Nvidia Officially Launches AI Computing Partnership Program, Expanding Revenue-Sharing to Overseas Cloud Providers
Miles Bennett
Nvidia on July 2 launched its AI Compute Partnership, guaranteeing loans for smaller cloud operators in exchange for a share of their cloud revenue — starting with firms in Australia and Indonesia. This means Nvidia is shifting from pure chip sales toward locking in long-term recurring income tied to downstream compute operations.
How does this deal actually work?
Nvidia provides financial guarantees so smaller, lower-credit AI cloud operators can secure bank loans to buy GPU infrastructure.
In return, Nvidia takes a declining share of their cloud revenue over the contract term — the cut shrinks as the deal matures.
If demand falls short and GPUs sit idle, Nvidia commits to leasing back the unused capacity, absorbing the downside.
In plain terms = Nvidia is both supplier and guarantor, tying its own revenue to the success of its customers' businesses.
Why pick small firms in Australia and Indonesia?
The first partners are Sharon AI (Australia) and Firmus Technologies (Indonesia) — both independent of the hyperscale giants.
This means → Nvidia is deliberately cultivating a cohort of "neoclouds" that don't depend on Amazon, Google, or Microsoft, building a demand base outside the big three.
The backdrop: all three hyperscalers are accelerating in-house AI chip development, turning Nvidia's biggest customers into potential competitors.
What does Meta's cloud push mean for this chessboard?
Meta is reportedly preparing to launch its own cloud business, selling surplus AI compute to outside customers.
This reflects a broader shift: big tech is not just designing its own chips — it is starting to compete directly with independent GPU cloud operators.
This means → Neoclouds face a two-front squeeze: upstream chips get replaced, downstream clients get poached. Nvidia's guarantee program is essentially a lifeline.
Who else has Nvidia backed?
Earlier this year Nvidia invested $2 billion in GPU cloud operator CoreWeave and expanded their partnership agreement.
The target: accelerate construction of more than 5 gigawatts of AI data-center capacity by 2030.
CoreWeave has signed multi-year deals with Meta and Anthropic, among others, and raised its capital-expenditure guidance.
In plain terms = Nvidia is not just offering guarantees — it is investing directly, treating the entire neocloud ecosystem as a strategic moat.
The revenue mix is changing — are the risks changing too?
Revenue-sharing lets Nvidia capture upside from long-term cloud usage, creating a recurring income stream beyond one-time hardware sales.
But the contingent liabilities — the potential bills if partners can't repay — building up through these guarantees are a key variable the market needs to track.
This means → if AI compute demand disappoints, Nvidia doesn't just miss out on revenue shares — it also faces real costs from leasing back idle GPUs.
Does this model reduce risk — or create it?
Per *Barron's* analysis, the program helps partially hedge the demand slowdown risk as hyperscalers shift to in-house chips.
But whether it truly lowers over-building risk — or instead accelerates disorderly expansion among smaller neoclouds — remains an open question.
This reflects Nvidia's core tension: sustaining GPU sales growth requires helping customers lever up — but that leverage itself could create the next round of overcapacity.
Content is for reference only, not financial advice.