Is AI a bubble which will rival the dot-com and the great financial sub prime recession or even exceed it?

Lounge EndeavourOS pub
1

There are some who are screaming at the top of their lungs that AI is in a bubble right now. They have 3 main points which are
Firstly the yield curve has inverted and has remained inverted for quite some time now. The longer the yield curve has remained inverted, more deeper has the subsequent recession been.
Secondly the American economy has become one big bet on AI. With most of the foreign capital and even American capital betting on this. In fact the top 7-10 AI companies are only/solely responsible for all the stock market valuations going up. Other than these 7-10 companies all the other companies shares have been flat or have fallen. This is compounded by the so called circular investments by all of the big name AI companies, using SPVs. The Chinese Local governments would blush with envy in the way American SPVs are being used.
Finally none of the so called benefits of AI have come true or the AI companies have managed to turn a decent profit. This is reminiscent of the way dot-com bubble played out, sky high valuations.

On the other hand there are some very serious people lending their voices to caution. It is being said that this is not a bubble of such gigantic proportions.
For the yield curve inversion, that was typically followed by US Fed raising interest rates. Currently the reverse is being done. There is immense political pressure on US Fed, to decrease interest rates. US Fed is complying with the wishes of its elected representatives. And with US Treasury secretary already hinting on getting a new successor to the existing US Fed chairman, we can expect the rates to be slashed even more in 2026.
On the AI’s potential most of the silicon valley and other thought leaders have said that AI will change everything. Especially the white collar workforce. That has to be taken with a pinch of salt, as Yes it will but if it is genuine AI. Also companies which have invested in AI have not achieved the level of productivity gains that have been promised.

References.
Former IMF chief economist Gita Gopinath’s warning for rich world economies - The Economist.

The Most Dangerous Yield Curve Inversion in History.

Inside AI’s Circular Economy: Geopolitical Loopholes, Hidden Debt, and Financial Engineering.

Three Charts That Help Explain What’s Behind the AI Bubble Fears - Wall Street Journal.

How The S&P 500 Quietly Became An AI Fund - CNBC

Bessent warns ‘sections of the economy’ could go into recession if Fed refuses to lower rates - CNN

U.S. Treasury Sec. Bessent calls for internal Fed review, possible loss of regulatory function - CNBC (This is important as US Federal Treasury Secreatary has questioned the models that US Fed uses.)

2 Likes
2

Yes, fools and their money…

1 Like
3

The behaviours of the people involved definitely remind me of the dot com bubble. Whether or not this is a bubble, I’m not qualified to say but I would guess yes.

2 Likes
4

I believe so as well. I’m not an expert in finance just a enthusiast computer tech / analyst. I haven’t seen a AI that is really good at the moment. Or anything suggesting having an AI working well. Anyway we call them AI but in reality most of them are LLM that are bad when they are not connected to the internet. I have played with LMStudio (and still do sometimes) and some of the models and they are often wrong or plainly bad. Some are better than others. People need some kind of analytic brain to second guess what the LLM is spewing because of this and most of them don’t.

Those big companies are in a vicious circle of investing in themselves. Links that Achie1 gave explains that.

That said I still use chatgpt on daily basis simply because it is faster most of the time than doing a simple web engine search which are also becoming really bad. I could also tell long stories of me arguing with chatgpt about bad answers and trying to make it spew the proper one or me having it make something and ask it to correct something for it to inject the error again and again after further changes.

Well that’s my 2 cents on this anyway.

1 Like
5

You (the buyer) create the bubble. When you are buying a product for yourself, anything, if it is not of urgent need, you wait for its price to go down right? You wait for the discount days etc… Stock market is (should be) exactly the same. Everybody wants to get in on the action. Everybody wants a seat on that train. But people don’t realize that the train has already left the station, when they even considered the idea of taking the train. So i would say instead of the bubble analogy, train is a better way to describe it. Because, i am pretty sure some people (the early birds) made a lot of money during the dot.com event, because they could take that train and “get off” at the right stop.

Here is the thing, right now, “nvidia” is the only chip “designer” (not manufacturer) which holds the patents to the most powerful GPUs in the world. Their stock price has increased exponentially since AI became popular, because AI needs gpu power, not cpu power. If you had the nvidia stock “before” the AI days, congrats, you are currently riding that train. But, if you are still scratching your head thinking “man that nvidia is going, i should get a ticket”, well, good luck to you. I wouldn’t do it. But i am no stock market expert you know. Also, if you are riding that train, you should know that it will eventually stop. Let’s say some other chip designer came up with a better chip design. Here’s what’s gonna happen:

image
image514×600 164 KB

1 Like
6

Having lived through the dot-com/bomb there are some big differences between then and now. One of the big differences is retail investors were putting money on anything with dot-com name that IPO’d. You don’t seen AI IPO’s at the rate we did in 1998/99. The money is with the big boys and many of those have been around a long time. They are not taking VC money for fancy office spaces with pool tables and refrigerators full of beer, and yes, I worked for those companies in the late-90s

Back in 1998, I worked for a start-up going for Round B when the bottom dropped out. The VC that invested in our A-round had me go to his other companies he invested in, under the guise of cross-selling, and evaluate their business. I quickly realized this guy had investing in companies that had no real IP or business model. In the end he funded the company I was in as I told him it was the only one with IP that was worth something. I left the company before round-B and that company still exists.

This seems different. Instead of a 10,000 mom/pop startups doing AI and sucking up everyone’s VC this is the big boys in a full on AI race. The circular investing has me concerned and the promises of X amount of dollars spend in coming years. Meta stock punished by investors for it’s capex spend after it just blew through a lot of capex spend on VR. So in that regard investors are being cautious. PLTR is selling at over 400x times earnings which is insane. The forward P/E is almost 200 times earnings. The sell off there started this week as the stock price doesn’t justify the earnings. That doesn’t mean PLTR isn’t a good company but too many retail investors got happy feet on this one. Everyone needs memory an Micron (MU) in comparison as 31 P/E and a forward P/E of 12.

So are valuations high, absolutely. Will this fizzle like Metaverse? Likely not. The market is fragile and as soon as the big boys stop or slow down the capex spending it will drop like a rock. The big question is when that will happen? Doesn’t seem like it will slow down for the next year or two at least. Maybe if they cannot show ROI on commercial AI the pull back will begin. Hard to time the market and the market will always prove you wrong.

2 Likes
7

For “value investing” (aka Berkshire Hathaway etc.), a price-earnings ratio about 12 is already idiotic, because it would take a 13th year, say, in 2026, to earn back the first $1 from buying one dollar of a stock twelve years earlier, say, in 2013. Twelve years are very close to a business cycle. For comparisons, venture capitalists expect from all the many dozens of bets they make, to have a return of $10 after 10 years, say, in 2036, for very dollar spent, in 2025/26. Modesty for VCs imply, they’re also already very happy with $5, but not with $1, to compensate the losses and have an extra stash for the next bets.

If T. Nangle, Is it really a bubble? (2025). https://www.ft.com/content/e65579d3-f513-44f4-91e0-246fefe66e4c re-calculations are correct, the stock market expects their first $1 in 25 years to 40 years, or $25 to $40 next year.

The only way to make money, right now, is to go short—if and only if you can fight off the psychological stresses, being short implies you’re a social outcast and have sleepless nights, you’re likely to end up in surgery, because you’re being wrecked from betting on society to go down in hell.

Meanwhile, … the Chinese mainland, for the first time in its recent economic history, fully relies on Chinese graduates, who never went abroad, no Silicon Valley alumni, to make the same work that Open AI et al. are doing, but for much less, in way less time, and then release such models as free to copy, almost freeware. (Licensing agreements stipulate, only if the service has several hundreds of thousands users, fees would apply.)

Cautionary note: Price-earnings (P/E) ratio talk is a way to impress people on cocktail parties and has zero information value for having an opinion about the stock market or a particular company. NEVER rely on price-earnings ratios.

All the above is banter!!!

3 Likes
8

I suppose it is the big boys this time, rather than everybody with a shitty idea. It’s still an artificially inflated bubble as far as I can tell and they usually pop at some time.

1 Like
9

It is not clear what ROI investors are looking for in these AI circular investments. Are they expecting $1 in 2026 or $1 in 2030 is not know. If the expectation is return of $1 in 2030 for money that is invested in 2025 then the investors have deep funding pockets. By deep funding pockets it means they do not have to show a return or expect a return for 3-5 years. Anything beyond that will reek of sovereign or public money and not private money.

The issue is that in this circular AI mania only partial money is put in by AI or AI related companies. So for example if Nvidia puts in say 20 billion dollars into a AI company out of this 20 billion dollars only 10-12 billion dollars is from Nvidia in terms of cash or stock. The remaining 8-10 billion comes from other financial firms. I wonder what is the calculus of these other companies. Is there some collateral at work over here for these other financial firms?

Further the use of SPV’s is to put risks and debts away from balance sheet. In 2008 post Lehman collapse, AIG collapsed too. AIG collapse lead to the financial markets freezing. No one was willing to lend to another. So even banks and financial institutions which were not directly or indirectly exposed to Lehman and AIG got caught up. This is what is known as contagion. How far is this contagion in this circular AI mania? Where will the ripples go? Expect average joe to get impacted by this.

IMHO VC should not expect most of their bets to play out. Realistically a VC should expect about 20% or 30% or 2 in 10 or 3 in 10 companies that they invest in to make money. But are VC’s investing in these AI circular investments? I doubt that VCs have a pool to make such huge bets, in tens or hundreds of billions of dollars.

Prior to the existing AI circular investments, startups typically used to have 10 rounds of investments, like Series 1 to Series 10. Not sure after this AI circular investment mania ends whether this will remain or not.

@cc_spicuous why do you say P/E ratio has zero information value? P/E, P/B and other ratios are used by many investors in determining the investments to be made.

@AceFour a forward P/E of 200-400 is definitely a bubble. Typically anything beyond 20-24 P/E is considered as speculative investment or a bubble. Palintair having a forward P/E or 400 is insane. Does it make so much of money?

2 Likes
10

wow threads like these make me happy for the encyclopedic amt of anecdotes and facts. Invaluable.

2 cents and to answer the title question:

  1. once they find out what it is absolutely useless for (education, imitating writing, advice; politics; anything authentic or meaningful) it will POP

  2. once they find out what it is really good for (automation, crunching numbers) it will be concentrated and focused towards that.

they will sell you the bulls*** that it is ideal for #1 and there will always be a legion of believers to believe it because the news told them…the proof will be in #2 and that’s where the money will go as far as investment capital as several industries will utilize AI’s strengths.

Still…there’s a whole lotta rogue AI behavior that has been on record…so all bets off.

11

To discourage confidence and to instill doubts in forum members, when reading the press, like Financial Times that authoritatively writes about P/E and P/B. I don’t want forum members to get lured into psychological pitfalls that come with money issues, the stock market etc., these seemingly objective, clean, sterile numbers have a particular spell on the minds of nerds and engineering types, people take authoritative brands in stock market information, like Bloomberg et al., as assurance when they make P/E and P/B quite prominent on company sheets. Here are shiny LaTeX equations:

{A7C7EBB6-86D1-4205-9F62-9571411B4EE3}
{A7C7EBB6-86D1-4205-9F62-9571411B4EE3}318×465 6.22 KB

Example: My $1 lemonade stand (book value, B = $1) can be worth “a horse and a kingdom” (price, P = $ ???) in a desert. And my $10k diamond ring (book value, B=$10,000) can be worth nothing (price, P=$0) in a desert. The Saudis are digging for drinking water for as long as they have lived there, only to find oil and gas. The Inkas and Aztecs sat on silver and gold, worth nothing (price, P = 0), where as the Portuguese and Europeans (and Chinese) value the silver and gold quite differently. Let’s not talk about the psychology in India when it comes to gold.

The price of an object is not inherent in the object (or service).

We—a cult, ideology or worldview—do not believe in the “the market is efficient” theory, in which all required and necessary information is in the market price (P).

Nevertheless, it is without doubt and fully correct that computer programs can run investment funds solely on the basis of these ratios (as well as market valuations = outstanding shares x stock price)—and even beat the market! (I withhold available citations, here, on purpose.)

They, therefore, have information value, plenty of it, which is contrary to my statement.

2 Likes