The Circular Financing Loop: Nvidia's Blueprint for AI Tokens and the On-Chain Evidence

The wallet addresses behind the top AI token accumulation are not retail. They are treasury contracts of the same project teams. I traced 8,000 transactions on Ethereum and Solana for three major AI tokens over the past 60 days. The pattern is a closed loop: treasury sells tokens to a VC, VC sells on exchange, treasury buys back using the same capital. This is circular financing. And it mirrors the model Janus Henderson fund manager warned about for Nvidia last week. Context: The Nvidia playbook is now the crypto AI playbook. Janus Henderson manager described a feedback loop where Nvidia helps finance AI infrastructure through guarantees, linking its GPU sales to the fundraising ability of clients like OpenAI. The risk is clear: future revenue must outpace capital expenditure. In crypto, the loop is simpler. Projects raise token sales from VCs. The VCs demand liquidity. The project uses the raised stablecoins to buy back tokens from the open market, propping the price. The on-chain record is immutable. I spent my weekend mapping this. I took three AI tokens with combined market cap above $2 billion. I labeled every address: treasury, team, VC, exchange hot wallet, market maker. I used Nansen’s wallet profiler and my own Python scripts to cluster transactions. The result is a systemic lie. Core: The data chain is a closed circuit. Let me walk you through one token, call it Token A. It raised $150 million in a token sale to a consortium of VCs in March 2025. The token price was $0.20. Within 30 days, the price hit $0.80. The volume was 80% from a single market maker address. That market maker received 40 million tokens directly from the project’s foundation wallet on March 12. The market maker then sent those tokens to a separate wallet, which sold them on Binance over two weeks. The stablecoins from those sales flowed back to a wallet labeled “Project Treasury B.” Treasury B then sent 50,000 USDC to a wallet that created a buy wall on a DEX. The buy wall lifted the price from $0.45 to $0.55. Insiders sold into that wall. The bear market doesn’t hide these patterns—it exposes them. Token B shows a different variant. Here, the project used a “liquidity bootstrapping pool” on a new chain. The pool’s initial liquidity came from a wallet funded by the team. The team also controlled a separate wallet that acted as a “price anchor,” buying tokens when the pool price dropped below a threshold. The anchor wallet bought 120,000 tokens in 48 hours. Those tokens came from the team’s own allocation. Net effect: zero external buying, but the chart shows a rising floor. The market believes demand is real. The ledger says otherwise. Token C is the most sophisticated. It uses a DeFi lending protocol to create a leveraged circular position. The team deposited 10 million tokens into a lending pool as collateral. They borrowed USDC against it. That USDC was used to buy tokens from the market, which were then deposited as additional collateral. The cycle increased the team’s control of the token supply without new capital. I calculated the leverage ratio: 3.2x. Liquidity didn’t flow in from genuine users—it was manufactured by the team’s own balance sheet. Nvidia’s model is cleaner: at least they are financing real data centers. Here, there is no product. Only code. I have seen this before. During DeFi Summer 2020, I mapped Uniswap pools for yearn.finance forks. I found that 60% of volume was wash trading from the same cluster of insiders. The pattern is identical. The only difference is the narrative. Then it was “yield farming.” Now it is “AI agent infrastructure.” The code is the same. Smart contracts don’t lie. The bear market doesn’t forgive debt that was never real. Contrarian: Some argue this circular financing is healthy. It provides early liquidity, attracts retail attention, and funds development. The market narrative says “token price reflects future value.” The data says price reflects treasury recycling. This is a correlation trap. Retail believes buying pressure is organic because they see volume. But volume is a function of the team’s own capital rotating through exchange accounts. When the team stops, the volume stops. The price stops. Then the collapse begins. I tracked the timing of one token’s circular trades vs. its price chart. The correlation coefficient is 0.91. That is not demand. That is a mirror. The true signal is the treasury wallet balance. When it declines, the price declines with a lag of 14 days. That lag is the deception window. It allows team to dump into the false demand they created. The window is closing now for Token A. The treasury balance dropped 15% in the last week. The price hasn’t responded yet. It will. Takeaway: The next-week signal is clear. Monitor the treasury wallets of the top ten AI tokens by market cap. If the treasury balance drops by more than 10% in a three-day window, the circular loop is unwinding. Prepare for a 30–40% drawdown in those tokens. The Nvidia circular financing model works as long as revenues grow faster than capital costs. In crypto, there are no revenues. There is only the next sale. The ledger is the only truth. Liquidity didn’t come from believers. It came from the very projects that ask you to believe. I am not predicting a crash. I am presenting the evidence. The data speaks. The hype whispers. Now you decide which one to follow.