
NVIDIA’s Neocloud Squeeze Gives Leverage Back to Hyperscalers

NVIDIA‘s Q3 Data Center revenue of $107.6B hides a strategic split. Hyperscalers—Google, Amazon, Microsoft, and Meta—bought $60.1B worth of NVIDIA hardware, up 71% year over year, while racing to replace those GPUs with custom ASICs. The remaining $47.5B came from AI Clouds, Industrial, and Enterprise customers: neoclouds and frontier labs that build no competing chips but need NVIDIA‘s balance sheet to scale. That segment nearly tripled year over year on strong Vera Rubin shipments.
The company’s own breakdown notes that this $47B bucket increased 18% sequentially and nearly tripled year over year. But the mix shifted slightly: after non-hyperscaler share peaked at 45.3% of Data Center revenue in Q2, it ticked down to 44.1% in Q3 just as NVIDIA signaled tighter financing terms.
NVIDIA‘s financial exposure to neoclouds is substantial. It holds $102B in cloud and data center lease commitments, including $32B directly with neoclouds, alongside $390B in purchase commitments. Interim CEO and CFO Colette Kress set a firmer line: “We have supported the development of this ecosystem, and we will continue to be disciplined and selective in our support.” For operators that lack hyperscaler cash flows, a more selective backstop could moderate growth for smaller, capital-constrained companies.
The catch is strategic: a diversified buyer base is NVIDIA‘s best defense against hyperscaler custom silicon. Squeezing neoclouds lowers NVIDIA‘s risk on leases and financing, but it hands leverage right back to Google, Amazon, Microsoft, and Meta. NVIDIA‘s hardest problem is not demand; it is choosing which customer to feed and how much of its own balance sheet to put behind each one.

