Hook Bloomberg dropped a chart last week that should send chills down every crypto infrastructure investor’s spine. It showed that nearly 40% of total revenue for top-tier AI startups comes from other AI startups—a closed loop of spending where no real end-user demand exists. This isn't innovation; it's circular financing. And if history teaches us anything, it’s that when capital loops feed on themselves, they eventually snap. The 2000 telecom bubble was built on exactly this model: companies borrowing billions to build fiber optic networks for each other, only to discover that actual consumer demand filled less than 10% of capacity. The ensuing crash wiped out $2 trillion in market value. Today, crypto’s decentralized compute networks—Render, Akash, io.net—are the new fiber optic cables, and their primary customer is the AI hype machine. I’ve seen this movie before. In 2017, I dissected the ParagonCoin ICO as a high school junior, and I learned that when the narrative overshadows the revenue model, the technical foundation is sand. 2017’s dream is today’s regulation. Now the dream is AI-driven demand, and the regulation is market gravity.
Context The circular financing dynamic is deceptively simple. A startup like Mistral raises $500 million from venture capital. It spends $200 million on cloud compute from Azure or AWS—both owned by Microsoft and Amazon, who are also heavy investors in AI. Those cloud giants then reinvest their profits into AI research, which fuels more startup rounds. The cash cycles around like a carousel, but no new money enters from outside the system. Meanwhile, crypto’s decentralized GPU networks have positioned themselves as the cheaper, permissionless alternative. Projects like Akash offer compute at 60% of AWS costs, yet their revenue data shows a suspicious correlation with AI token price movements. When AI tokens pump, usage spikes; when they dump, nodes go idle. This is not organic adoption—it’s a macroeconomic derivative. The same pattern emerged in DeFi during 2020: liquidity flowed to yield farms, not to real economic activity. I led the liquidity crisis response at my fund then, mapping cascade failures across protocols. The lesson? Any market that depends on recycled capital for growth is a house of cards.
Core Insight: The Liquidity Dependency of Crypto Infrastructure Crypto infrastructure—especially GPU-based DePIN networks—is now a leveraged play on AI funding cycles. Let’s look at the numbers. Render Network’s active GPU nodes have grown 120% year-over-year, but its revenue per node has dropped 30% in the same period. Why? Because the supply of nodes ballooned in anticipation of AI demand that remains artificially sustained by VC money. Akash’s TVL in its compute markets similarly correlates with three-month trailing AI fundraising totals. When Circular Financing Is Active, crypto infrastructure appears healthy. When It contracts, the entire sector faces a liquidity vacuum. Based on my audit experience analyzing tokenomics for 50+ projects, I can tell you that the fundamental question is: how much of this compute demand is truly “sticky”? If an AI startup is only renting GPUs because it has fresh funding to burn, the moment that funding stops, the compute lease cancels. The crypto node operators are left with sunk costs. This is not speculation; it’s a repeat of the 2017 ICO bubble where projects raised millions for “blockchain-based” ideas, spent on marketing and dev salaries, and left investors holding empty tokens. Today’s tokens are GPU-hours, but the structural fragility is identical.
Contrarian Angle: The Decoupling Thesis That Isn’t Some analysts claim that crypto infrastructure is diversifying away from AI—that autonomous agent payments, on-chain machine learning training, or decentralized data storage will absorb the capacity. They point to the rise of “AI x Crypto” protocols like Bittensor or Allora as evidence of organic synergy. But this is a decoupling thesis in name only. Since my 2022 Terra collapse analysis, I have maintained a rule: any market that argues it can decouple from its primary demand driver during a bust is lying to itself. The telecom crash analogy is instructive. After the bubble burst, fiber optic capacity was eventually absorbed by streaming media—but that took five years. Meanwhile, the companies that survived were those with real end-user revenue, not those serving other busted firms. In crypto, the only validated end-users for decentralized compute today are AI startups and other crypto projects. That’s a closed loop within a closed loop. Every market cycle is a stress test for claims of decoupling. When the circular financing unravels, the “AI x Crypto” narrative will not protect asset prices—it will accelerate their decline.
Takeaway: Positioning for the Inevitable Adjustment The clock is ticking on circular financing. Major AI players like Microsoft and Google have already started slowing capital expenditure growth as their cloud revenue faces scrutiny. If the macro tide turns, the first casualties will be the most leveraged—those DePIN projects with high node inflation and low real user income. My framework suggests rotating toward infrastructure with diversified demand: storage networks (Filecoin) that serve enterprise backup, or oracle networks (Chainlink) that provide data to multiple sectors. But even these are not immune. The core insight remains: crypto is a macro asset, and the macro today is a fragile AI funding cycle. I wrote in 2024 that “2017’s dream is today’s regulation,” and now I see the same pattern in AI. The dream of infinite compute demand is built on a fantasy of perpetual capital. When the music stops, the GPU nodes will go dark, and the true survivors will be those who built for a reality where end-users pay real money—not just other startups playing pass-the-bag. The question isn’t if this correction happens; it’s whether you have positioned your portfolio for the pivot from narrative to fundamentals.
