp on multiple disparate production technologies simultaneously, so it makes no practical sense.

Basically, you're saying Musk is akin to an old-fashioned carnival snake-oil barker, just to get attention. I'm having trouble believing that.

You may be right... Lets see what he actually does vs his hype. How big a factory does he plan? What technologies does he actually plan to ramp.

And yes he is a carnival barker imho. No one is going to mars and the terafab, if it is built, will be smaller than tsmc fabs. just ask him for specifics and the actual plan will come out

 

[user nl]

For now, the project remains aspirational. As Moorhead put it, “I’ll start believing when I see confirmed… equipment orders, a named process technology partner, and a head of semiconductor manufacturing.”



https://www.datacenterknowledge.com...cture-shift-build-the-chips-control-the-stack

Musk’s ‘Terafab’ Proposal Sparks Debate on the Future of AI Infrastructure​

The ambitious Terafab effort targets AI’s next constraints – chip supply, power limits, and scaling – by bringing silicon production in-house.

Shane Snider,Senior News Writer,Data Center Knowledge
March 23, 2026

 

[Tiana Rahaga]

Running 2nm fab (and beyond) is not similar to making Tesla autonomous vehicle (not yet by the way) or Space X falcon or Starlink.

For now there are TSMC and ASML hegemony. For TSMC it is a kind of insurance against China invading

Even China is struggling with the 2nm challenge, despite it being a vital priority for them. Better to invade Taiwan and own everything and put thread on the others.

Intel is faltering, despite Lip Bu-Tan communication and Jason Huang photo

Samsung seems to be succeeding but suffers from a critically low yield. Elon still communicate the possible engagement with Samsung

Beyond lithography, talking about leading HBM, Chiplet and interconnec, 3D packaging...

Elon has money, sure, is the solution is to buy ASML, Samsung or TSMC and Memory manufacturer?

Time will tell how it will turn out. A real genius or an above-ground.

 

[Daniel Nenni]

rafab’ Proposal Sparks Debate on the Future of AI Infrastructure
The ambitious Terafab effort targets AI’s next constraints – chip supply, power limits, and scaling – by bringing silicon production in-house.

Shane Snider,Senior News Writer,Data Center Knowledge
March 23, 2026

It still looks like a head fake to get a better wafer agreement, better than his AI competitors.

He can pre pay TSMC to build fabs, like Apple did, and get a special process that others cannot use which would have a 99.99% success rate.

 

[soAsian]

Tesla car sales have been on the decline year over year. BYD is eating Tesla's lunch everywhere outside of US. This Terafab felt like a desperate play for Elon to pump Tesla's stock. Not to mention in the US, people who buy EV car are mostly liberal/democrats and they all hate Musk and Trump.

Trump is all about drill baby drill. Ev credits canceled by Trump as well.

 

[essiah]

As some of you know I grew up an Intel fan early on in my Semiconductor life. My first PC was an IBM PC XT and my fandom grew from there. I actually got to meet Andy Grove when he was CEO of Intel at a semiconductor conference. I had seen Gordon Moore speak a couple of times in 1984/5 but he did not write books like Andy Grove. I read all three of Andy's books and they full on inspired me throughout my career. The man could communicate.

Intel is one of the reason's why my writing and SemiWiki became famous. BK took over as CEO in in 2013, 2 years after SemiWiki was founded and you can say I was not fond of BK from the very start. I pushed for Intel to buy Nvidia and make Jensen Haung CEO. Imagine that, Jensen CEO of Intel in 2013. That would have been world changing, for better or worse we will never know.

I gave BK the benefit of my doubt at first but his dishonesty was impossible to support. He lied to my face about 14nm yield and that was it for me. He was an intelligent man and he knew where the Intel bodies were buried. Some say he was manipulated by the Intel Board of directors ( Frank Yearny joined the Intel board in 2009).

The result was Intel employees hated me. I don't blame them. Even today I get comments on how hard I was on Intel back then. So I guess I should thank BK but I won't. He was the worst Intel CEO in the history of Intel.

As ex intel employee I fully support your efforts. There is still a lot of persistent rot within intel management

 

[Barnsley]

[News] Tesla Reportedly Ramps TeraFab Hiring in Taiwan; TSMC Talent Seen as Primary Target

Please note that this article cites information from Economic Daily News, Tesla on LinkedIn, and UDN News. Elon Musk’s TeraFab chip plant, reportedly...

www.trendforce.com

Tesla Reportedly Ramps TeraFab Hiring in Taiwan; TSMC Talent Seen as Primary Target

Nice , hope they get good signing on fees in case it amount to nothing.

 

[Xebec]

Tesla car sales have been on the decline year over year. BYD is eating Tesla's lunch everywhere outside of US.

To be fair, the Chinese EV makers are eating every automakers lunch where they're allowed to compete.

 

[Xebec]

Your theories might be correct, but they don't make sense in the context of what Musk has been talking about. He's been calling it a "terafab" because he thinks his volumes will be very high. He must be thinking hundreds of millions of chips to be in high volume territory. In fact, his rhetoric reminds me of Sam Altman's from a couple of years ago. If you're very high volume, as a terafab would be if it earned the name, cost is a big factor, because even a small cost adder on hundreds of millions of chips multiplies out quickly.

Elon is thinking or saying - billions of chips per year:

"On Saturday, Musk also estimated that Tesla will eventually produce 10 to 100 times as many Optimus robots as electric vehicles, potentially producing as many as 10 billion robots a year."

What is Elon Musk's Terafab chip project? Here are his "most epic" goals for the factory.

Elon Musk unveiled plans for a massive chip manufacturing facility in Austin, Texas, that he said will power his "galactic" aspirations.

www.cbsnews.com

..

And yes, he is starting to sound like Altman, minus the fake vocal fry.

 

[MKWVentures]

Elon is thinking or saying - billions of chips per year:

"On Saturday, Musk also estimated that Tesla will eventually produce 10 to 100 times as many Optimus robots as electric vehicles, potentially producing as many as 10 billion robots a year."



..

And yes, he is starting to sound like Altman, minus the fake vocal fry.h

He should just say "we will have a JILLION robots ..... and they we will colonize Mars and have hyper loops from Austin to Chicago"

being aspirational is OK as long as no one spends money based on your wild aspirations.

 

[Xebec]

being aspirational is OK as long as no one spends money based on your wild aspirations.

Off topic from this forum - but I'll disagree here partially; a lot of things are achieved with aspiration. Aim for the moon -- and if you end up in Earth's orbit, you still achieve something. Aspiration (properly applied) can motivate people to try or do things they shouldn't. Experienced Angel investors make decisions based on aspirations -- as they learn that 99% of the financial projections put in front of them are not true.

I do think Elon is falling to the 'previous success = future success' fallacy, but at least he is trying something new. I would have thought opening a brand new bleeding edge fab not named "TSMC, Samsung, or Intel" was completely impossible - but Rapidus has a chance at disproving this already. And that's with layers of government red tape undoubtedly slowing the process vs. a purely commercial effort.

 

[MKWVentures]

Off topic from this forum - but I'll disagree here partially; a lot of things are achieved with aspiration. Aim for the moon -- and if you end up in Earth's orbit, you still achieve something. Aspiration (properly applied) can motivate people to try or do things they shouldn't. Experienced Angel investors make decisions based on aspirations -- as they learn that 99% of the financial projections put in front of them are not true.

I do think Elon is falling to the 'previous success = future success' fallacy, but at least he is trying something new. I would have thought opening a brand new bleeding edge fab not named "TSMC, Samsung, or Intel" was completely impossible - but Rapidus has a chance at disproving this already. And that's with layers of government red tape undoubtedly slowing the process vs. a purely commercial effort.

I have been hired by a couple VC/Angel companies to investigate ("is this real, what do the number say").... So they require numbers to back aspirations. and they have a goal of <90% of start ups fold. So Musk is held to a higher standard.

Rapidus is not close to being a successful leading edge foundry yet. but we will see how the mass production in 2027 goes (have they published the data and process details from their pilot line from last year?)

But back to the thread: TSMC and even samsung were not willing to spend money on the aspiration, so Musk said he will do it himself. Lets see how that goes or if its just leverage

 

[Jim351]

I have been hired by a couple VC/Angel companies to investigate ("is this real, what do the number say").... So they require numbers to back aspirations. and they have a goal of <90% of start ups fold. So Musk is held to a higher standard.

Rapidus is not close to being a successful leading edge foundry yet. but we will see how the mass production in 2027 goes (have they published the data and process details from their pilot line from last year?)

But back to the thread: TSMC and even samsung were not willing to spend money on the aspiration, so Musk said he will do it himself. Lets see how that goes or if its just leverage

At least Rapidus has IBM and IMEC in their court for process and design technology - IBM still maintains the Albany Nanotech facility as far as I know. The idea of having an ecosystem partner or partners for the process and chip design is just one of two major hurdles.. the other is related to the scale Elon is imaging and answering the question of where all the tools and chemicals would come from to feed a fab of this scale.

I tried arguing with the Elon fanboi numbnuts on X but the vast majority of them are speculators who have no shared sensibilities with actual semi industry veterans. They read the announce by Lace on their He atom beam litho idea and think it's a solved problem. A lot of them think Elon can conjure up some kind of alternative litho platform on his own... as if nobody smart ever tried that before (paging David Lam, and the bones of Mapper which ASML danced on). Even if Elon were to be a first mover with a new light source like Xlight for his EUV tools, with ASML actually agreeing to sell tools stripped of the Tin X-ray sources, there are so many other supply chain bottle necks. No entity is more motivated than China to find a means to manage leading edge litho without EUV... but they haven't figured it out yet either. And then there is the lack of high purity H2SO4..... and on, and on.