Analyst Rick says NVIDIA is functioning as a "quasi-central bank" for the AI
era: it cannot issue currency or set rates, but it is reshaping allocation of
the scarcest capital input—GPUs—and accelerating financialization of the AI
industry. The industry is shifting toward an infrastructure financing model:
financing to buy GPUs → GPU leasing produces cash flows → cash supports further
financing → more GPU purchases, creating a self-reinforcing capital flywheel.
GPUs are acquiring infrastructure-asset characteristics—long-duration cash
flows, collateral value and securitization potential—which helps explain
NVIDIA’s continued lock-up of HBM, advanced packaging and supply capacity. Cloud
giants (Microsoft, Google, Meta, Amazon) are expanding capex and using orders,
prepayments and deals to ease upstream financing; NeoCloud, HBM and
advanced-packaging firms are raising debt and signing long-term contracts;
private credit, infrastructure funds and sovereign wealth funds are moving into
AI data-center investment. Market implications: sustained, large-scale capital
demand may keep long-term funding costs elevated even if short-term policy rates
fall; long-duration capital will flow not only to GPUs but to data centers,
power, fiber, cooling, HBM and packaging. Financing costs are likely to
bifurcate—lower for AI firms with predictable cash flows, relatively higher for
traditional sectors—raising the risk of widening credit spreads.
