Is AI going to blow up the economy?

An article in FT today highlights the increasingly creative ways Big Tech are coming up with to finance their AI build-out while keeping the debt off their books.

instead of just issuing bonds off its own balance sheet, Meta formed a joint venture with Blue Owl called Beignet that would develop and own Hyperion. Meta owns just 20 per cent of Beignet, but made a rock-hard commitment to lease Hyperion for at least 20 years. That guarantee allowed Beignet to issue an amortising $27bn bond, but this debt doesn’t actually appear as debt on Meta’s balance sheet, even if it is on the hook for the payments.

Goldman Sachs ran the numbers and counted up a trillion dollars of purchase commitments which don’t appear on the balance sheets of Big Tech.

The sheer scale of AI borrowing is believed by some to impact US Government borrowing as Big Tech debt issuance competes with Government bond offerings potentially boosting interest rates at a time when the government is trying to keep interest rates low.

On top of that, AI investments plan to ‘pay off’ by replacing employees which potentially undermines future government tax receipts if employment decreases as well as potentially increasing unemployment benefit costs.

4 Likes

Well duh, Goldman, wasn’t this obvious from day one? And besides the reduced tax revenue and increased unemployment benefit spend, what about consumer spending that’ll crater? If I am caught on a bad mood moment, like I am now when reading this, I’d say Zuck et al don’t give a shit, this road seems to be the road to technofeudalism. Caper in gilded palaces and let them plebs eat cake.

3 Likes

Forever loans could be the next crises. But who knows.

https://caredge.com/guides/auto-loan-crisis-32-year-record

https://www.autoblog.com/news/americas-car-loan-crisis-is-getting-worse-and-lenders-dont-seem-worried

The WSJ has a nice article on this, too: Why Big Tech’s AI Spending Is $3 Trillion Higher Than It Seems

I’m not exposed to any of these companies but I am exposed to some (Life) insurance companies that buy the securitized debt of this shit. Started scaling down on those.

3 Likes

Very interesting. Since most of the lease of datacentres has not been paid for the 1st time yet, it’s not on the balance sheets. Since it’s not on the balance sheets, the terms of payment are not publicly known. No quarterly reports!

Anyway, the companies have rather large expenses in the pipeline as contracts to buy future services. Hopefully, there’s an exit ramp to get out of these contracts without too much damage.

Some years ago I consulted for one of those and one of the main questions was “what are the exit ramps of this contract?”. Hopefully, they are managing risk in a similar way today. I maybe wrong, but purchase commitments are just a loose contract where the software company accepts to buy 1x of services at X price, and reserves the option to buy until 100x at Y price. So, the datacentre developer takes this contract to a bank/investment fund and asks for money to build it based on some happy revenue estimates.

The 1.52 trillion USD purchases commitments number may look ugly, but the real liability for alphabet, MS, amazon, and meta may be smaller. Datacentre developers are already making their money, the ones who should worry is the people lending money to build the datacentres.

PS, I wrote hope many times in this reply :confused:

1 Like

All these shenanigens in the US contrast with China where AI development is largely state run and controlled. And rather than seeking quick returns from AI, their goals are technological self-reliance, domestic-chip production, boosting industrial productivity and national security.

And then there is Europe, which is now waking up. The EU is proposing up to 7 large AI gigafactories across the EU, with public/private funding of more than €30 billion for those facilities.

Bill Gates has thoughts:

https://www.gatesnotes.com/work/make-ai-work-for-everyone/reader/a-turbulent-ai-era-and-critical-choices-to-make?WT.mc_id=20260826_ai-overture-2026-med-med

Essentially, he is saying that “AI will either be the greatest equalizer ever invented, or the worst source of injustice” - and we are not prepared.

More on Gates’ essay from this article in the NYT, titled ‘Bill Gates Warns That AI Is More Dangerous Than Big Tech Admits’:

https://www.nytimes.com/2026/08/26/technology/bill-gates-ai-risks.html

1 Like

Right now my only question is: ‘buy, sell or hold’?

Schrödinger and quantum computing stocks to the rescue: in one quantum superimposition, you can hold all three states (or more) simultanously.

If unsure about what specific stock to buy for this, I’m sure there’s an ETF for that. Or nowadays probably more ETFs than stocks for this theme 



Slightly orthogonal: also just put this into my bucket list, easy to check off for me as a Zurich resident:

“A life-size cat figure in the garden of Huttenstrasse 9, Zurich, where Erwin Schrödinger lived from 1921 to 1926. Depending on the light conditions, the figure appears to be either a live cat or a dead one.”

(Source)

5 Likes

Thanks for this!

1 Like

How does the cat thing even work?

My AI gives the following explanation:

The “Dead” State: Because the sculpture consists of thin, cut metal layers, when you look at it from a sharp side angle or under specific dim, overcast lighting conditions, the flat perspective collapses. The cat silhouette completely vanishes or blends into the background shadows, making the cat “disappear” from sight—representing the unobserved, collapsed, or “dead” state.

3 Likes

SELL! sell! Sell!

I would like to see the money magically evaporating. The thing is that money always only moves, if there’s a dip, just wait

Don’t count on it. As long as China is a threat to US national security it will be a fight, oh sorry, an investment to the finish. Of course with tax payer $s.