Breaking

NVIDIA to Deploy AI Factories With Clouds Under Revenue-Sharing Model

July 2, 2026 at 01:00 EDT

Infra & Chips
Sales & Business
Funding & M&A

On July 1, 2026, NVIDIA unveiled a new business model to deploy large-scale, multi-tenant "AI factories" in partnership with AI cloud providers. Beyond selling hardware, it combines cloud revenue-sharing and credit support to make NVIDIA infrastructure more accessible to emerging AI companies and research institutions.

NVIDIA · AI Factory Buildout

A New Model to Deploy AI Compute at Scale

NVIDIA turns from one-off GPU sales to revenue-sharing and credit-support with AI cloud providers — easing upfront capital and opening always-on "token production" to a fast-growing ecosystem.

~20×

Lower cost per token via full-stack software optimization + latest GPUs

25×

Revenue increase within just four weeks of optimization

+40%

More GPUs under power constraints with DSX MaxLPS

Cost per token — before vs. after optimization

Full-stack optimization drives roughly a 20× reduction. Taller column = higher cost.

20×

Before
(baseline)

1×

After
optimization

How the model works

Provider procures NVIDIA platform with credit-support & reduced capital burden

→

Runs multi-tenant AI factory producing tokens at scale as a service

→

NVIDIA earns ongoing share of cloud revenue from covered capacity

Partners scaling AI factories — six continents

Firmus

170,000-GPU liquid-cooled campus across Australia / APAC

CoreWeave

Early Vera Rubin adoption; physical AI support

Nebius

Provides a Physical AI Workbench

Why it appeals

Multi-tenancy lifts GPU utilization & monetization
Lowest token cost becomes a competitive edge
Flexible clouds for regional & sovereign AI demand

Open questions

Long-term sustainability of revenue-sharing
Undisclosed revenue-share rates
Real-operation proof beyond blogs & case studies

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.