
What happened
Amazon set up an $8b special purpose vehicle to buy Nvidia Grace Blackwell chips and lease them back to AWS clients, while Morgan Stanley reinstated Nvidia as a Top Pick.
Why it matters
The structure keeps AI hardware off AWS's balance sheet, which appears to let AWS keep building AI capacity as cloud capital demands rise, and points to new funding structures supporting large Nvidia orders.
What to watch
The test is whether other hyperscalers or sovereign AI buyers set up similar chip vehicles that explicitly name Nvidia hardware with disclosed sizes and timelines, which would show repeatable off balance sheet funding.
WHO IT HITSCloud infrastructure teams and AI capacity planners at hyperscalers can look to this structure as a way to fund AI hardware purchases without holding it all on their own balance sheets. Nvidia investors get an early signal that order financing may become repeatable, though the body flags vendor financing and data center constraints as open risks.
Summaries like this, in your inbox every morning.
The Amazon vehicle sits inside a wider Nvidia AI infrastructure story, alongside Morgan Stanley's reinstated Top Pick on Nvidia and Nvidia's own US$500b ecosystem financing plans. Nvidia positions itself as a data center scale AI infrastructure provider, supplying the chips and systems that power cloud AI workloads for hyperscalers and enterprises across the US, Europe and Asia. That role makes its hardware a core input for financing experiments like Amazon's $8b vehicle, and a focal point for broker views on AI data center demand.
The structure is designed as an off balance sheet financing tool, which appears to reflect rising cloud capital demands. The body also notes that Nvidia is far from the only business tied tightly to this AI infrastructure build out, and it flags risks around vendor financing, data center constraints and customer vertical integration that remain in view alongside the catalysts.
What the outcome hinges on is whether the structure proves repeatable. If other hyperscalers or sovereign AI buyers set up similar vehicles that explicitly name Nvidia hardware, with disclosed sizes and timelines, that would suggest off balance sheet funding is becoming a normal part of Nvidia's order pipeline. Concrete multi year lease commitments on Grace Blackwell or future platforms would be the clearest early sign. The body also flags two warning signs, one major, worth knowing about at Nvidia, so the financing trend and those signs may need to be read together.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Manulife Financial Corporation launched a first-of-its-kind CoverMe travel insurance plugin in ChatGPT in late…

Google said free Gemini app users will be restricted to the "Flash-Lite" model starting October 9; "Flash" nee…

Nvidia expects AI spending to grow to $3 trillion to $4 trillion annually by 2030, while it reported $96 billi…

David Robinson, who led safety reporting on OpenAI's main products, resigned, saying the company's culture is…

On September 25, Bank of America reiterated a Buy rating on Advanced Micro Devices and raised its price target…

On September 22, 2026, Anthropic announced Claude Opus 5.5 at $4 input / $20 output per million tokens, then O…
