Let's Know Things podcast

Circular Finance

0:00
16:02
15 Sekunden vorwärts
15 Sekunden vorwärts

This week we talk about entanglements, monopolies, and illusory money.

We also discuss electrification, LLMs, and data centers.

Recommended Book: The Extinction of Experience by Christine Rosen

Transcript

One of the big claims about artificial intelligence technologies, including but not limited to LLM-based generative AI tech, like ChatGPT, Claude, and Gemini, is that they will serve as universal amplifiers.

Electricity is another universal amplifier, in that electrifying systems allows you to get a lot more from pretty much every single thing you do, while also allowing for the creation of entirely new systems.

Cooking things in the kitchen? Much easier with electricity. Producing things on an assembly line? The introduction of electricity allows you to introduce all sorts of robotics, measuring tools, and safety measures that would not have otherwise been available, and all of these things make the entire process safer, cheaper, and a heck of a lot more effective and efficient.

The prime argument behind many sky-high AI company valuations, then, is that if these things evolve in the way they could evolve, becoming increasingly capable and versatile and cheap, cooking could become even easier, manufacturing could become still faster, cheaper, and safer, and every other aspect of society and the economy would see similar gains.

If you’re the people making AI, if you own these tools, or a share of the income derived from them, that’s a potentially huge pot of money: a big return on your investment. People make fortunes off far more focused, less-impactful companies and technologies all the time, and being able to create the next big thing in not just one space, but every space? Every aspect of everything, potentially? That’s like owning a share of electricity, and making money every time anyone uses electricity for anything.

Through that lens, the big boom in both use of and investment in AI technologies maybe shouldn’t be so surprising. This represents a potentially generational sea-change in how everything works, what the economy looks like, maybe even how governments are run, militaries fight, and so on. If you can throw money into the mix, why wouldn’t you? And if that’s the case, the billions upon billions of dollars sloshing around in this corner of the tech world make a lot of sense; it may be curious that there’s not even more money being invested.

Belief in that promise is not universal, however.

A lot of people see these technologies not as the next electricity, but maybe the next smartphone, or perhaps the next SUV.

Smartphones changed a whole lot about society too, but they’re hardly the same groundbreaking, omni-powerful upgrade that electricity represents.

SUVs, too, flogged sales for flailing car companies, boosting their revenues at a moment in which they desperately needed to sell more vehicles to survive. But they were just another, more popular model of what already came before. There’s a chance AI will be similar to that: better software than came before, for some people’s use-cases—but not revolutionary, not groundbreaking even on the scale of pocketable phone-computers.

What I’d like to talk about today are the peculiar economics that seem to be playing a role in the AI boom, and why many analysts and financial experts are eyeballing these economics warily, worrying about what they maybe represent, and possibly portend.

The term ‘exuberance,’ in the context of markets, refers to an excitement among investors—sometimes professional investors, sometimes casual investors, sometimes both—about a particular company, technology, or financial product type.

The surge in interest and investment in cryptoassets during the height of the COVID-19 pandemic, for instance, including offshoot products like NFTs, was seemingly caused by a period of exuberance, sparked by the novelty of the product, the riches a few lucky insiders made off these products, and the desire by many people—pros and consumer-grade investors—to get in on that action, at a moment in which there wasn’t as much to do in the world as usual.

Likewise, the gobs of money plowed into early internet companies, and the money thrown at companies laying fiberoptic cable for the presumed boom in internet customers, were, in retrospect, at least partly the consequence of irrational exuberance.

In some cases these investors were just too early, as was the case with those cable-laying companies—the majority of them going out of business after blowing through a spectacular amount of money in a short period of time, and not finding enough paying customers to fund all that expansion—in others it was the result of sky-high valuations that were based on little beyond the exuberance of investors who probably should have known better, but who couldn’t get past their fear of missing out on the next big thing.

In that latter case, that flow of money into early dotcom startups did fund a few winners that survived the eventual bursting of that bubble, but the majority of companies tagged with those massive valuations went out of business in part because their valuations were based in part on optimism, hot air, and illusory financials.

Which is to say, their financials were based on a lot of money being added to their account sheets and tallied in the places investors would see those numbers, but the numbers didn’t mean what most people thought they meant.

A company could receive tens of millions of dollars in orders, for instance, but that money and those orders might never be received and fulfilled, or that money might be mostly illusory: maybe it was borrowed from another company to spend on advertising, and that money would then go right back out the door, to the company from which it was borrowed, to pay for their ad services.

That kind of arrangement could be beneficial, as the company doing the borrowing might give up a relatively small number of shares in exchange for money, which looks good on its balance sheet, especially if the money is given at a high valuation, even if that money was mostly just a loan from a company providing ad services, with the full knowledge that money would then be spent on their own ad services. And the ad company giving the money could usually afford to buy in at a high valuation, because it knows it will get that money right back, and when it does, it will get to record that money as income on its own balance sheets.

So Company A gets millions of dollars from Company B, that money is then paid to Company B for some type of service, and both companies get to record favorable figures on their accounting sheets, as if real sales took place and real outside money changed hands, despite it being a circular move, with very little or no actual value being created.

These sorts of relationships are also often good for investors in companies that do this sort of thing, because it makes their investments, the companies they’ve bought into, look even more valuable.

Check it out, Company A, which I own shares in, is worth more than it was last month because of all the business it’s conducting, and because this other company bought into it at a higher price per share than I paid! Even though that increase in valuation is predicated on circular financing, the numbers still go up, and they go up for everyone involved, so there’s little reason to crack down on this not illegal, but shady behavior, and even less reason to want anyone else to know about it, because then they might not add their own money to the circular money-cycling, number-increasing machine.

The major concern amongst some analysts right now is that the AI boom, especially in the United States, might be essentially this kind of circular cycle, but much larger than previous versions of the same.

In the US right now, investment in AI infrastructure like data centers accounts for a huge portion of overall growth—the numbers vary, depending on who you ask and what numbers they look at, but some say that about 90% of total US economic growth, and around 80% of US stock market growth, are predicated on these sorts of investments this past year. Without these investments, the US economy would be basically flat, or worse, and the US stock market would be flailing as well.

This situation isn’t ideal whatever the specifics, as too much reliance on just one industry, or one small collection of industries dominated by just a handful of companies and their investors, makes for a precarious financial foundation.

If anything goes wrong with just one company, the whole house of cards could collapse. And if anything goes wrong with the industry, things could get even worse, and fast. All that investment, all that construction, all those employees and all that money sloshing around could disappear, could stop being spent, could make all those numbers fall and fall and fall more or less overnight.

If this industry is in fact in a bubble, and if it’s being propped up by this kind of circular financing, where companies are fluffing up their own and each other’s accounting books by rotating the same bundle of money and on-paper money from company to company to company, that would portend pretty bad things for the US economy and market, if anyone involved stumbles, even just a little.

This is why recent deals between the biggest players in this space are raising so many eyebrows, and causing so much sweat to bead on so many foreheads.

In September of 2025, ChatGPT-maker OpenAI announced it had formalized a $100 billion investment deal with AI chipmaker Nvidia, the latter expanding on its existing investment in the former. In October, OpenAI announced it was purchasing billions of dollars worth of AI hardware from Nvidia-rival AMD, and that it’s taking a 10% stake in the company.

Microsoft is already heavily invested in OpenAI, to the tune of $13 billion; it takes 49% of OpenAI’s profits, and gets more than that until its original investment is paid back. Microsoft also accounted for nearly 20% of Nvidia’s annualized revenue, as of the fourth quarter of 2025.

Oracle, another computing company which has become hugely influential in this space due to its investment in cloud-based AI datacenters, has a $300 billion deal with OpenAI for future infrastructure buildouts and access, and OpenAI’s Stargate datacenter project was co-funded by Oracle and SoftBank. Nvidia also owns part of CoreWeave, which is an AI infrastructure supplier for OpenAI, and which has Microsoft as a massively important customer.

All of which is very…tangly. It’s an interconnected mess, and OpenAI and Nvidia are at the center of it, but there are a lot of weak spots, threads that, if pulled, would cause the whole thing to unravel. Which is why this feels like such a dangerous setup to many analysts right now.

Consider that in 2025 alone, OpenAI has made around $1 trillion-worth of AI deals. A lot of these deals are plans to invest: commitments to buy data center construction or the use of data center bandwidth, or they’re financial ties with competitors, clients, and providers—companies that would otherwise be competing with, selling to, and buying from each other, rather than linking arms and creating financial and infrastructural interdependencies.

Many of these deals are predicated on debt and what are generally considered to be over-inflated IPO valuations, too: money that isn’t money in the traditional, accounting-book sense, in other words. Numbers that make activity, use, and income for these companies look a lot bigger than they concretely are, on balance sheets, which in turn helps their investment numbers go up up up.

This dynamic has become overt enough that many of the biggest investors in AI companies, and the heads of said companies, like Sam Altman of OpenAI, have said, outright, that it’s probably a bubble, and that a lot of companies will probably go under in the relatively near future. No one knows when, but it’s a good thing, they’re fond of saying, because that shakeout will kill off the deadweight, allow the survivors to scoop up their former competitors’ assets at fire sale prices, and the whole industry will be further centralized around just a handful of the best and the most impactful, just like in the post-dotcom years. Monopolies and mini-monopolies, which, for the people creating and profiting from those monopolies, at least, seems like a good thing.

That optimism glosses over what those in-between years look like, though, especially for smaller investors, employees who are laid off, en masse, and the folks who aren’t profiting directly from the surviving business entities, and who see their stock portfolios collapse and overall growth in their country decrease.

Most of the stories in the tech world right now in some way tie back to the promise and concerns surrounding AI. It’s become such a big story because there’s a chance it will be the next electricity, but there’s also a chance the warning signs we’re seeing are real, and things will get a lot worse before they maybe, possibly, for some people, at some point, get better.

Show Notes

https://finance.yahoo.com/news/a-20-billion-clock-is-ticking-for-openai-as-microsoft-talks-turn-fractious-130006071.html

https://www.sfgate.com/tech/article/circular-deals-bay-area-tech-21089538.php

https://www.theguardian.com/business/2025/oct/08/openai-multibillion-dollar-deals-exuberance-circular-nvidia-amd

https://www.ft.com/content/950e3a36-7141-4426-b7c5-08fad5d83919

https://finance.yahoo.com/news/very-troubling-ais-self-investment-spree-sets-off-bubble-alarms-on-wall-street-160524518.html

https://www.cnbc.com/2025/10/15/a-guide-to-1-trillion-worth-of-ai-deals-between-openai-nvidia.html

https://insights.som.yale.edu/insights/this-is-how-the-ai-bubble-bursts

https://www.bbc.com/news/articles/cz69qy760weo

https://www.nbcnews.com/business/economy/openai-nvidia-amd-deals-risks-rcna234806

https://www.bloomberg.com/news/articles/2025-10-08/the-circular-openai-nvidia-and-amd-deals-raising-fears-of-a-new-tech-bubble

https://flowingdata.com/2025/10/13/circular-deals-among-ai-companies/

https://www.nytimes.com/2025/10/07/business/dealbook/openai-nvidia-amd-investments-circular.html

https://sherwood.news/markets/analyst-a-lot-more-disclosure-needed-on-these-circular-ai-deals/

https://www.barrons.com/articles/nvidia-microsoft-openai-circular-financing-ai-bubble-5d9a4e7c

https://www.investopedia.com/wall-street-analysts-ai-bubble-stock-market-11826943

https://www.goldmansachs.com/insights/articles/ai-may-start-to-boost-us-gdp-in-2027

https://finance.yahoo.com/news/most-us-growth-now-rides-213011552.html



This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit letsknowthings.substack.com/subscribe

Weitere Episoden von „Let's Know Things“