Report: Sam Altman seeking trillions for AI chip fabrication from UAE, others

WSJ: Audacious $5-$7 trillion investment would aim to expand global AI chip supply.

See full article...

 

[jjmorris2000]

JFC, this has WeWork 2.0 written all over it.

When I saw Softbank on the list that was exactly my thought - Altman is hoping to replicate Adam Neumann's success.

And by success I mean "getting Softbank to write him huge checks that he got to keep even after he got removed from the C-suite of his company". Don't underestimate how successful Neumann is seen in "entrepreneurial" circles. Many of them would dream to "fail" like Adam Neumann did.

 

[nights]

When I read a Sam Altman headline my brain is always trying to figure out how he does all of that from jail. After a couple of seconds I realise that’s the other guy. Might be a problem with my personal brain-implemented LLM that headlines with these two are statistically pretty close together.

 

[jtwrenn]

I'm sure none of these politicians have any hidden investments to push their ideas at all. No bias here I am sure. Just good old fashioned proper spending right? Homelessness? Food insecurity? Clean energy? nah, small potatoes, I need to be able to ask a computer to make fake porn of celebrities and have it happen really fast.

 

[sigmasirrus]

Putting aside the nauseating idea of spending that much money on computers, I remain convinced that GPU/TPU hardware will not solve GAI. I think we have to look at advancing neuromorophic hardware too slash the power requirements.

I also don’t think that LLMs are going to magically become general AI just by running bigger models on more powerful chips. It would be better to put that sort of crazy investment into really tediously studying the biological neural networks of lower animals and work your way up. You could perhaps begin to see how brains do rudimentary reasoning, memory representation, etc, if you start small enough and build on what you learn. We have already mapped a few connectomes, but interpreting them is tough. $7 trillion would go a long way towards that and probably lead to a genuine breakthrough, rather than another stupid copy pasta parrot.

That being said, we don’t need a $7 trillion AI. How about some humanitarian aid? Livable wages? Environmental cleanup? $7 trillion could go a long way to improving lives. It’s just sad that it’ll be wasted on this crap.

 

[BicyclingGreen]

Current AI projects have added significantly to energy demand and caused coal plants that were due to be retired to stay online. If this goes ahead (even in a much smaller form) that’s not going to help with rapidly worsening climate change.

 

[xPutNameHerex]

What do you think the rate of return will have to be to justify investing 6 trillion dollars upfront? That's more than the spending budgets of many countries.

Most countries? That's more than any country other than the USA spends per year.

 

[citizencoyote]

I work in DSS. Where people don't have a house, health care, or food and ask for help constantly. And this fucker is asking for trillions in investments for fuckin' chips?! Fuck this place.

See, you would use that money to help people. What are people but bags of meat that take up resources?

He, on the other hand, would use that money to create value. That's really what's important here.

(/s obviously)

 

[fredrum]

Most countries? That's more than any country other than the USA spends per year.

If you don't ask you don't get.
Maybe in a pinch he could do it for...say 4 trillion?

 

[jjmorris2000]

That being said, we don’t need a $7 trillion AI. How about some humanitarian aid? Livable wages? Environmental cleanup? $7 trillion could go a long way to improving lives. It’s just sad that it’ll be wasted on this crap.

But think of the business owners and how many workers they could replace with digital slaves! Not having to worry about pesky human concerns and can concentrate entirely on how much money they are earning as they run up the rich person scoreboard to compete with each other has got to be worth a few trillion to them.

/s. Except cynically not really.

 

[Purpleivan]

I doubt if any business calculation in history, other than hypothetical naval gazing exercises, have come up with a budget of $7,000,000,000,000.

The closest thing would be calculating the total US government expenditure for a year ($6.3 trillion in 2022 according to Wikipedia).

 

[tuple]

If anyone wants to know if AI is going to be successful this time around, just note that the head of OpenAI appears to need a whole new, multi-trillion dollar industry created for AI to be a viable product...

 

[ArsPropaganda]

But think of the business owners and how many workers they could replace with digital slaves! Not having to worry about pesky human concerns and can concentrate entirely on how much money they are earning as they run up the rich person scoreboard to compete with each other has got to be worth a few trillion to them.

/s. Except cynically not really.

This is truth not sarcasm or cynicism. no need for the "/s"

 

[ColdWetDog]

I also don’t think that LLMs are going to magically become general AI just by running bigger models on more powerful chips. It would be better to put that sort of crazy investment into really tediously studying the biological neural networks of lower animals and work your way up. You could perhaps begin to see how brains do rudimentary reasoning, memory representation, etc, if you start small enough and build on what you learn. We have already mapped a few connectomes, but interpreting them is tough. $7 trillion would go a long way towards that and probably lead to a genuine breakthrough, rather than another stupid copy pasta parrot.

That being said, we don’t need a $7 trillion AI. How about some humanitarian aid? Livable wages? Environmental cleanup? $7 trillion could go a long way to improving lives. It’s just sad that it’ll be wasted on this crap.

This money is supposed to create something that makes more money. That's the entire game. Now, the funny thing is (for certain sardonic definitions of funny) is that by actually giving your population a clean environment, decent education, nutrition and health care you would grow your economy much faster and more robustly than creating the world's largest LLM.

But that makes everyone better off, not just the investors. So, no sale.

 

[jtwrenn]

Kind of interesting that TSMC is reported as a fab, but Intel isn't. I know they have had a lot of issues but they seem on track to finally start growing their fab business. We need more fab companies not for TSMC to be even more dominant at this point. I hope the recent visit to Intel's fab talks means he is spreading the money around.

All of this is a bit ridiculous but if you are going to go there then at least globalize it rather than putting everything into a Taiwan company that China is eyeballing and sharpening it's knives. We need to make Taiwan a whole lot less important globally or a major pressure cooker will just get hotter.

 

[Jeff S]

I wonder if we will hear from any conservatives/libertarians about the perils of central planning?

I mean while this isn't exactly the government doing central planning, if one single corporation gets investment to the tune of trillions of dollars on as-yet-unproven technology, where if that corp fails it's going to substantively impact the entire world economy. . . that starts to look an awful lot like the sort of central planning that they like to warn about the hazards of. . .

 

[Photon_plumber]

Convince the US military it's of extreme strategic value and they'll foot the bill. Obviously requiring US locations for all development, research and production. While part of me is still questioning many uses of AI, part of me says if it all pans out, you want to be at the top of food chain. Being left behind may be disastrous.

 

[DesiLurker]

and in return we'll be hostage to one of the most regressive regimes on earth for all eternity, you know because AI. Also special provisions to convert everyone to 'preferred religion'.

 

[MicroManager]

That guy stinks like poopoo. His plan must be to ripoff Nvidia tech. He is way behind.

 

[Photon_plumber]

All of this is a bit ridiculous but if you are going to go there then at least globalize it rather than putting everything into a Taiwan company that China is eyeballing and sharpening it's knives. We need to make Taiwan a whole lot less important globally or a major pressure cooker will just get hotter.

Many people misconstrue the unspoken purpose of the CHIPs Act (and IRA Act). It's, as someone else mentioned on the other thread today, 'a means of getting independence from China as the sole provider of certain technologies and materials that are critical to defense and industry, with the idea that Taiwan may be compromised in the future'. Best to move some things back here to US shores.

 

[lmmo]

Only $5-7 trillion? Why not $50-70 trillion?

 

[arcite]

If anyone wants to know if AI is going to be successful this time around, just note that the head of OpenAI appears to need a whole new, multi-trillion dollar industry created for AI to be a viable product...

Also, he’s relying on Sheikh oil money as a key backer. In the land of sand and mirages…. Automatic alarm bells.

 

[matthewslyman]

I'm beginning to suspect everything Sam Altman does is overstated by multiple orders of magnitude, including the capabilities / goals of OpenAI.

It's such a relief to me that I'm not the only one thinking this. Thank you. Once again, my confidence in the ultimate efficiency and rationality of the market is somewhat restored (even if we often take some bizarre detours over time horizons of a few months or years). There's more to AI than Sam Altman, in the same way that there's more to the Internet than Bill Gates.

 

[psarhjinian]

Did he come up with the idea? Or is the AI in the back office starting to flex its muscles?

Roko's Capitalisk.

 

[Legatum_of_Kain]

Did he have an AI write the proposal? Maybe that's why the numbers are so large, LLMs seem to be bad at math.

Joking aside I don't see how this is in any way a reasonable plan. Where is he expecting the money to come from? What's the ROI for investors expected to be? How does any of this even begin to make sense?

Edit: You know what, to put these numbers to scale, Microsoft's market cap is currently about $3 trillion. He's asking from about double that to more than double that. For a single venture: AI chips.

He’ll deliver ROI in 70 years. Or after investors die of old age, whichever comes first.

 

[ktmglen]

To hit these ambitious targets—which are larger than the entire semiconductor industry's current $527 billion global sales combined...

Completely off topic but to put things in perspective about the size of the semiconductor industry, Walmart's FY23 annual revenue was $611.3 billion. Walmart has almost always been larger than the entire semiconductor industry.

 

[Viki Ai]

I'm going to give what will obviously be a very unpopular opinion. Everyone here saying this is ridiculous sounds a lot like Bill Gates saying no one will ever need more than 640k in a PC. I don't know if he'll get it, but if he does, I think his investors are going to be very happy in 20 years.

Um.... You are aware that Bill Gates quote is an urban legend, with no evidence that he ever said it, right?
Of course you aren't.

 

[matthewslyman]

Oh good, the UAE really needs more money!

Investing 10% of the amount Altman demands will make either a material positive or negative difference to the national economies of those who invest. Sam Altman has no clue how to build a business so large and complex that it needs more than $1 Tn in market capitalization. Further, he has no clue about cyber security, let alone semiconductor manufacturing engineering, or the highly theoretical specification & verification that's needed to get semiconductor hardware designs right first time (and not manufacture [m|b]illions of chips with hard-to-detect fundamental defects that didn't show up in the battery of unit tests they applied). The algorithms, knowledge, and organizational qualities needed to get this right are completely different from an AI startup. Prove me wrong.

In the current high interest environment, investors might start to take these exaggerated demands as a clear signal that the market bubble is topping out. Is it time to sell?

 

[nash076]

Currently "AI" is not profitable. What's keeping it afloat is airdrop after airdrop of investor capital, and that's going to dwindle in time. There's no real proven business model yet for making all this stuff financially self-sustaining, if there ever will be.

And Altman wants another $5-7 trillion from multiple foreign sources including other governments.

Remember when the scam was just to make a semi-functional app, con Google or Facebook into buying you out, and then run like hell?

 

[brewejon]

I had to do a cartoonish double-take upon reading the headline, because $5-7 trillion with a t is the kind of monetary value I would come up with as a kid... like according to Statista, a (very rough) estimate of global GDP in 2023 is something like $100 trillion. You're telling me that this Altman guy is seeking 5% of that (or roughly 20% of America's GDP) for producing semiconductor chips for ML processing?????? I swear he's on the same shit that SBF was (and probably is) on, because this doesn't make even an iota of sense!

He's probably trying to speedrun the singularity like the longtermist he is, and thinks that once we hit the singularity then money won't matter. In which case money doesn't matter now, and it's his responsibility to make the singularity happen as fast as possible.

Except that money does, of course, for very many people, matter now.

 

[Cheshire Cat]

Ok, $1.8 trillions to buy NVIDIA. Fair enough.
What's the other $5 trillions for?

 

[lurknomore]

The pentagon would like to point out they could also solve all of the world's problems after getting a 10x raise (solve from a certain perspective, obviously).

Thanks for informing us that Altman has severe issues with reality. You may now ignore all future BS coming from him.

 

[Dr.Appleseed]

Ridiculous. Apple or MSFT market cap $3 T each. NVidia $1.8 T TSMC $0.5 T Building a latest semiconductor fab costs $0.1T. US Federal Budget 2023: $6 T. US defense budget $0.8T So what he requests is more than yearly US federal budget and 10 years of US defense budgets.

 

[Andrewcw]

Let say you get a 1 Trillion on investment. $300 billion goes into building a fab within a year. Because well you're going to spend 30X the amount trying to rush this fab into production. Leaving you 700 Billion of what is that money going to.

Investors want ROI and with Twitter and FTX fresh in large investors minds.

At even 7 Trillion you could just buy Intel, TSMC, UMC and control a majority of the chip manufacturing without even having to build a Fab.

 

[Sadre]

This crop of tech masters have really sophomoric ideas about what it means to be outside the box. American mega-billionaires and tech geniuses, thinking they are clever by making geopolitical moves?

DARPA already has whatever they think they are inventing. And the NatSec info world is not impressed by a high-level meeting with a foreign wealth fund. They will be inside that deal before you even hear the yes.

A lot of shit can happen between conception and implementation, tech billionaires. But thanks for stating the obvious and getting the spigots flowing.

 

[eas]

Sounds like a stunt to hype OpenAI. Another pump-and-dump.

 

[ColdWetDog]

Convince the US military it's of extreme strategic value and they'll foot the bill. Obviously requiring US locations for all development, research and production. While part of me is still questioning many uses of AI, part of me says if it all pans out, you want to be at the top of food chain. Being left behind may be disastrous.

FOMO, to the tune of a significant fraction of the world's economy, strikes me as a rather foolish investment strategy.

 

[stk5]

Ok, $1.8 trillions to buy NVIDIA. Fair enough.
What's the other $5 trillions for?

Orbs, I assume.

Slightly more seriously, I think he looked at the $527 billion number and decided “we need at least ten times as much compute” and that’s as deep as he got.

 

[hambone]

This Altman guy has a serious messiah complex. And it causes him to make dangerously cavalier statements about the consequences of his actions.

Like climate change "just requiring some geoengineering and oh yah fusion power" so he can burn enough power for his AI farms.

The pattern has been consistent too. He has acknowledged massive structural disruptions to free elections and the labor force while simultaneously saying it'll all just have to sort itself out, as if releasing AI is an inevitability humans can't control.

It's really unhinged. And kind of dangerous given Altman's influence at the moment and the thrall with which less sophisticated politicians want to believe that AI will make bread and honey fall from the economic clouds.