23 Comments
User's avatar
Dart Driver's avatar

If I am not mistaken, Grant Williams recently interviewed Mike Green, who had a lot of information about index funds, the number of which out number the names that on the markets, and how this dynamic influences market action. The names these fund invest in are dictated by their charters. Strictly a mechanical reflex based on market cap. That was a brilliant thing about SPCX. It opened with a market cap that instantly put it on every index fund listed thus providing huge demand instantly.

So it is for AI names, all of whom occupy a large portion of the indexes. And who supplies the most capital for index funds? I suspect it is IRA's and other captive vehicles for retirement funds established by the usual brokerage houses. The participants pick from the available funds but not individual stocks and are going to be the bag holders when the Bubble finds its pin.

After the financial collapse of the builders of the AI infastructure (it will still be humming), I expect the usual suspects- government agencies included, will be ready to pick up the data centers for pennies on the dollar all the while claiming what generous and nice guys they are.

Lau Vegys's avatar

Thanks, Dart. Agreed on all of it. The index funds don’t decide to buy Nvidia; market cap decides for them, and the paycheck deductions arrive every two weeks whether the CAPE is 15 or 42. I actually wrote about exactly this back in June, once on the SpaceX listing (and the passive money as exit liquidity), and once on why your 401(k) has to own whatever the index owns (employer match and all). Links below if you haven’t caught them. On who ends up owning the data centers afterward, I suspect you’re right, and we’ll be told it was for our own good.

https://www.snafuinvesting.com/p/ai-ipo-mania-is-here-youre-the-exit

https://www.snafuinvesting.com/p/what-the-blackrock-ceo-just-told

Dart Driver's avatar

I went back and reread your article. At the time, I did not appreciate how devious the financing scheme is, and advantageous it is to Wall Street - everyone wins except people trying to build their retirement funds. The next owners of the data centers have probably been decided and as always in a bankruptcy, the stockholders are wiped out. All of it does not bode well for the future of individual freedom.

Thomas St.Yeng's avatar

good warning, don't get swept up in the euphoria of companies that "make" $10 billion by spending $50 billion. "Buy analog", which I mean companies that produce things and do stuff people want. Yes, AAPL, AMZN, TSLA, etc but only because they produce revenue from real things.

...and of course the small mining operations talked about here, because all that tech gadgetry requires rare earths, power, and hard currency to support it. Buying the dip (today) takes a unique long term perspective on the market.

Lau Vegys's avatar

Thanks, Thomas. “Buy analog” is a good way to put it. Every data center still needs the copper, the uranium and the rare earths, and nobody’s paying a hundred times earnings for the companies that dig them up. That’s the unglamorous end of the AI trade, and it’s where I’d rather be standing when the music stops.

frank hayes's avatar

At this point, all i can say is "you da' man!

thank you,

DFH

Lau Vegys's avatar

Appreciate it, Frank. Let’s see if the market agrees with me soon, too.

Ricardo Alvarez's avatar

100%100 de acuerdo con tu análisis,saludos desde Argentina

Lau Vegys's avatar

Gracias por leer, Ricardo. Saludos!

Jay Bremyer's avatar

Excellent warning and all true. I don't think all of the big incensuous tech companies will fail, or even go down significantly before rebounding, but they might, and certainlysome will crash all the way. So, it's pretty insane to not be very cautious in these frothy times. Thanks,again.

GlenB's avatar

And well diversified

Lau Vegys's avatar

Thanks, Jay. I agree with pretty much all of that.

Xavier Narutowicz's avatar

You take the complicated, much thrown around, make it simple and interesting with a personal appraisal.

Lau Vegys's avatar

Thank you, Xavier. That’s about the best compliment I could hope for.

Attila Rebak's avatar

I think CAPE may actually understate how extreme current valuations are. The unusually large fiscal deficit is supporting corporate profits and profit margins through the Kalecki-Levy mechanism, which mechanically boosts the “E” in valuation ratios. If you instead look at total market cap relative to GDP, which is less affected by elevated profit margins, today's valuation is meaningfully above the 2000 peak. So the comparison with 1999 may actually be too generous to today's market.

Lau Vegys's avatar

Good point, actually. Thanks, Attila!

Attila Rebak's avatar

Thanks, Lau! And I forgot to say in my first comment, what a great article. Very thought-provoking, as always.

Lau Vegys's avatar

Thank you, Attila. Which reminds me, I still haven’t gotten back to you on your earlier comment. I’ll do that as soon as I clear my desk.

John D's avatar

The CAPE answers how expensive shares are relative to a decade of CPI-adjusted reported earnings but it does not answer how expensive shares are relative to the supply of fiat money (i.e. fiat debasement). How can CAPE be a constant measurement unit when we have changes in discount rates, profit margins, accounting standards and index compositions?

Lau Vegys's avatar

Thanks, John. Well, CAPE isn’t a constant, which is why I called it the closest thing we have rather than the answer. On debasement, though, it mostly cancels: price and earnings are both in the same inflated dollars, so printing lifts the top and bottom of the ratio together. And when the price runs ahead anyway, well, that’s the bubble.

On discount rates, that was the case for a high CAPE when the 10-year was at 1%; at 5.2% it’s the case against. Margins actually make it worse, as Attila noted above: deficits are propping them up, which flatters the E. Accounting and index composition are the standard Siegel critiques, and I’d grant they take a few points off the raw number. But you don’t need CAPE to be perfect for 42 to be a problem. Market cap to GDP, which doesn’t use earnings at all, is above its 2000 peak.

Simon Stevenson's avatar

I think there is an issue with the fiber to AI comparison - fiber had a long lifespan, while the depreciation rates of chips (and therefore a key component of datas centres) is much higher.

Lau Vegys's avatar

Thanks, Simon. Fair point, though I’d argue it actually reinforces mine. Fiber laid in 1999 is still carrying traffic today, so the people who paid for it at the peak at least got a long-lived asset for their money. A GPU bought in 2024 is already two generations behind, and the big four are depreciating them over five or six years, which some would argue is generous. So the AI buildout has to earn its money back much faster than the fiber buildout did, and it’s starting from the same lopsided ratio of spending to revenue. Faster depreciation makes the problem worse, not better.