Who's Actually Paying for the AI Boom? A Look at the Money Going in Circles

Opinion · July 1, 2026 · 7 min read · By Elizabeta Dimoska

The AI buildout is the biggest infrastructure spending spree of our lifetimes. Hundreds of billions of dollars a year are pouring into data centers, chips, and power. The stocks tied to it have driven the entire market higher.

So it's worth asking a simple question that surprisingly few people stop to ask: where is all that money actually coming from — and where is it going?

When you trace the flows, you find something odd. A lot of it appears to be moving in a circle, among the same small group of companies. This has a name — "circular financing" — and in the past couple of weeks it went from a niche worry among analysts to a warning from the most establishment institution in global finance. Let's unpack it calmly, because it's genuinely important and genuinely misunderstood.

(As always: this is an opinion piece meant to help you think, not a prediction and not advice to buy or sell anything.)

The circle, in plain English

Here's a simplified version of the loop that has analysts talking:

• Nvidia agreed to invest as much as $100 billion in OpenAI to help fund its data-center buildout. OpenAI, in turn, committed to filling those data centers with Nvidia chips. (Noahpinion)
• OpenAI also struck a roughly $300 billion cloud deal with Oracle — and Oracle spends billions buying Nvidia chips to build those very data centers, sending money right back to Nvidia. (Noahpinion)
• OpenAI signed a chip deal with AMD and is set to become one of AMD's largest shareholders. Nvidia has also put money into Elon Musk's xAI. (Noahpinion)

See the pattern? A supplier invests in its customer, and the customer uses that money to buy from the supplier. The revenue looks real on an income statement — but some of it is, in effect, the same dollars going around the track. Asset manager Man Group described the modern AI cycle bluntly as a closed, recursive financing loop, where a handful of mega-caps act at once as suppliers, customers, investors, and cheerleaders. (Man Group)

Why this suddenly got serious

Two things happened in the last week of June 2026 that turned up the volume.

First, OpenAI signaled it might delay its IPO to 2027, and markets wobbled — all three major US indexes opened in the red on the news. (Fortune) When the poster child of the AI boom hints it's in no rush to face public-market scrutiny, investors notice.

Second, and more striking: the Bank for International Settlements — often called the "central bank of central banks" — used its flagship annual report to warn about a roughly $1 trillion AI investment boom and the web of private arrangements underneath it. The BIS specifically flagged the circular financing structure, and warned that if the big spenders (the "hyperscalers" like Microsoft, Amazon, Google, and Meta) slow or halt their capex, the whole chain — chipmakers, AI labs, construction firms, and the private-credit lenders behind them — could face revenue shortfalls at the same time. (Fortune) It drew a direct historical line to the canal, railway, and dot-com booms: real breakthroughs that attracted more money than near-term profits could justify.

When the BIS starts comparing your favorite trade to the railway mania of the 1840s, it's at least worth understanding the argument.

The other side — because there's a real one

Now, before anyone panics: "circular" does not automatically mean "fake." There's a serious counter-argument, and it deserves a fair hearing.

• The demand is real. Advanced AI chips are genuinely scarce, and building this infrastructure is genuinely expensive. In that world, companies don't just place orders — they lock in supply by pairing long-term purchase commitments with financing. Some call it a virtuous circle that lines up suppliers and customers to meet exploding demand. (Bloomberg)
• The scale isn't crazy — yet. UBS estimated the headline Nvidia-OpenAI arrangement at up to 13% of Nvidia's projected 2026 revenue — significant, but not the whole business. And crucially, much of it is performance-based, released as real milestones are hit, unlike the fixed, speculative vendor loans that sank telecom companies in 2000. (UBS)
• These are strong companies. Today's AI leaders have vastly healthier balance sheets than the telecoms that blew up a generation ago. This isn't a pile of profitless startups.

Reasonable, smart people are on both sides of this. That's exactly why it's worth understanding rather than cheering or panicking.

The part that should give you pause

Two details keep me cautious even after hearing the bull case.

1. The spending is eating almost all the cash. Combined hyperscaler capital spending has been running so high it reportedly consumes the large majority of these companies' operating cash flow after dividends and buybacks. (Sourcery Intel) That leaves little cushion if demand disappoints. And a lot of the broader buildout is debt-financed — AI-related companies tapped debt markets for at least $200 billion in 2025 alone. (Sourcery Intel)

2. The depreciation bill is coming. Those chips and data centers don't last forever — they wear out and go obsolete. As all that hardware ages, it turns into a mounting depreciation expense that eats into profits. By 2030, AI-related assets on corporate balance sheets could generate on the order of $400 billion a year in depreciation. (Sourcery Intel) The capex shows up as exciting growth today; the depreciation shows up as a drag on earnings tomorrow.

There's also a connection to a risk we've written about before: some of the money financing speculative data-center builds is coming through private credit — the same fast-growing, less-transparent corner of finance now being sold to everyday investors. (See our piece on the 10% yield being sold to regular investors.) If AI demand stumbles, the pain wouldn't stay neatly inside big tech.

Why this lands on your doorstep

Here's the uncomfortable punchline for the average investor: you almost certainly own this already.

As we covered in "You own an index fund, so you're diversified — right?", the top handful of stocks now make up about 40% of the S&P 500, and most of them are the very AI names at the center of this circle. If you hold a plain S&P 500 fund — VFV, VOO, SPY — a large slice of your money is riding on this buildout paying off. You didn't place that bet at a casino. It arrived quietly, through the "safe, diversified" index fund everyone recommends.

That's not a reason to sell. It's a reason to know. The healthiest position here isn't cheerleader or doomer — it's clear-eyed:

• Understand that a big chunk of your portfolio's recent gains came from AI, whether you chose that or not.
• Don't add a second concentrated AI bet on top of the one your index fund already gives you, thinking it's diversification.
• Watch the real tell: hyperscaler capex plans and actual AI revenue. As long as the spending is backed by genuine, growing, outside-the-circle revenue, the boom has legs. The day the biggest spenders start trimming budgets is the day to pay very close attention.

The canals were real. The railways were real. The internet was real. Every one of them changed the world — and every one of them also handed enormous losses to the investors who assumed the money could keep going in circles forever. AI can be genuinely transformative and a place where a lot of people lose money. Both things have been true before.

Know what you own. That's the whole game.

RiskStock is educational, not financial advice. We're not licensed advisors, and nothing here is a recommendation to buy or sell any specific security. Always do your own research.

Sources: Fortune — BIS annual report on the AI investment boom; Man Group — The AI Bubble; UBS — Should recent AI financing deals be a cause for concern?; Bloomberg — AI Circular Deals; Noahpinion — Should we worry about AI's circular deals?; Sourcery Intel — The Hidden Financial Bubble in AI Infrastructure.