These companies are spending more money than budgets of many countries enough to add 2+% to the US GDP so the amount of loss for if it comes all crashing down will be huge.
the market cap of a company is computed by the current price of a company's shares, the last price paid; not all the shares of the company were bought at that price, the ones who got shares cheaper are showing paper profits, unrealized. Those who have already cashed out have money in their bank accounts that was transferred from people who wanted to get in. If the company goes bankrupt, their shares will be worthless, but the money they paid for them still remains in the accounts of people who sold their shares: the money was not lost even if some people lost money.
I'm not going to keep going through it but the reason it works to value things the way we do is that the values are comparable and they frequently work out, so snapshots of the economy and the participants are comparable. But "losses" are not like taking gold and feeding it into some deep fold in the earth where it will disappear into the molten middle of earth.
Stock valuations are "expectations for the future". Those expectations weren't money, they were lottery tickes where the lottery consisted of human creativity and human effort. People buying and selling share are moving real money around to trade the expectations. The money didn't go anywhere, it's still there, it's just that expectations for the future have been reduced. It all boils down to humans trading some of their time and potential on a bet that things work out. Some people's effort gets more rewarded than others. Not every team wins the world cup, but people like to play and like to watch.
If they fail then the negative impact ripples through the economy due to misallocation of resources.
consider all the companies in a market and those that feed that market to be one virtual mega company, add up all the valuations and revenue streams, costs, etc and aggregate all the investors into one. Nothing changes about the picture I drew. We simplify models to make the real world understandable.
>negative impact ripples through the economy due to misallocation of resources
free or relatively free financial markets are the only way, the best way, the ne plus ultra of ways we know to allocate capital, we have no better way than for the owner of the capital and the reapers of the loss or reward to make a considered opinion that is risk "impedance" matched. By definition, the market does not "misallocate" capital, it optimally allocates it.
your theory is that we could somehow know the future, but that's a fallacy.
Free market efficiency is inherently tied to having multiple companies. Treating the entire economy as a single company gives nonsensical results because it fundamentally differs from what actually occurs. You might as well compare the economy to a game of tick tack toe, inherent complexity isn’t something you can simplify it has meaningful consequences.
Your ideas like many other ideas are simply wrong.
> could somehow know the future
Perfect accuracy isn’t the only possibility here, there’s levels of error.
Our system involves intermediaries between the actual owners of capital and the allocation of that capital who have very different incentives. When the worst possibility is missing a bonus there’s little difference between losing 10% of an investors money and 100%. That results in inefficiency through the misalignment of incentives.
That is actually true, and thus there’s no way to gloss over that truth without simply being wrong.
trust me bub, I've studied much more econ than you. If a competitive market sets the prices (check, that's what is happening), and you want to analyze statistics of a sector (check, that's what we are doing), you can take those competitive prices as "given" and hold them constant, and consolidate the assets of in industry into one virtual entity. No claims were being made about competition, the claim is that "it is validate to consolidate statistic of what you are trying to study.
"how much did the AI sector make last year? how much will it make next year?" is not answered by running a simulation of competitive marketplace with production functions.
>>could somehow know the future
>Perfect accuracy isn’t the only possibility here, there’s levels of error.
if you deviate from the market's prediction of the future, you are increasing your levels of error; why do that?
Then try and justify why you say shit this clueless:
> how much will it make next year?" is not answered by running a simulation of competitive marketplace with production functions.
Profits next year very much depend on the number of companies involved 1 vs 100 is not going to give the same results. Like I hope you realize how false what you just said was. Because if you actually believe this there’s literally no point in talking with you.
in a competitive marketplace, economic profit will go to zero. so whether an AI company buy or rents outside infrastructure, or builds it itself doesn't matter, it makes no difference. Therefore, if your argument is "outside infrastructure X", you can see the meaning of that by looking at "assume the company bought up the whole industry including outside infrastructure, then go back and look what I said and it still applies" A company monopolizing outside infrastructure for its own use would not abuse its monopoly against itself, but even if it did, makes no difference the extra profit and extra loss would balance out. Would it abuse its monopoly against downstream customers? if we use the existing market prices unchanged in our example, that is analyzing the case where it does not, which is the case that is comparable to the current situation.
or to put it another way, let's say these are all publicly traded companies competing. What if I told you "hey, i've investigated the ownership of all these public shares, guess what, Elon Musk owns them all, he owns every share of every company in AI, and all the infrastructure suppliers. Does that change the analysis from what we see in the marketplace? no, it doesn't. You want to draw a bigger circle around more affected parties, the suppliers to the infrastructure suppliers: OK, Elon owns those too it turns out.
nobody is analyzing the future here, we're talking about the case of AI going bust, not trying to predict AI going bust.
if were were going to project the future, we still would not do it with a simulation using functions to model companies to try to come up with meaningful profit numbers, we would project profits (and costs and revenues) based on margins of similar industries
Lots of words to say you don’t understand what you’re talking about.
At the most simple level monopolies extract profits through raising prices above that of a competitive market. This price increase reduces total sales even as it drives up profits.
As such the existence or non existence of a monopoly would change how much energy/etc AI was consuming among a host of other effects completely independent of how much utility it provided.
> nobody is analyzing the future here
That’s genuinely funny.
if you want to sound like you know what you are talking about, don't talk about "total sales", you want to talk about quantity demanded at the market clearing price vs the monopoly price, and the effect that has on producer surplus.
and what is bad about a monopoly in economic terms is not that they extract a higher price and profit from their smaller number of customers (those customer choose to purchase the product because it's worth it to them), it's the dead weight loss which represents unmet demand which slows the economy overall
We can’t look at next year’s power bill today. Hell finding their power bills from 1930 isn’t trivial either, thus we model the universe and test those models rather than just making up nonsense models that look pretty.
> that's the impact they have on the power grid.
That’s only the direct effect, the indirect effects get way more complicated.
The fact you’re constantly demonstrating profound ignorance is why I am treating you like a 5 year old. You understood what I said and had no defense, thus my use of simple terms was entirely justified. Using more words to say exactly what I just said doesn’t change my opinion. Instead try and apply that line of reasoning to your earlier statements and find the issue, that would demonstrate some actual understanding.
PS: The economic harm from monopolies extends to R&D etc but …
Let me guess, you sit down next to Kobe Bryant and start by saying you're going to tell him about winning basketball?
you tried to dismiss me by saying "oh but you're doing well" as if that meant anything. You brought it up, not me, but inasmuch as it does means anything, it suggests I'm winning the race that you purport to be an expert at.
I do not come from wealth, my family is largely working class. I have grown my wealth dramatically because I understand how the market works. I didn't know a priori what would happen, I just took what they taught me in school and applied it with extreme discipline and without fear. Turns out that works.
>The equity of the economy is not very similar at all to a game
the economy is about efficiency, and supply meeting demand, and fair exchange of factors and products for pareto optimality. that is what equity should mean but it's not what you mean by it. Your equity lifts only some boats and at the cost of lowering and even sinking others. Nobody can prove except by simple observation that your equity does not in fact lift boats.
You assume too much, that I am going to argue for centrally planned economies or something. I never claimed or implied I was an expert, or to what degree I’m “winning the race” (what a horrible way to think about human society!). I think it’s either an absurd failure of imagination or simple invested ideology, that we have to have either hardcore “free” markets (free for who?) or strict Soviet-style planning (typically with the assumption that we have only the knowledge and technology from that period too, for some reason). I think we can do a lot better than both.
Your impressive-sounding words about efficiency quickly fall apart for anyone who has actually looked at the dirty end of capitalist processes. Inefficiencies abound; the market optimises for only money which a lot of the time is a stupidly poor abstraction of the stuff of life that actually matters. And that abstraction enables and justifies untold cruelty and exploitation.
If you were a sort of capitalist-pessimist, saying that you didn’t like it but this seemed to be the least-worst option, I’d think you were woefully unambitious, but at least some way understandable. But you arrogantly defend this system, and brag about how your massive brain managed to exploit it. Welcome to HN, I guess.
so, according to you, you are not arrogant, you just legitimately know what's best for everybody else, and you are virtuous to boot?
writing tip: take all the emotionally charged words out of your prose, they don't have the effect you think they do.
I won’t be taking “writing tips” from you; my words are an expression, not pure calculated rhetoric. I never thought I was going to change your mind, I wanted you to face a teeny tiny bit of resistance.
Expressing an opinion and making observations is not arrogance. I said “I think” and “I’m against”, I never claimed any virtue, I didn’t brag, and I didn’t tell you that you don’t know what you’re talking about. According to me, I just have a view about part of the way the world works, that’s it. Reading tip: get better at it.
And I think we passed the threshold for crash down for AI, even if AI companies wont be that profitable. Nvidia/cloud providers will be profitable as long as there is demand for AI.
If AI companies aren't that profitable...then they're going to stop spending so much money on GPUs to train AI models. A gigantic amount of Nvidia's profits would go bust overnight.
Claude code and others are here to grow even if they don't do any further training.
The cost(and size) to train models is also increasing and is still 60% of the cards that Nvidia is selling. Losing 60% of your most profitable revenue stream I think would do bad for a company regardless of how much inference is increasing "dramatically"(all this means is the GPUs are dead sooner and the cost to do this massive inference increases too)