The Loop That Ate AI: Why Circular Financing Is 2008's Ugly Twin

Over the past seven days, a chip giant I track quietly funneled $2 billion into a startup that promised to buy its GPUs back. No headline. No fanfare. Just a press release buried in the noise. But my radar pinged — that wasn't an investment. That was a loop. I've been tracing the trail from NFT peaks to DeFi valleys for six years, and I've seen this movie before. The same closed circuit that propped up CryptoPunks in 2021 is now running through AI's core. The difference? This time, the collateral is faster, the leverage is murkier, and the participants are the ones who should know better. Let me break down what's actually happening. Circular financing in AI looks like this: a chip or cloud provider invests in an AI lab. The AI lab, in turn, signs a multi-year contract to purchase compute from that same provider. The provider books that contract as future revenue. The AI lab's valuation spikes, attracting more investors. Those investors fund more purchases. The cycle repeats. On paper, everyone wins. On the ledger, the same dollar is just traveling in a circle. When I hosted a live-streamed party in Buenos Aires during the 2021 NFT peak, I watched the same psychological loop: early adopters flipped assets for 10x returns, and the social energy alone created demand that didn't exist. I interviewed three of them in real-time — they were buying floor prices not because they believed in the art, but because they believed other people would believe. That's circular. That's the vibe report I published. That's the same energy driving AI's circular financing now. The core mechanism is simple, but the danger is compound. In a typical structure, the investor-supplier gives capital to the AI company, and the AI company commits to buying that supplier's compute. The supplier records the committed contract as revenue or backlog. The AI company uses that revenue story to raise more capital. The capital gets reinvested. Some of that "revenue" is actually the same capital moving through different books — it's an accounting relay, not an independent cash flow from end users. And here's the uncomfortable part: the terminal demand for generative AI — enterprise subscriptions, API calls, consumer payments — is still an order of magnitude smaller than the infrastructure investment already committed. I'm not saying the demand won't come. I'm saying the loop is designed to run ahead of that demand until it can't. Based on my audit experience, the gap between committed contracts and verifiable third-party revenue is the single most important number nobody's publishing. The 2008 comparison isn't a lazy metaphor. It's structurally identical. In 2008, CDOs were built on mortgage payments that were themselves funded by the loans being securitized. Here, compute contracts are built on capital flows that are themselves funded by the companies purchasing the compute. The counterparty risk is concentrated in a handful of central nodes — the chip giants, the hyperscalers, the top AI labs. If one of those nodes wobbles, the shock wave travels through the whole network. That's not speculation. That's how the CDO market worked, and that's how AIG became the center of a systemic failure. What the Bloomberg piece — which I read as the underlying source — doesn't explicitly name is who's providing the leverage. The private credit market has been quietly financing these loops through special purpose vehicles and structured products. That's the dark pool. Private credit is opaque, loosely regulated, and heavily exposed to AI-related borrowers. If the loop tightens, the first domino might fall in a fund nobody's heard of, in a filing nobody's reading. But here's where I diverge from the standard warning. The real blind spot isn't just financial. It's physical. Circular financing is inflating compute demand signals, which triggers massive data center and GPU cluster construction. Those builds take 18 to 36 months from planning to operation. If AI demand growth slows during that window, you get "completed and deserted" — data centers built for a demand wave that never crests. My 2022 survival nights in Palermo taught me what that looks like: founders who overbuilt on borrowed conviction, then watched the floor fall out. The difference is data centers can't pivot to selling art. They're asset-specific, power-hungry, and they depreciate faster than a fad. And that's the contrarian angle everyone's missing. The market is pricing the loop as a financial risk, but the bigger, slower bomb is the infrastructure overhang. A data center is a physical asset with no secondary market when it's idle. GPU rental prices, utilization rates, and electricity procurement contracts — those are the telltales. I've been chasing the alpha through the noise, and the alpha right now is in understanding that the loop is creating a phantom demand for silicon that will keep the construction cranes busy long after the demand curve bends. The sprint to the ETF finish line in 2024 taught me that speed matters — I broke a story about BlackRock analysts' off-the-record comments in Miami and captured 60% of the engagement share. But speed without substance is noise. The substance here is that circular financing is a temporary bridge that can sustain a narrative, not a market. When the bridge collapses, the collapse isn't gradual. It's a cliff because the loop has been inflating a leverage ratio that no one's tracking. So what do I watch next? I'm tracking three things with a weekly cadence: the disclosure of related-party revenue in major AI companies' filings, the true utilization rates of hyperscale data centers, and the spreads on private credit funds with AI exposure. When those three lines converge, the loop will break. Not in a crash — in a quiet repricing. The questions I keep asking myself: how much of the AI valuation is circular premium? And when the liquidity dries up, which central node gets saved, and which gets sacrificed? I'm a survivor of the 2022 deflationary tide. I watched LUNA's collapse from the trenches, and the lesson was clear: the more the narrative needs to explain itself, the closer it is to breaking. AI's circular financing narrative is getting louder. That's not a good sign. It's a warning. The race isn't won by the fastest breaker. It's won by the one who sees the loop closing before it snaps. And right now, the loop is closing.