Bloomberg’s latest chart tells a story investors don’t want to see: AI startups are burning cash to buy each other’s services. This circular financing loop is eerily similar to the 2000 telecom bubble. And crypto’s infrastructure tokens are sitting right on the fault line.
I’ve seen this pattern before. In 2017, I front-ran the ICO bubble by auditing MelonPort’s staking contract—an integer overflow that let me buy $150k of MELON at the bottom. Code executes promises; men make excuses. The lesson: when the underlying mechanics are broken, the price is just an echo. Today, the mechanics of AI financing are broken.
Context: Crypto’s GPU-dependent projects—Render Network, Akash, Livepeer—have ridden the AI narrative to multi-billion valuations. The thesis is simple: AI needs compute, crypto provides it cheaply and in a trust-minimized way. But who is paying for that compute? A recent Bloomberg analysis reveals a disturbing trend: AI startups are raising money from VCs, then spending it on cloud services from other AI startups, which in turn use that revenue to raise more funding from the same VCs. It’s a closed loop. No external revenue from real customers. Just recycled capital.
This is the circular trade. It worked for telecom in the 1990s—until it didn’t. Overbuilt fiber networks went bankrupt. The same fate awaits overbuilt GPU capacity if the financing stops. Crypto infrastructure tokens are leveraged plays on that capacity. If AI venture capital tightens—and macro headwinds suggest it will—the demand for decentralized compute evaporates.
Core: Order flow tells the story. I ran the on-chain data last month: large wallets holding RNDR, AKT, and LPT started moving tokens to exchanges in the first week of April. Total inflows across these three tokens hit $12 million—a 3x increase from the monthly average. Meanwhile, spot prices barely moved; the narrative still held. That’s a classic divergence: smart money distributing into retail buying the story.
On-chain eyes saw the mania before the crowd did. The same pattern emerged with NFT derivatives in 2021—wash trading inflated volume, whale wallets sold into the hype. This time, the hype is AI, but the mechanics are identical. I’m tracking exchange inflows for these tokens daily. If they breach 20% of circulating supply, I’ll short the front month futures.
But the deeper problem is protocol revenue. I checked the income statements of the top three decentralized compute platforms: revenue per GPU hour is declining, with average utilization below 40%. Compare that to AWS or Azure, which hover around 60-70%. The difference is that traditional cloud providers have sticky enterprise contracts; crypto platforms rely on spot demand from AI startups that may not exist in six months.
Contrarian: The market is bullish on AI + crypto. Coinbase’s recent research report called it ‘the next mega-theme.’ But the contrarian view—my view—is that this financing loop is a ticking bomb. When it explodes, the contagion will hit crypto infrastructure first. Why? Because the same VCs funding AI startups are also backing crypto projects. They’ll pull liquidity from both. Retail token buyers will be the last ones holding the bag.
I’ve been here before. In May 2022, I modeled the over-collateralization risks of Anchor Protocol and hedged with Deribit puts. The Terra/Luna collapse wiped out $40 billion. I survived because I saw the circular nature of the Anchor yield—new deposits paying old depositors’ yields. AI circular financing is worse: at least Anchor had real deposits (albeit unsustainable). AI circular financing has no external revenue at all. It’s a pure Ponzi on VC money.
Takeaway: Is your GPU token a bet on genuine demand or a leveraged play on a circular trade? The chart is just the echo; the code is the voice. Code that shows declining revenue, concentrated wallet distributions, and reliance on a single narrative is code you shouldn’t trust. I don’t trust it. I’m sizing my shorts and waiting for the first major AI startup to miss its next funding round.
Yield farming was the only shelter in the storm. This time, it’s cash and short positions. Survival isn’t about being right; it’s about staying solvent. And right now, solvency means staying away from tokens that depend on AI venture capital to survive.

