Big Tech Needs $2T in AI Revenue by 2030 | Hacker News Reader

Big Tech Needs $2T in AI Revenue by 2030

wheresyoured.at

"Jensen Huang claims — without any real breakdown as to who is buying them — that NVIDIA has over $500bn in bookings for its AI chips, with little worry about whether there’s enough money to actually pay for all of those GPUs or, more operatively, whether anybody plugging them in is making any profits off of them."

Does NVIDIA care if the company that bought their chips cannot make money using their chips? As long as the company's money is transferred into NVIDIA's accounts, that's all that matters to NVIDIA.

Unless it was NVIDIA that gave said company the money to buy said chips. The “money” transferring into NVIDIA’s accounts in many of these cases is NVIDIA’s money being used to buy its own chips.

Those sort of circular deals, at the scale they’re happening, is one of the things that’s freaking folks out at the moment.

Out of interest, do you have any specific examples of this?

Nvidia “invested” up to $100B in OpenAI. OpenAI uses much/most of that money to buy NVidia chips.

The MS Azure deal is of the same nature. MS “invests” into OpenAI who then buys Azure services for the investment. OpenAI evaluation is increased and the musical chairs will continue for another round.

Investors assume that people will continue to buy their products, so yes presumably they do care that nvidia’s customers do not go bankrupt.

That’s before even getting into the fact that many of these gpu purchases were partially funded by equity.

Investors, yes. The leadership, highly motivated to show massive revenue growth, not so much. It’s nothing new, telco industry has seen these mechanisms before, for example

They know they aren't going to get it, the goal however is to be the one that does get their share of it back.

If OpenAI gets their $500B back, they don't care that everyone else lost $1.5T.

This is not true because of how money works. Take mortgages. You might think that if the bank borrows you $100 to buy a house, it is now dependent on you paying back $100. In fact the reverse is true. Fractional reserve banking, you can find plenty of explanations that this "creates" money, that the banks can use or pay out to investors.

The wealth of a small set of banks is essentially the total money that's lent out. The wealth of a set of banks PLUS the government literally is the total money lent out. In other words: if houses start being owned "naked" (meaning no mortgage), either through bankruptcy or just paying off the loan, that's when banks

So let's say you have a circular loan. Nvidia -> MS -> OpenAI -> Nvidia. You have just created money (since it won't ever be paid back), and what you're dependent on is the valuation of the companies in the chain going up always, there is some leeway of course but it can't really drop.

If the money keeps circulating, the money "rotating" effectively becomes money these companies can spend (whether on stock buybacks or extra chips or management raises or ...) You hopefully also see that you or I, or anyone external cannot cause this situation to collapse, only the companies in the chain can (but have extreme incentives not to). The system collapses, of course, if one of these companies goes bankrupt, but until then there's nothing that can stop it.

And because of this it will make those companies "Too big to fail".

In essence, this is what the investments are meant to accomplish ...alignment of interests.

The amount of Big Tech cash flow being diverted into capital investment coupled with the lack of viable business models to actually fund that outflow of cash is outright scary.

A lot of parallels here to infrastructure buildouts leading up to the .com implosion. A lot of dark fiber got dusted off later and used, but unclear if the present buildout of GPUs will have similar decades-long utility.

When did AI become a thing? After it turned out blockchain is not going to change the world after all and the price of a gpu operation dropped dramatically as infrastructure became disused. Something else, currently prohibitively expensive, will drop below a viability line and see massive uptake and improvement.

This is a lengthy introduction. The complete post requires a subscription.

Just a caution, this guy is a Gary Marcus-like perennial chicken little, this isn’t some discovery or realization, it’s just an ongoing part of someone’s identity as an AI bear.

This seems like an ad hominem attack, are his numbers incorrect?

Most criticism of Zitron tends to boil down to either “he’s mean/vulgar” or “he has an opinion”. Recently I saw an article that implied that Ed Zitron, owner of Ed Zitron PR and author of a couple books about public relations, was hiding the fact that he’s a PR guy.

Off the top of my head I think that he might’ve gotten inference costs wrong in an article and then corrected it, but usually I don’t see many people poking holes in his numbers.

There's a pessimistic slant on them like he says:

>By the end of the year, Microsoft, Amazon, Google and Meta will have spent over $400bn in capital expenditures, much of it focused on building AI infrastructure, on top of $228.4bn in capital expenditures in 2024

but doesn't mention those companies made over $320bn profit in 2024 so they are not really going bust over this stuff.

This is the same guy who did the totally unhinged article (with even more unhinged headline) a while back.

https://www.wheresyoured.at/the-men-who-killed-google/

You can see how HN ate up that article back in the day - https://news.ycombinator.com/item?id=40133976

I read that article when it came out and thought it was well written and had good evidence and sources. It's the opposite of unhinged IMO, and explains a lot as to why Google is as terrible as it is now.

terrible.... but profitable.

Yeah, total looney. I had a collection of questions and retorts until I got bored with it.

"During a goldrush, sell shovels"

Reply on news.ycombinator.com