NVIDIA's business model is now a circular financing operation. Over 90% of revenue comes from data center hardware, split roughly evenly between hyperscalers like Microsoft, Google, Amazon, and Meta, and a second tier of 'neoclouds' and sovereign AI customers that exist primarily to borrow money, buy NVIDIA GPUs, and rent compute to theoretical future customers. Jensen Huang has projected $1 trillion in combined Grace Blackwell and Vera Rubin sales by end of 2027, while simultaneously raising chip prices 17% and investing directly in customers like CoreWeave to ensure they can keep buying.
The circularity is the story. NVIDIA invests in Poolside at $6 billion, eyes Perplexity at a $30 billion valuation, builds agentic AI platforms, and funds open-source tooling. None of this is product diversification. All of it is demand generation for GPU compute. More than 70% of NVIDIA employees are now millionaires from stock appreciation, creating internal pressure tied directly to whether the rationalization for AI spending holds. The company is not just selling to the AI boom. It is financing the boom that justifies the selling.
Ed Zitron's full piece is worth reading not for the conclusion but for the structural breakdown: how hyperscaler and neocloud spending is categorized, what Bank of America analyst Vivek Arya said at the BoFA Global Technology Conference about the split, and why claimed sovereign AI buildouts lack evidence. Part two covers the history of circular financing schemes. If you track NVIDIA, AI infrastructure investment, or the mechanics of how tech bubbles self-sustain, this is the primary source to read this week.
[READ ORIGINAL →]