Nvidia's GPU Financing Loop Draws Skepticism
The Nvidia-CoreWeave-Nebius circular financing story got a lot of attention. The core claim: Nvidia invested $2B into CoreWeave for a 9% stake, and CoreWeave is spending $35B in capex in 2026, most of which flows back to Nvidia for GPUs. The implication is that Nvidia is effectively financing demand for its own products.
The HN discussion was split. Some called it a nothing-burger: 'all financing is circular,' and Nvidia's $2B is only 5.7% of CoreWeave's single-year capex. Others pointed out that the more interesting question is whether the GPU buildout is being sustained by financial engineering rather than genuine end-user demand pulling through the whole chain.
This connects to a broader anxiety in the threads: the AI infrastructure boom may be more fragile than it looks, because so much of the money is moving in loops between a small number of players rather than being grounded in profitable end-user applications. That's a different kind of risk than traditional capex cycles.
So what?
If you're building on top of GPU cloud providers, understand that their financial stability depends on this loop continuing. A demand air pocket at the top of the stack, fewer developers actually paying for AI inference, could ripple down faster than a traditional infrastructure cycle would suggest.