Elon Musk's Terafab semiconductor project could cost $5 trillion
Posted by sr_local@reddit | hardware | View on Reddit | 188 comments
Posted by sr_local@reddit | hardware | View on Reddit | 188 comments
Dev_Im@reddit
You know what's funnier? All those Tesla suckers that are praising this project will cry when they announce a share dilution. It is what happens when you buy stocks driven by a cult instead of a logical analysis.
muntaxitome@reddit
I'm really curious on what his plan is to get his hands on enough lithography capacity on a reasonable timeframe. The claim seems to be that they will develop their own but that is easier said than done. Money isn't the only constraint.
Chipay@reddit
Look, Elon Musk is the man that brought fully self driving cars in 2017 and the first Mars colony in 2019. If there's any person in the world that sticks to deadlines, it's him.
bad1o8o@reddit
let's not forget the hyperloop and a fleet of self-driving trucks or swapping out the EV battery within seconds instead of waiting for it to charge
vnmslsrbms@reddit
Vegas without the Vegas loop would be in such bad shape right now. What would the world be like without Elon. Have you ever thought about that? I shudder even trying to start
ours@reddit
At least we got regular rocket flights from NY to Tokyo.
benjamin_noah@reddit
And I can’t imagine what COVID would’ve been like without all those Tesla ventilators.
ElementII5@reddit
Didn't he also get those kids out from that cave?
Marcoscb@reddit
While exposing a pedophile at the same time!
an_angry_dervish_01@reddit
That was just fucking strange.
vr00mfondel@reddit
That was the point where any sane person would've dropped anything with his name attached to it.
But that is not the timeline we are in...
EconoMePlease@reddit
People aren’t betting on him to succeed in anything other than duping investors into pumping up his stocks. He’s useful to people as long as he is making them more money than they spend propping him up.
an_angry_dervish_01@reddit
Yet it made zero difference, we live in an odd world.
ours@reddit
Oof, that one aged spectacularly with the Epstein files.
splendiferous-finch_@reddit
And let's not forget his honourable civic service is cutting trillions in waste from government spending.
Roxalon_Prime@reddit
And the most damning was the whole 150 years old people getting money from the government. One would have thought that he is tech savvy/ strong in sciences enough to at least suspect that it is a programming language related issue, but no
splendiferous-finch_@reddit
He did the right thing we all know US Aid was evil when they started investigating him for possible fraud.
Who the hell would dare to even think Elon could be commit fraud
TemuPacemaker@reddit
Good thing there were 0 new cases by April 2020.
st8ic88@reddit
Don't forget how awesome all of the Tesla Roadsters have been, for all those people who put down $50k deposits eight years ago.
Master_Put_6283@reddit
hyperloop to be fair to the ass of a man whos only words coming out of said ass are lies is that he wanted to fuck over public trains and metros and he did get dems to pay for the loop idea
Rx-Nikolaus@reddit
This. It was protecting his business interests and also helping to ensure that we can't have nice things.
im_just_thinking@reddit
He singlehandedly eradicated all of fraud, inefficiencies, and woke mind virus in the whole country
bad1o8o@reddit
and only thanks to him comedy is alive and well
frudi@reddit
And all the alien dreadnought giga factories where robots move so fast you need strobe lights to even see them.
tumble00weed@reddit
Elon is a businessman. He, himself, hasn't really brought us much, but he is brilliant no doubt.
EmployeeCautious6314@reddit
Yeah man, I wouldn't give up my Tesla Roadster for anything!
an_angry_dervish_01@reddit
Hahah he has a lot in common with Sam Altman, this made me laugh.
blaktronium@reddit
His only constraint is the english language, which limits his ability to tell ever grander lies
orbvsterrvs@reddit
That's why he and the other billionaire bros speak gibberish! :D
tumble00weed@reddit
They call it something. Thrump does it too. Circle talk? (Circle jerk)
banditcleaner2@reddit
Space data centers!!!!!
StaysAwakeAllWeek@reddit
It's the fact that he occasionally follows through (eg. Starlink) that keeps people listening
U3011@reddit
Which you can witness by another person who tells very grand lies but has the vocabulary of a youth.
banditcleaner2@reddit
I imagine if you had $10 trillion dollars it would probably increase speed of getting lithography machines, but still not by much
SpaceyMcSpaceGuy@reddit
Yeah this is a really long tail project - like 10-20 years if they don't run out of money.
The point of building the TeraFab is that the world's total capacity for chip-building is under what Tesla/SpaceX project for their own demand. World capacity is generally limited by ASML's production, so Tesla using ASML doesn't solve the problem. If they want that many chips they have to build their own lithography.
Special-Call494@reddit
The spaceX/Tesla demand is based on completely unrealistic growth in a bunch of their services. You might see 10% as the actual demand at best.
The other issue is why use a company completely new to the industry to invest in to try and make Fab machines. Then also trying to make the chips themselves is just not going to happen. It may have been possible in the 2000s but not chance at it today.
SpaceyMcSpaceGuy@reddit
Agree that this only makes sense if inference compute explodes in demand. I think the answer to that is "yes," but not necessarily through Tesla's internal product lines. They can sell the chips externally as long as they are good chips.
The way I think about this is that building their own fab is an asymmetric play - worst case they blow $50B-100B building a pilot plant, realize it's way harder than they thought, and give up. Best case they build a new vertical that itself is worth > $1T. It is more likely they fail, but the mag 7 do this kind of stuff all the time. Biggest fail I can think of is the MetaVerse. Biggest win I can think of is AWS. Tesla's strength is in manufacturing, so they are inclined to take on asymmetric manufacturing challenges.
cdoublejj@reddit
if it's for semiconductors maybe they don't need cutting edge lithography? maybe an older much bigger NM would work?
wintrmt3@reddit
If the plan is to make AI hardware it won't, because the bottleneck is electricity and smaller nodes have a significant advantage there.
CheesyCaption@reddit
Google makes their own AI hardware and it's more efficient than Nvidia cards because they are less general purpose devices. Turning your software into hardware improves efficiency more than die shrinks.
arg_max@reddit
Google designs their own AI hardware that is being manufactured by tsmc. Other AI labs are working on similar chip designs and if X.ai would want that I wouldn't call it unrealistic that they'd get there in a few years. Manufacturing is a completely different beast though
CheesyCaption@reddit
That's not the argument I was disputing. The idea that whatever x can come up with on a chip being inherently higher power consumption than current gen, general purpose chips running an AI workload is apples v. oranges. I've got a USB stick (maximum available power 5W) that runs image recognition faster and better than my 16 core CPU (maximum power ~200w).
Manufacturing is a different beast, you're right, but they don't need to be cutting edge for it to be viable. What's the latest gen chip that's no longer covered by patents? They can start there and work their way up. And, given that there is plenty of demand for previous gen lithographies, there's a market for the intermediate products they'd be able to produce off the line.
For example, the US Military would certainly prefer electronics manufactured in the US and usually prefer older, more reliable tech.
arg_max@reddit
Yes specialized hardware can be better than general purpose chips. But specialized hardware made with state of the art processing technologies is still gonna be vastly superior than specialized hardware made with 5 years old processes.
On smaller scale, this might be okay. E.g. in a car you can probably have x% higher power consumption, but in a data center this quickly becomes infeasible since cooling and power consumption are such a large part of total cost there.
giritrobbins@reddit
Isn't that advantage less? Isn't leakage current a huge issue with the most recent nodes?
jmlinden7@reddit
Newer nodes aren't actually smaller physically, they just use more complicated 3D transistors that have less leakage and faster performance. The node name is just an abstraction, i.e. 'if leakage didn't exist, then our transistors would have the same performance as a 2nm transistor'
cdoublejj@reddit
i was thinking resistors and capacitors or big NM chips for basic tasks, i can't see his claims being for AI
wintrmt3@reddit
CheesyCaption@reddit
There's an implications your making that isn't present in that language. Artificial intelligence projects use a lot more "chips" than just the gpu and memory modules. "Will be used in" doesn't mean that's all there is.
muntaxitome@reddit
Yeah I bet Tesla could do with having their own semiconductor fabbing for simpler items too, especially with the whole NXP mess. However the terawatt compute power premise of the undertaking suggests making advanced chips?
cdoublejj@reddit
yeah i didn't read, part of that being lazy and part of that being if i read articles on everyone of his grandiose claims i'd have no time for other news. apparently the claims are advanced chips. either they cracked something with air handling and building design integration or it's the bullet proof cyber truck glass all over again.
TheFaithlessFaithful@reddit
He said 2nm and beyond.
WingedGundark@reddit
This. If it would be that easy, there would be new fabs popping up whenever there is an opportunity in the market. The truth is that it is complex and costly business at any meaningful large scale.
splendiferous-finch_@reddit
He doesn't really need lithography from the damned dutch, they will build it from first principle!
muntaxitome@reddit
It's just aiming some light, how hard could it be?
splendiferous-finch_@reddit
Exactly, we can just rewrite the stack!
jinjuwaka@reddit
We'll solve it with the software!
This is a great plan!
Nothing can go wrong with this plan!
AfterAssociation6041@reddit
Just rebuild the the atoms and bits.
Masters of the Universe.
splendiferous-finch_@reddit
Atoms have the woke mind virus - Elon Musk
110397@reddit
@Grok
Deciheximal144@reddit
*Throws a blob of tin, takes notes*
neshi3@reddit
just hire some indians to point a laser pointer at a CD (no pun on indians, but it was funny that some of the more "novel" ai ideas where just indians in a call-center)
Boreras@reddit
EUV is woke, he'll use incandescent bulbs.
6198573@reddit
He just needs to watch the Veritasium video on DUV
Those fools practically gave the technology away for free!
splendiferous-finch_@reddit
Space based silicon factories next!!
Impossible_Raise2416@reddit
he learnt physics in college !
Swimming-Tax-6087@reddit
Develop their own lithography machines. Okay.
KeyboardG@reddit
Lie to investors. Hype himself up on paid for social media. The modus operandi.
lNTERLINKED@reddit
Don’t forget lying to the US government! He’s definitely aiming to get taxpayers to pay for this, and the US gov might be stupid enough to throw money at him in the name of homegrown “secure” chip manufacturing due to all of the China scare rhetoric.
ThePlanck@reddit
Elon Musk's superhuman ability is somehow always find himself and the Dunny-Kruger misplaced confidence peak
CheesyCaption@reddit
Like that time he said he could make electric cars desirable or land rockets after launching them.
Remarkable-Host405@reddit
i'm not sure you can put ev's on elon, that was more government subsidies. i've had 4 ev's and none of them have been teslas.
really, we should be putting gm on the back, obviously.
fair enough for the rockets
CheesyCaption@reddit
None of those EVs would have existed without the Tesla Roadster. Before the Roadster, EVs were undesirable, ugly cars. The model, not even being fully electric, is the Prius. After the Roadster proved that a market for high-end electric vehicles existed, then the subsidies came and the market exploded.
Sure Musk isn't the only one to get credit for that, he didn't create Tesla, but, without his funding (at the very least, arguably his leadership and definitely his marketing), Tesla probably would have never been able to sell any sort of volume and the electric car would have come and gone like it had so many times before.
How many of those 4 EVs you've had were released before the Roadster?
I've also never owned a Tesla and drive an EV, for the record.
splendiferous-finch_@reddit
We should rebrand it to "dunning -kruger presents the Elon Effect"
Whazor@reddit
He plans to first build a small, end-to-end chip fab in a single building. To make designing and testing a chip much faster.
I think this is smart from a point of chip design. You could make chips significantly more efficient with faster iterations. If you chip is 2x better than you need half the fabs.
he_must_workout@reddit
It took ASML over 30 years from the original idea of UV lithography to launching a commercial product. Veritasium did a great video on it a couple months ago
dfuqt@reddit
That video was great. I knew the machines were incredible, but I didn’t realise they were close to magic. Some of the technology and concepts that they talked about sound absurd.
classifiedspam@reddit
Not only that, but where will he get the power from? And the water? There's already not enough power (and power grid/infrastructure) to feed all the other datacenters that are currently being built. So he'd have to build some nuclear power plants nearby. And these things also need a lot of water to operate.
Stupid plan.
TheCh0rt@reddit
He's not, that's the point. That's not how Elon works. He grifts his way into money over time. Maybe it will happen, maybe it won't. But raising 5 trillion is silly. He says these numbers to move investment away from AI so he can start syphoning money away from Jensen Hwang as he looks to try to slay the beast. Basically try to give investors another safe haven piggy bank to dump their money into.
an_angry_dervish_01@reddit
Honestly the guy might pull it off somehow. Seems like people that don't believe it's too hard sometimes can. He has shown to be way over optimistic though at times and by the time he could deliver this would it even matter?
Street-Nothing-5213@reddit
By the time elon delivers on homegrown 2nm. Everybody else would probably be in the 2 angstrom range.
DZCunuck@reddit
Was thinking the same thing. Did anyone bother telling ASML about these plans? Or is Musk going to invent his own extreme ultraviolet lithography machines?
onegumas@reddit
Yeah, why not 30T? Use your imagination. Looks like us money becoming worthless.
soggybiscuit93@reddit
DXY shows the USD is modestly increasing in value
NoNameSwitzerland@reddit
He easier does nothing or goes the boring approach. Buy some old machine and start small.
jaaval@reddit
He’ll ask grok how to do it.
General-Demand7758@reddit
All these techno bro billionaires are pathological liars.
psilonaut0@reddit
It’s actually insane how many people call this guy an idiot and act like he hasn’t accomplished anything meaningful. I know this comment will get a bunch of hate because Reddit is a complete echo factory of leftist ideology. I am not right leaning at all. I don’t think Elon is a good person. He’s obviously not stupid and his lead in the advancements of important technology should at least be respected for the benefit it could have for the future of our species.
Teddy_RGB@reddit
I can’t believe people keep throwing money at this loser. Can’t even get an invite to the pedo island, but sensible people fall al over themselves to report on whatever fraud he cooks up
jinschizman@reddit
Holy shit I thought this was related to Terraform Labs haha... although it could end up with the same fate
BuchMaister@reddit
Put aside those preposterous numbers, just the scale of of those data centers that are planned and built, are just not the way to satiate the need for more capable AI. This just brute force and soon they will hit a wall.
Interesting_Cheek943@reddit
5T sounds kinda insane ngl 😭
feel like its more of a long term vision than actual cost rn. if he even pulls off small part of this it would still be huge for chip industry
EmployeeCautious6314@reddit
So, this is the new Tesla Roadster, Hyperloop, neuralink, etc. Roll up, gullible investors! Got money burning a hole in your pocket that you never want to see again?
neueziel1@reddit
lol remember when liberals worshiped the guy?
stoic_suspicious@reddit
I think when this part of history is written, Elon will either be a ponzi scheme villain no different from Madoff or a genius. He's on a knife's edge, and he's very impactful.
SirActionhaHAA@reddit
This man said his fab will let him smoke a cigar right next to the lithography machines without a clean room. What other lies do ya need? Fusion power in 5yrs?
nanonan@reddit
His claim is that current fabs are doing clean rooms wrong. Calling it a lie is simply ignorant.
cdoublejj@reddit
if by semiconductors they mean 1950s style 10watt resistors sure that should be true.
noiserr@reddit
Jeri Ellsworth (an electrical engineer / inventor) made a youtube video 15 years ago of making a home made transistor. So it's possible:
https://www.youtube.com/watch?v=-Qph8BNrnLY
So I'm sure they can make a transistor in the next 5 years.
Jaded_Celery_451@reddit
The Huygens Optics guy on youtube is doing some stuff in this space too. Buying lithography equipment from the 80s and actually doping silicon in your basement is a hobby now.
cdoublejj@reddit
she also did a video on installing a roll cage in camaro, shes so talented
splendiferous-finch_@reddit
Pretty sure the fusion power grift is sam altman as well... It's so hard to keep track of all the different conmen and thier grifts
jenesuispasbavard@reddit
You really think someone would do that? Just go on the internet and tell lies?
splendiferous-finch_@reddit
You are right... what was I thinking?... They even say it during investor calls and we know it's a sin to lie in those..
frogchris@reddit
So elon musk with no experience in semiconductors, patents is somehow going to compete with tsmc, Samsung, Intel and smic in leading edge semiconductor process. Lol.
He's somehow going to reinvent the field that many smart people, companies, governments are pouring hundreds of billions and trillions of dollars into making extremely efficient.
Then on top of that he's going to need hundreds of billions of dollars in r&d to keep his fab up to date and competitive since semiconductor is a highly capital intensive business.
Majestic-Volume9996@reddit
I mean to be fair, that's not too far off from what Rapidus is doing. As long as he doesn't think he's going to do it all from scratch, he might be able to at least pull some sort of something off.
frogchris@reddit
Rapidus is supported by the government of Japan with the help of other technology from other companies . And they are not targeting mass production. They are for some niche application where cost does not matter. It will mlst likely fail to compete with tsmc, Intel, Samsung in the long run as they do not have experience in advanceded packaging and fabrication techniques. And they have no history of custome support and meeting deadlines, which is 95% of what fabless companies care about.
For tesla and SpaceX to use their own foundry, one of which must be true. Either their foundry is super cost effective. Or their foundry is much better performance than something they can order off tsmc. If tesla is using expensive chips then their products will not be cost competitive. If it's poor in performance then their products will not be competitive application wise.
Then they need to spend tens of billions of dollars every year to upgrade, perform r&d just to keep up with tsmc.
shmeu@reddit
So Elon Musk with no experience in rockets, somehow started a rocket company that is leading the tech in the field.
He might overpromise and have behavioural issues but that doesn’t mean he didn’t help build a lot of value for electric transportation and space exploration.
soggybiscuit93@reddit
Totally incomparable. Rocket flights were mainly a government venture for research. He entered into the market right at the time where rocket launches now had reasons for the private sector to pursue - such as deploying low cost satellites.
Reusable rockets are also way easier than building and running a leading edge fab
CheesyCaption@reddit
That's backwards. People have wanted to put small satellites in orbit forever but couldn't do it due to costs. SpaceX didn't create the market, it has existed since forever. Sputnik would have launched on a Falcon 9 if it were available.
soggybiscuit93@reddit
Well, initially the only organizations capable of starting the space race were governments because there was no path to profitability and the upfront capital was massive. The US was spending close to 1% of its GDP on it, plus had to fund and build up the semi-conductor industry to support it as well.
A fleet of Sputniks wouldn't be profitable. Once technology matured to a point where private companies could deploy imaging and wireless internet satellites at reasonable speeds and costs was there a demand for private satellites
CheesyCaption@reddit
And that happened decades ago.
soggybiscuit93@reddit
Right. SpaceX was founded 24 years ago to disrupt an underserved market mainly controlled by DoD contractors, with a market that now had non-government customers
CheesyCaption@reddit
Sorry, that happened more than half of a century ago.
CubeSats have been launching since 1998. A satellite spec designed by scientists to ease of cost and difficulty of launching multiple small satellites (so that they can share the cost of the launch) because they had a shitload of experiments they wanted to run in space. The desire to cheaply launch things into space, especially for research purposes but, eventually, commercial purposes, has existed since Sputnik.
There has always been money in selling orbital launches, ask Russia. SpaceX didn't enter the market at the perfect time at all, SpaceX was the first and only company to be able to capitalize (make a profit on) on the demand that's existed since before the space race even began.
dandeeago@reddit
To be fair, this is exactly what people said about creating a private space company as well, especially one that will be creating their own engines. It can’t be done.
TheFaithlessFaithful@reddit
Space is a much less established industry and the competition there was pretty weak as it was solely a few inefficient DoD contractors.
Chips are actually profitable, with a bunch of companies, and the cutting edge even has 3 highly competitive companies (TSMC, Samsung, and Intel).
jocnews@reddit
And the technology is only getting harder and harder every generation. That's why so few manage to stay in the business there.
CaptainMonkeyJack@reddit
Except spaceflight is trivial vs fabs, and was deeply outdated. TSMC is not some legacy business waiting for disruption.
Remarkable-Host405@reddit
"spaceflight is trivial"
shenmethefucc@reddit
Compared to leading edge nanofabrication, it is trivial
jocnews@reddit
Actually this. The bleeding edge nodes are some of the most complex technology ever.
Check how crazy mechanism the EUV machines use to produce the radiation used to etch this super small structures. I find it crazy we can even solve the "boring" things like how to properly align the machines so that the light etching and deposition arrives at the right places...
TemuPacemaker@reddit
We managed to do it in the 40s.
splendiferous-finch_@reddit
It's because he used fraud, lobbing and connection to essentially eat NASA, both interms of funding and people.
You can't do that here simply because you have actual competition. So no that is not a fair comparison, it was a niche industry where innovation was stifled mostly due to burecrartic reasons.
Look at YMTC and CXMT, both started having realistic understanding of where they are and what the need to do and even with all the support they had in terms of money and a rather "relaxed" view of patents and IP they are still only now after 10 years getting there in terms of memory, and all the GPU companies from China are still wat way behind.
Yet somehow Elon will be able to do much much more because his other company can build a rocket engine ?
Hey remember the time Elon made fun of BYD cars? when he had like a 10 year head start, How did that turn out I wonder
noiserr@reddit
Rocket engines are a 50/60s invention. This is completely different level of technology.
soggybiscuit93@reddit
He'd be better off focusing on setting up a permanent moon colony with Space X. It'd be cheaper and easier than entering the leading edge fab market
frogchris@reddit
There were barely any private space companies. Semiconductor is pretty mature and tons of people working on it. Every problem is being looked at and many researchers are working on it. He's not going to reinvent the wheel when he can't even make 22nm nodes lol. Not even byd can do that, they buy their chips from Nvidia or license from an existing foundry for leading edge.
In order for your fab to be profitable and pay back the r&d cost, you need to sell a lot of chips. His companies SpaceX and tesla don't sell enough or produce enough chips to warrent an independent fab. There's a reason why companies switched to the fabless model. It's like he's trying to repeat the mistakes of the past.
The only hope for Elon is that he makes a joint fab with Intel, using intel r&d but fab for his companies only. Or he buys global foundry and start from where they are until he reaches something competitive with tsmc in 15 years and hundreds of billions burnt at the same time producing worse chips performance and power wise. Both will be extremely expensive.
KeyboardG@reddit
Expect it right after the Roadster 2
-S-P-E-C-T-R-E-@reddit
Don’t forget he stole Eberhards Roadster (after stealing his company) and shot it into orbit on his stupid rocket. He is beyond vile.
filisterr@reddit
And the Hyperloop.
FIFofNovember@reddit
And his mars base and moon base
-Suzuka-@reddit
And space datacenter satellites.
ZarK-eh@reddit
A moon base makes sense. Space 1999!
filisterr@reddit
He and his sycophant friends can all embark on a journey to Mars, they won't be missed.
Fragrant_Equal_2577@reddit
They should restart the 450mm wafer size project. This would halve the number of wafers. Few years back, semiconductor industry estimated the 450mm migration to cost ~$20bn.
sketchysuperman@reddit
$20bn isn’t a drop in the bucket for what it’d cost the industry. It would cost more than that to “convert” a fab from TSMC or Intel.
You’re talking about completely new tooling and all of the infrastructure surrounding it. Not even taking into account the R&D cost of making the new found wafer real estate useful.
Fragrant_Equal_2577@reddit
The $20bn estimate was about developing the 450mm equipment+ unit process module capability. 450mm consortium was fairly advanced before TSMC effectively pulled the plug from the 450mm.
splendiferous-finch_@reddit
Building them from first principles I am sure.
There has never been a project that will have more cost overruns and still not delivered anything.
Also this 5 trillion number makes no sense.
Cute-Pomegranate-966@reddit
That number is the estimated cost based on him saying how much capacity they will be able to print.
He claims it'll only be 200B. Independent cost estimates are 5T due to his claims of capacity.
splendiferous-finch_@reddit
I know it just seems more like throwing a biggest possible number then actually making a informed guess.
One againbits a Elon claim which means it doesn't even have to make sense.
jocnews@reddit
The other techbros (OpenAI, Nvidia) actually started that meme. I suspect it's actually a show of arrogance - they want to flaunt how they are wasting the energy. Because they are so huge and they can while you struggle with your perpetual credit card debt. Maybe it's supposed to be some "Silence bottom (and pay the subscription to our crap service)." power move?
splendiferous-finch_@reddit
See this sounds very Machiavellian of them.....we know tech bros are too dumb for that... I suspect one of them saw a big number in thier internal technical presentation and decided to present it to the investors who equally dumb told other VC bros and round and round it went.
jocnews@reddit
Are they sophisticated enough? Dunno.
Are they arrogant and sadistic enough? Probably.
splendiferous-finch_@reddit
The only source of enjoyment for some in particular I agree
panckage@reddit
You can measure compute power in several ways. He decided to use power input so he could make a comparison on the Kardeshev scale.
And remember it's energy that they are saying is the #1 bottleneck in producing AI computer faster.
Cute-Pomegranate-966@reddit
Trust me I don't believe him whatsoever. He must be in danger of over leveraging assets again and needs a scam.
splendiferous-finch_@reddit
I thought going public with spacex was the current out
Sorry_Soup_6558@reddit
I mean I think it would be 5 trillion but it doesn't exist and won't exist like self driving cars and Tesla robots
Chipay@reddit
And Mars colonies, and hyperloops
splendiferous-finch_@reddit
Don't forget the automated trucks and space data centers.
kensaundm31@reddit
I'm really enjoying the + 1 million robotaxis that were on the road since 2019...
splendiferous-finch_@reddit
Yeah all the early Tesla buyers were a me to run Thier cars as robotaxis have the used prices are not more then new... Unbelievable
splendiferous-finch_@reddit
Next year(TM)
Simple_Feed_3426@reddit
what are the chances that, to free up some power and availability constraints, they switch to using Canon's Nano imprint? IIRC Canon was able to get a somewhat working chip.
Lowetheiy@reddit
More anti Elon propaganda, as if some armchair internet genius knows about semiconductor fab costs than the experts lol.
noeldr@reddit
Twitter all over again but at a higher level, Teratwitter ?
Classic_File2716@reddit
If anyone can it’s Elon .
cil1@reddit
Sounds like an Elon Musk too big to fail plan after he gets the US government fully invested.
rootkeycompromise@reddit
I don't get this Terafab idea, if not just another stunt to pump the Tesla stock price. There are already very experienced players in the market - with existing operations - expanding their own capacity on proven technologies and processes, that they have no doubt spent considerable time and effort in developing.
Why in the world can Tesla not just reserve capacity like any other? What is it that Tesla does, that somehow gives them an edge over incumbents in semiconductors?
By the way, 5 Trillion USD is 1 Trillion dollar more than the market cap of nVidia. He can buy that instead and keep the change.
panckage@reddit
Because of vertical integration. As he learned with SpaceX, you can't always rely on others.Elon loves risk and It's also much cheaper.
The last thing the world needs is another MS to buy up the competition.
rootkeycompromise@reddit
If vertical integration was always a cheaper option, why would anyone rely on external suppliers and not just vertical integrate? Because vertical integration is about control, not cost.
And typically it’s better for a company to not overextend itself into too many areas of specialty. There is no reason why Tesla/SpaceX should expect to outperform existing semi industry, and by that logic, they should just procure what is already in the market.
adh1003@reddit
It won't cost anything like that, because it'll obviously never be built - at least not per Musk's latest "don't look at Tesla's numbers, don't look at Tesla's safety record, don't look at the Epstein files" grifter fever dream.
seb21051@reddit
Peanuts. Once the "Culture" society arrives (in 10K to 100K years), money will have no role.
SourceScope@reddit
Ok
I assume us tax payers will end up paying for it
MumrikDK@reddit
At least for the only part that actually gets paid.
Z3r0sama2017@reddit
If we pay for it we own all the profits.
their were, infact, no profits
jocnews@reddit
I suspect that at best it will be "owning the libs" (ruining the country/environment/whole world)
DymlingenRoede@reddit
Only if he actually starts building it.
Actually building it is incidental to the primary purpose, which is juicing stock prices with big promises.
AnechoidalChamber@reddit
Remember folks, Musk once said "Mars by 2022".
Still waiting...
Electrical-Leave-986@reddit
Five trillion is a wild number, even for Musk. If this is even remotely true, he’s probably looking at a timeline that spans decades, not years. The real question isn't just the money, but where the talent and raw materials for a project that scale would even come from. It sounds more like a multi-generational move to secure the future of AI and robotics than a standard fab project.
EmergencyCucumber905@reddit
When someone says “1 terawatt of power per year,” what they’re doing is revealing—quite spectacularly—that they don’t understand the difference between power and energy, which is, you know, Physics 101.
Z3r0sama2017@reddit
Damn I kinda wished each chip used one TW. Then the next scam would be "how many suns do we need to power our datacentres?"
Wait_for_BM@reddit
I was going to say the same, but it is just bad writing. I parse the badly written sentence differently as "Each year he would make enough chips to waste 1TW of power".
greasyee@reddit
The point is obviously that the total power draw of all the chips produced by that factory per year is 1 TW.
splendiferous-finch_@reddit
Physics is woke - Elon Musk
twoddle_puddle@reddit
Would be good to fix world hunger first.
Ok-Lynx-7484@reddit
Yeah this is never going to happen lmao. Remember when he said spacex was going to launch reusable rockets? How’s that going
soggybiscuit93@reddit
Of all the things you chose to criticize Musk for, and there are many good reasons to, you chose something he actually did do lol. SpaceX does have and launch reusable rockets
itsaride@reddit
Of all the sarcastic replies you chose the most sarcastic to fly over your head.
soggybiscuit93@reddit
Poe's law and all that.
On second reading, while he may have been sarcastic, there definitely do exist people who would say that earnestly.
ie-redditor@reddit
It´s ok, with the money from the space datacenter that is like some spare cash to him.
edvlili@reddit
Look at Mr space genius here: wants to build a chip fab on the moon but earth has more Silicon concentration than the moon.
spiker611@reddit
Lots of people commenting on the "$5 trillion" figure which according to the article is:
So it's napkin math based on what is current tech at today's prices. Obviously this endeavor won't materialize within a few years so to say it'll cost $5T is absurd.
Fritzo2162@reddit
Just invent Tesla shares to pay for it...you know, like they do for everything else.
bargu@reddit
Why stop there, let's build a $500T fusion reactor and a $10000 quadrillion wormhole between earth and Andromeda galaxy, numbers don't mean anything anymore anyway.
Deciheximal144@reddit
Pheh. They'll just make something smaller than this, like they did with the Gigafactory.
dustingibson@reddit
Never trust a vaporware salesman.
DimMak1@reddit
This would cost taxpayers $5 trillion. Musk has never paid a penny of his own money for any technology he lied about developing
GongTzu@reddit
Lmao. The fantasy money these billionaires are talking about are such a long way from normal society
AutoModerator@reddit
Hello sr_local! Please double check that this submission is original reporting and is not an unverified rumor or repost that does not rise to the standards of /r/hardware. If this link is reporting on the work of another site/source or is an unverified rumor, please delete this submission. If this warning is in error, please report this comment and we will remove it.
I am a bot, and this action was performed automatically. Please contact the moderators of this subreddit if you have any questions or concerns.