Hook
Ed Zitron just dropped a bomb on CNBC. The CEO of EZ Primary Research laid out a simple, terrifying thesis: NVIDIA isn't just selling shovels in the AI gold rush. It's also lending the miners the money to buy those shovels. And the gold? It's sitting in a vault owned by two companies that have never turned a profit.
Markets don't lie — they just take their time revealing the truth. The truth here is a feedback loop that looks eerily familiar to anyone who survived the 2022 Terra collapse. Speed is the only currency that never depreciates, but in this case, the speed is all illusion. The capital is circulating inside a closed loop: NVIDIA sells GPUs → CoreWeave/Lambda borrows → borrows to buy more GPUs → NVIDIA books revenue. The end customer? OpenAI and Anthropic, burning cash at a rate that would make LUNA's Anchor protocol look conservative.
Context
For context, NVIDIA has become the single most important hardware supplier in the AI revolution. Its GPUs power the training and inference of every major large language model. The company's market cap has skyrocketed past $2 trillion. But the supply chain that feeds this growth is not a simple vendor-buyer relationship. It's a complex web of investments, long-term procurement contracts, and financing arrangements where NVIDIA plays three roles: supplier, customer, and financing facilitator.
CoreWeave, Lambda, and other cloud compute providers are NVIDIA's biggest customers. They buy enormous clusters of H100 and B200 GPUs. But they don't pay for them upfront. Instead, they secure funding from venture capital, debt markets, and — critically — from NVIDIA itself. NVIDIA doesn't directly lend cash; it lends its credit. It signs multi-year contracts with these providers, guaranteeing future GPU purchases. Armed with those contracts, providers can go to banks and bond markets to raise debt. That debt is then used to buy more NVIDIA hardware. The GPU giant gets its revenue today, while the providers carry the risk of future demand from end users.
Core
Here is the key fact: the entire structure depends on a single assumption — that the demand for AI compute from companies like OpenAI and Anthropic will continue to grow exponentially. Based on my experience auditing the EOS IEO in 2017, I can tell you that concentrated demand is a ticking time bomb. Back then, the entire EOS ecosystem was propped up by a few whales. When they stopped buying, the floor fell out. The same dynamic is playing out here, but at institutional scale.
Let me be specific. Over the past 12 months, NVIDIA has invested directly in CoreWeave and Lambda. It has also provided them with procurement contracts worth billions. These contracts are used as collateral for debt financing. The debt is then converted into GPU orders. The orders go back to NVIDIA. The cycle repeats. According to public filings, CoreWeave raised $1.5 billion in debt earlier this year, secured by contracts with NVIDIA and its AI clients. That debt is now funding more GPU purchases. Sentiment is the invisible ledger of value — and right now, the ledger shows a single entry: "AI demand will grow forever."
But the end-user demand is anything but diversified. OpenAI and Anthropic account for an estimated 60-70% of all compute hours rented from CoreWeave and Lambda. Both companies are still burning cash. OpenAI's revenue is growing, but its costs are growing faster. Anthropic has yet to report meaningful revenue. They are surviving on venture capital, not on profitable operations. If either company slows down its spending — or, worse, collapses — the entire debt structure built on their future compute needs will implode.
Contrarian
Most analysts are focused on NVIDIA's earnings beat and the AI capex boom. They see growth and extrapolate linearly. But the contrarian angle is that this is not a technology story. It is a financial engineering story. NVIDIA is essentially creating its own demand by financing its customers' purchases. This is not new. It happened in the telecom bubble of 2000, when equipment vendors like Cisco provided financing to telcos that bought their routers. It happened in the housing bubble, when banks lent money to people who could not afford to pay it back. The common thread: the seller controls the credit, and the buyer's ability to pay is never tested until the music stops.
In crypto, we call this a "liquidity loop." The DeFi summer of 2020 was full of them. Protocols like Compound and Aave would incentivize users to deposit and borrow, with the borrowed funds going back into the same protocols. The loop inflated TVL but produced no real economic value. When the incentives dried up, the TVL vanished. Markets don't lie — they just front-run the narrative.
What makes this situation more dangerous than a typical bubble is the lack of transparency. CoreWeave and Lambda are private companies. Their debt covenants, their client concentration, their cash flow — all opaque. The only public signal is NVIDIA's revenue, which is growing at 200%+ year-over-year. But that revenue is not "real" in the traditional sense. It is pre-funded by debt that is guaranteed by future demand that may not materialize. If you strip out the financial engineering, NVIDIA's growth rate looks more like a Ponzi scheme than a sustainable business.
Takeaway
The next watchpoint is not NVIDIA's next earnings report. It's the debt markets. If credit spreads on CoreWeave or Lambda bonds widen, or if rating agencies start downgrading their debt, the feedback loop will reverse. NVIDIA will be forced to either write down its investments or absorb losses from cancelled contracts. The AI narrative will shift from "infinite demand" to "capital efficiency."
We have seen this playbook before. Terra's Anchor protocol offered 20% yields. It attracted massive deposits. The deposits were used to buy LUNA. The loop collapsed when new deposits stopped flowing. The same principle applies here: any system that relies on continuous capital inflows to sustain itself is fragile. Speed is the only currency that never depreciates — but only if you can exit before the loop breaks.
The question is: who is the exit liquidity this time?
