Sam Altman seeking 5-7 TRILLION in backing for Open AI CPU Creation

[tkitch]

Summary

 

 

Sam Altman is shopping around ideas and trying to get backers to support OpenAI creating new AI hardware for future development of the space.  Report say up to 5-7 TRILLION is a number being mentioned

 

Quotes

Quote

OpenAI CEO Sam Altman is seeking to raise trillions of dollars from investors, including the United Arab Emirates government, to boost the world’s capacity to produce advanced chips and power artificial intelligence, The Wall Street Journal has reported.

Altman’s “wildly ambitious tech initiative” could require raising as much as $7 trillion, the WSJ reported on Thursday, quoting people familiar with the matter.

 

My thoughts

 

 

AI is cool and all, and I'd love to see new players in the game (Might force NVidia to do something more interesting than they've been doing), but jesus christ...  HOW MUCH MONEY!?!?!

 

 

Sources

https://www.aljazeera.com/economy/2024/2/9/openais-sam-altman-seeking-trillions-to-fund-chips-for-ai-report-says

 

https://www.cnbc.com/2024/02/09/openai-ceo-sam-altman-reportedly-seeking-trillions-of-dollars-for-ai-chip-project.html

 

https://www.nbcdfw.com/news/business/money-report/openai-ceo-sam-altman-seeks-as-much-as-7-trillion-for-new-ai-chip-project-wsj/3457888

[manikyath]

i'd love to have 7 trillion as well.

 

i think this man has been around big numbers he forgot their meaning..

 

SpaceX quoted their first generation of falcon 9 rockets to have cost 300 million to develop

NASA quoted that if they had developed such a platform using their own strategies, it would have been in the 3.6 billion category

SpaceX estimates that starship will cost between 5-10 billion to develop

nvidia spends 7.34 billion on R&D each year

 

so.. what this man is suggesting, is that his endavour will cost the same as:

- nvidia funding their R&D for the coming 950 years

- spacex developing starship, estimating they double their original budget at 20billion

- NASA copying SpaceX's homework and making a falcon bureaucracy edition.

- still have enough money left over for spaceX to throw away and re-invent falcon 9 not once, not twice, but 11 times.

 

please.. someone quote me to tell me that the order of magnitude got lost in translation somewhere...

[Skipple]

I... what? This has to be a mistake. $7 trillion is more money than the entire US Federal operating budget. It's a quarter of the US GDP. The concept doesn't make sense sense. Like, $7 trillion on a future evaluation I can maybe see, but what could possibly be the scope of a project could that costs $7 trillion?

[da na]

Looking past the absolutely ludicrous amount (which is about 10x the wealth of the 10 wealthiest countries combined), why do we need an OpenAI CPU?

[ToboRobot]

If you compare the expense to GDP it's comparable to the Manhattan project for the people who think it's unreasonable.

I'm too lazy to do the math. sorry.

[Lunar River]

Sam has lost his mind

[starsmine]

10 minutes ago, ToboRobot said:

If you compare the expense to GDP it's comparable to the Manhattan project for the people who think it's unreasonable.

I'm too lazy to do the math. sorry.

is it? manhattan project was 2.2billion. 

That's only 31 billion ish today.

[ToboRobot]

4 minutes ago, starsmine said:

is it? manhattan project was 2.2billion. 

That's only 31 billion ish today.

GDP was smaller back then.  

 

stolen from hacker news comments.

"For context in WWII, industrial defense production reached 10% of GDP in the US.

World GDP is now $100T.

If the $5T is spent over 5 years, that's 1% of GDP — incredibly high, but within the realm of possibility if we become singularly focused on building out AI. "

[Eigenvektor]

If every person in the world would simply chip in the small amount of $1,000 he'd have that money tomorrow. Surely new AI chips should be worth that much to all of us, right?

[starsmine]

14 minutes ago, ToboRobot said:

GDP was smaller back then.  

 

stolen from hacker news comments.

"For context in WWII, industrial defense production reached 10% of GDP in the US.

World GDP is now $100T.

If the $5T is spent over 5 years, that's 1% of GDP — incredibly high, but within the realm of possibility if we become singularly focused on building out AI. "

ok?

yes GDP was smaller back then. that doesnt change what I pointed out. 
GDP in 1947 was 250billion
2.2 billion was not even 1% of the total

5 Trillion today is almost a quarter of the our GDP. 

[ToboRobot]

5 minutes ago, starsmine said:

ok?

yes GDP was smaller back then. that doesnt change what I pointed out. 
GDP in 1947 was 250billion
2.2 billion was not even 1% of the total

5 Trillion today is a quarter of the our GDP. 

But you are thinking of this as American, rather than Global. 

 

"OpenAI CEO Sam Altman is seeking to raise trillions of dollars from investors, including the United Arab Emirates"  <- not USA

US GDP was like 120B in 1941, it took time to massively inflate from the war. 

5T over 10 years us 500B per year and current US GDP is $27T. 
 

[Lunar River]

38 minutes ago, ToboRobot said:

For context in WWII, industrial defense production reached 10% of GDP in the US.

Defence spending during the deadliest war in history

vs

a chip for a single company

 

what a terrible comparison

[porina]

When he says "making chips" I don't read it as design (like AMD/nvidia), but more like fabs? Think TSMC, Intel, Samsung.

 

However "trillions" seems order(s) of magnitude bigger than needed. TSMC has a market cap of just over 0.5T USD. Intel is loose change at 183B USD. ASML, who make critical optical systems essential for leading edge nodes has market cap of 350B EUR. One trillion USD could almost buy all those.

[NobleGamer]

If there is one place where OpenAI could get that kind of money, it would be the kind of people and businesses in or tied to UAE, if the amount of extravagance in Dubai's buildings & resorts and stuff is any indication of big oil money & such in the middle east.

 

Whether OpenAI would really use or "need" that kind of money is a different story. Maybe he just said something outlandish to get press from WSJ that would roll downhill to all other outlets. I'm a subscriber and with access to their relevant article, and it doesn't really provide anymore enlightenment beyond them wanting more chips so they can build AI more/faster.

 

Maybe this is Sam's Elon moment of sharing certain outlandish goals for electric cars and space stuff that isn't close to being achieved, but that's not the point and it's just an indication of a direction they're heading.

[Taf the Ghost]

As dumb as this sounds, if he really wants to make another entire silicon fab industry, that's actually what it would cost.

[GOTSpectrum]

5 minutes ago, Taf the Ghost said:

As dumb as this sounds, if he really wants to make another entire silicon fab industry, that's actually what it would cost.

Yeah and the wording seems to point to that

[tkitch]

1 hour ago, Taf the Ghost said:

As dumb as this sounds, if he really wants to make another entire silicon fab industry, that's actually what it would cost.

how do you figure?
TSMC has a total market valuation of like 1/10th of that?  

Or would he be trying to build something on a scale to dwarf TSMC and Intel combined?

 

[05032-Mendicant-Bias]

Sam Altman could instead ask to replace the US Dollar and the Federal Reserve with SamBucks, a crypto currency whose fiscal policy is handled by the SamAltman AI core. /s

[porina]

2 hours ago, tkitch said:

TSMC has a total market valuation of like 1/10th of that?  

I posted it earlier, TSMC market cap in USD is about half a trillion.

 

2 hours ago, tkitch said:

Or would he be trying to build something on a scale to dwarf TSMC and Intel combined?

At the moment the likes of TSMC and Intel still rely on external suppliers to make chips. Think things like substrates. Maybe a single company that does everything under one roof could inflate the cost.

 

If the goal is to "have a fab" then I think the cheaper/faster way to do it is to buy out existing ones. GlobalFoundries is a trifling USD 31B market cap, or UMC at 19.2B. If he wants more leading edge node access, I think Intel is relatively underpriced in the space, and the fab vs rest of company could be split off and sold if not of interest to recover some of that cost. Intel do have some noisy but thankfully insignificant shareholders who are pushing for them to spin off their fabs and do an AMD.

[Monkey Dust]

Assuming he could raise this kind of money, what is he planning to spend it on? Where would you find that many employees to reach that kind of wage bill?

 

Even if this was funding for initially 10 years, that's a burn rate of 500-700 billion a year. Even if he employed 1 million people on an average of 200k a year, that's 200 billion a year. Leaving 300-500 billion of capital investment a year. The odds of being able to spend that much without massive, massive waste is zero.

 

It would be cheaper and quicker for him to just buy Nvidia.

[leadeater]

6 hours ago, tkitch said:

how do you figure?
TSMC has a total market valuation of like 1/10th of that?  

Or would he be trying to build something on a scale to dwarf TSMC and Intel combined?

Market cap doesn't relate to how much it would cost to build and operate a leading edge fab facility ground up. TSMC doesn't recoup cost of a new fab technology line in one year nor do they build a new building for one every year either etc etc.

 

TSMC company value is a better thing to look at as that includes both cash assets and physical assets, their value being $16.75 trillion.

 

And from their Quarterly report:

Property, Plant and Equipment  $3,132,664

Total Assets  $5,484,556

 

TSMC also has long term existence benefit, much of the company and assets were built and obtained when things were cheaper and in places that were or are still cheaper. They will also not be trying to maximize certain things like land value on the books unless they have some future intention to sell, lower valuations can be better, can be worse too.

 

Either way to make another "TSMC" come in to existence today it will cost more than 5 trillion dollars, how much more depends on where you want to do it as well. So as crazy as it sounds 7 trillion dollars won't create a new fab company with the same capabilities and production capacities as TSMC.

 

What I don't agree with is needing to go this big from the get go. Guarantee actually getting chip fabrication going with be fraught with problems, delays and years of fine tuning so why waste 7 trillion on things that will be out of date by the time they are properly operational to their maximum capabilities.

[Sauron]

13 hours ago, ToboRobot said:

If you compare the expense to GDP it's comparable to the Manhattan project for the people who think it's unreasonable.

I'm too lazy to do the math. sorry.

The manhattan project was seen as vital for winning a potentially civilization ending war (not saying that was necessarily true, but it was the perception of the US government at the time).

 

This would do what, make the same stuff nvidia does but starting from scratch? I'm for having some competition in the space, but that money would be better invested in getting existing competitors up to speed, especially since right now the main hurdle is software and not hardware; CUDA is too widely used for others to make a dent in nvidia's market, even if the hardware were good enough.

[Needfuldoer]

I'll do it for half.

 

(And by "do it" I mean "make some custom labels for Nvidia L4s that cover up the 'Nvidia L4' branding".)

[ToboRobot]

58 minutes ago, Sauron said:

The manhattan project OpenAI was is seen as vital for winning a potentially civilization ending war against climate change (or Russia/China)

I am saying people are positioning this as an opportunity for a massive generational leap. 

[Sauron]

5 minutes ago, ToboRobot said:

I am saying people are positioning this as an opportunity for a massive generational leap. 

Maybe they're trying to bullshit you into believing that, but the premise is absurd. Like, actual lunacy.

 

We already have solutions to climate change, we just don't want to actually implement them because it would reduce powerful people's profits. Trying to "tech the climate change away" by funneling money into nebulous hypothetical solutions that delay any action taken by a factor of years or decades is part of the problem.

 

Russia and China have nukes. The US already has the military capability to easily beat both except entering a direct war with them means getting nuked. AI doesn't, wouldn't and can't ever change that situation. I also posit that if at all possible armed conflict should be avoided rather than looking for ways to win one before it's even on the table.