NVIDIA's biggest customers are, to a significant degree, funded by NVIDIA itself. Over 90% of the company's revenue comes from data center hardware, and Jensen Huang has projected $1 trillion in GPU sales from the Grace Blackwell and Vera Rubin architectures by end of 2027. To keep that number plausible, NVIDIA is not just selling chips: it is financing the demand for them. That includes lending CoreWeave the capital to buy GPUs, which CoreWeave then rents back to OpenAI, which is signing multi-billion-dollar agreements to keep the whole loop turning.
The customer base breaks into two roughly equal halves. Hyperscalers like Microsoft, Google, Amazon, Meta, and Oracle account for 50 to 60 percent of GPU sales. The rest comes from neoclouds and sovereign AI buyers, companies that raise debt, purchase NVIDIA hardware, and park it in data centers waiting for tenants. Bank of America analyst Vivek Arya confirmed this split at the BofA Global Technology Conference in June. NVIDIA has also committed $6 billion to AI startup Poolside, is discussing a multi-billion investment in Perplexity at a $30 billion valuation, and is building out its own agentic AI platforms. None of this is altruism. All of it is designed to manufacture GPU demand.
Ed Zitron's full piece is worth reading not for the conclusion, which is blunt: NVIDIA is running a circular financing operation. What earns your time is the structural anatomy of how it works, who is holding the bag at each link, and why it has not collapsed yet. Part two, covering the history of circular financing, is coming next week. The uncomfortable question Zitron is building toward: what happens to 70 percent of NVIDIA's workforce, more than 70 percent of whom are now millionaires on paper, if the loop breaks.
[READ ORIGINAL →]