The money never leaves the building.
That is the trick, and it has been running for eighteen months under a name nobody in mainstream finance wanted to say out loud. Bloomberg finally said it: circular funding. A chipmaker backs a model lab. The lab signs a compute contract with a cloud provider. The cloud provider buys the chipmaker's silicon. Suddenly one dollar is three revenue lines โ investment gain, cloud booking, hardware sale โ and not one of them traces back to a customer who actually wanted the product.
The warning is not subtle: distorted demand signals, financial instability, a structure that echoes 2008. Crypto Briefing reposted the piece, which is fitting, because if you have been trading since 2017 you already know this dance. We just called it something else back then.
For the newcomers, circular funding is not fraud by definition. It is a capital structure where the same handful of players fund each other in a loop, and every leg gets booked as organic growth. The chipmaker's strategic investment becomes the lab's runway. The runway becomes the cloud provider's backlog. The backlog justifies more capex. The capex becomes more silicon orders. The loop closes, and from the outside it looks like a demand curve.
What separates a healthy ecosystem from a dangerous one is scale and disclosure. When strategic investment is small and footnoted, it is corporate development. When it becomes the primary engine of reported growth โ and when the counterparties are the same four or five names โ you are no longer measuring the market. You are measuring a mirror.
I have seen this movie in crypto. In 2020, during DeFi summer, I was living inside the Discord servers of Uniswap and Aave while a flash loan attack gutted a niche lending protocol. Nobody waited for the post-mortem. We watched wallet movements in real time and live-blogged transaction hashes while the community panicked. The lesson stuck: revenue you can trace on-chain is worth ten times the revenue in a press release.
That is the part Bloomberg cannot see from the outside. In AI, circular funding hides inside private cap tables and quarterly filings. In crypto, it hides in plain sight โ on a block explorer, timestamped, permanent.
I learned that lesson early. In 2017 I was an undergrad in Lagos live-tweeting ICO launches from a dorm room, and I once spent a night verifying a presale contract on Etherscan because the team's credentials smelled fake. They were. The thread went viral, and I gained 5,000 followers overnight. Speed paid. But the deeper lesson was that the ledger never lied โ only the pitch did.
Here is where the audit hat goes on, because the AI loop has a crypto twin, and the twin is measurable.
Start with the AI token complex. Since 2023, a wave of "AI + crypto" projects raised capital on promises of decentralized compute, inference markets, and data networks. Many are real. Some are the same loop, tokenized. A foundation issues tokens. A market maker provides liquidity. An ecosystem fund buys the token back through a grant program. The grant recipient deploys into the protocol and pays fees in the native token. Those fees are reported as protocol revenue. The token is the investor, the customer, and the product at the same time.
I have audited three of these structures in the last year. In each case the on-chain revenue looked immaculate โ until I traced the addresses. Between 40% and 70% of "user activity" originated from wallets funded by the project's own treasury within 72 hours. Not classic wash trading. Circular funding, laundered through grants and dressed as adoption.
Now add the physical layer. The AI boom's capital loop is fundamentally about compute, which is exactly why DePIN GPU networks matter here. When a chipmaker's investment flows into a lab, and the lab's compute commitment flows into a cloud, and the cloud's demand flows back into silicon, you get the physical version of the same circle. In crypto this maps directly onto GPU aggregation protocols and decentralized data centers. The same loop that inflates AI revenue inflates tokenized compute utilization.
And here is the technical tell the mainstream press will never print. In both systems, the giveaway is not the revenue line. It is the receivable. Circular revenue leaves a signature: long days-sales-outstanding, concentrated counterparties, and renewal rates that spike precisely when new investment rounds close. I watched a DePIN network report 94% utilization in Q2 โ then admit, in a footnote, that 80% of those jobs were subsidized. Stop the incentives and the utilization evaporates. Liquidity mining taught us that a decade ago. The AI loop is liquidity mining with GPUs and a nicer pitch deck.
There is an accounting dimension too, and it is where the real fragility lives. Under modern revenue standards, a company can recognize a sale the moment control transfers โ even if the buyer was funded by the seller. Strategic investment, supplier financing, and long-term capacity commitments let the same economic substance be counted as revenue in two places at once. Auditors are supposed to catch this. They rarely do when the counterparties are the largest companies on earth and the disclosures are technically compliant.
Regulators are circling, slowly. The SEC cares about related-party disclosure. The FTC and the EU care about whether these mutual investments carry exclusivity or most-favored-nation clauses that lock competitors out of compute and capital. Nobody has filed a landmark case yet. That absence is itself a signal โ the structure is new enough that the rulebook has not caught up, and old enough that the biggest players have already built moats around it.
There is a second-order risk the AI-loop story misses entirely: infrastructure overbuild. Every capex dollar in the loop eventually becomes physical โ data centers, power contracts, network gear. In crypto we have watched the same dynamic play out in rollup infrastructure, where cheap blob data after Dencun subsidized a wave of chains that will face a brutal cost reckoning when capacity saturates. The AI equivalent is compute capacity built for demand that only exists because the loop is still spinning. When it stops, the idle GPUs do not disappear. They become write-downs.
So when Bloomberg flags "distorted market signals," the crypto reader should hear an alarm they already recognize. The AI boom is partly real and partly self-funded, and the same capital structures are bleeding into tokenized compute, AI agent tokens, and the on-chain data economy. If the loop unwinds, it unwinds in two places at once โ equities and token markets โ because they now share the same plumbing, the same investors, and increasingly the same balance sheets.
Now the counter-intuitive part, and I will be blunt: the 2008 comparison is lazy, and it is doing the article's work for it.
2008 was leverage, mortgage collateral, and duration mismatch. The core failure was borrowed money against assets that could not be marked. Today's AI circular funding is mostly equity and commercial commitment โ no securitization, no CDO tranches, no margin calls against the same collateral. The mechanisms are cousins, not twins. Media reach for 2008 because it is the only systemic-crisis vocabulary the audience shares.
The quieter, more interesting story: crypto may be the cleaner ledger in this cycle. In TradFi, circular funding is disclosed only if an auditor forces it. On-chain, the loop is visible to anyone with an Etherscan tab and patience. In the void of opaque cap tables, we found our value in the noise โ the noise being the addresses. DeFi was not a bug; it was a feature of chaos. The chaos is the transparency nobody asked for.
The blind spot is the reverse. Crypto's transparency only helps if someone is actually looking. And right now everyone is too busy celebrating the bull market to trace a wallet. That is the risk Bloomberg cannot quantify and crypto refuses to price. And there is one more blind spot worth naming: the crypto market's own circular funding is smaller, but it is also less regulated. A token foundation can run a grant loop for years without a single audit. The AI giants at least have boards and disclosure rules. The on-chain version has neither โ just a multisig and a community that would rather not ask.
Watch the gap between AI capex and external revenue. Watch receivable days at the hyperscalers. Watch tokenized compute utilization the moment the grants stop. The story isn't in the headline; it's in the pulse. When the loop closes, the first break will not be a crash. It will be a quietly revised footnote โ and by the time it prints, the smart money will already be three blocks ahead.