Artificial intelligence factories are turning into a class of investment assets led by Nvidia and global financial institutions

Nvidia has announced strategic partnerships with several major global financial and investment institutions, including Apollo, Black Rock, Blackstone, Brookfield, Goldman Sachs, and KKR. These partnerships aim to create independent financing platforms dedicated to raising over $500 billion in third-party capital over time to support the construction of artificial intelligence infrastructure. This development represents a pivotal moment in the technology sector, where the market is shifting from a phase of buying chips and building data centers project by project, to a phase of financing artificial intelligence factories as a productive class of investment assets based on repeatable platforms, long-term venture capital, and a diverse set of clients who leverage computing to generate revenue.

Nvidia's platform capabilities exceed the idea of simply supplying a silicon chip, to include an integrated platform comprising accelerated computing, networks, system software, artificial intelligence frameworks, and a global developer network. Artificial intelligence factories based on the NVIDIA DSX architecture can run a wide range of models and algorithms, from language, vision, and speech models to biology, physical AI, and robotics. This infrastructure is characterized by being flexible and replaceable, as it is built on a globally accepted architecture across major clouds, system manufacturers, and institutions, allowing the factory to be redirected to another client or operator when needs change.

Continuous software development through the CUDA environment contributes to improving computing efficiency and performance and reducing the total cost of ownership for previously installed infrastructure. This is evident in the A100 processor model based on the Ampere architecture, which was launched in 2020, and after six years, in 2026, is still in active commercial use for training, fine-tuning, inference, and high-performance computing operations, with expectations of extending its useful economic life to nearly a decade.

Market data shows stability and resilience in the economic value of Nvidia's computing across different generations.For example, the rental prices for H100 processors for one-year contracts rose from around $1.70 per processor hour in October 2025 to around $2.35 in March 2026. The average on-demand rental price across various providers also increased from around $2.00 per processor hour in October 2025 to $2.70 in June 2026. At the same time, the capacity of Blackwell processors is in high demand with a price premium, with B200 cloud processor prices ranging from $5.30 to $7.05 per processor hour.

This initiative aims to address the imbalance in financing available to artificial intelligence companies and specialized clouds that have a real demand for computing but lack appropriate financing mechanisms in terms of volume and cost. The partner financial institutions independently assess each project based on a study of the client, demand, usage rate, cash flows, and residual value.To ensure risk management discipline, Nvidia may provide a residual value support mechanism of up to 25% as a maximum for a single investment opportunity.This support is carefully evaluated on a project-by-project basis and is designed to support independent financial coverage, not replace it, leveraging the reusability, transferability, and upgradability of Nvidia's computing.

This investment vision is based on converting energy and data into actionable intelligence that serves artificial intelligence labs, institutions, and countries. The economic cycle of artificial intelligence is completed through an integrated loop, where increased computing leads to the development of better artificial intelligence models, better models lead to increased usage, and higher revenues are generated, which in turn drive the expansion of computing consumption and construction.