Скачать или смотреть Why NVidia circular financing is very profitable but also that's changing in 2026 with recent deals

NVidia has been engaged in circular financing, just like the other hyperscalers. However, where they're different is that they make 70% gross margin and they collect a huge revenue stream on the back end. This, combined with the fact that they always try to unlock outside investment (they are not the only investor), means that unlike the other hyperscalers, NVidia makes enough gross profit to more than cover their investment, and thus even if the equity goes to $0, they still make a profit.

This is NOT like the other circular financing deals.

The other thing we notice is that unlike Microsoft or Oracle's deals, which are very long term multi-year deals, NVidia's deals are all very short term and the circular financing goes right back to Nvidia very rapidly, usually within less than a year. The only reason NVidia would structure a deal like that is if they don't believe the AI bubble will last much longer, and thus want to ensure the circular part of the circular financing happens quickly.

We see in recent times, especially in the last few months, they've been announcing deals like with Coreweave and OpenAI where they're at substantial risk of equity loss in these new deals. I think this could be a sign that we might be near the top of the AI bubble if Nvidia feels compelled to prop up the bubble.

Sources:
NVidia IR: https://investor.nvidia.com/home/defa...

https://nvidianews.nvidia.com/news/op...

https://techcrunch.com/2025/10/08/eve...

https://www.cnbc.com/2026/01/31/nvidi...

https://www.wsj.com/tech/ai/the-100-b...

00:00 — The thesis: NVIDIA’s “investments” are designed to win even if equity goes to $0
03:00 — Definitions + the real comparison: vendor financing vs round-tripping vs circular financing, and why NVIDIA’s math is different
08:19 — CoreWeave: IPO anchor → capacity backstop → the 2026 equity backstop (and why that’s a red flag)
13:56 — xAI + OpenAI Deal #1: the “anchor” pattern, and the $10B-per-GW structure (money comes after capacity)
16:28 — OpenAI Deal drama + Deal #2 rumors: why the ratios are getting worse, why this looks late-cycle, and what breaks first