Nonprofit Plans To Lease 24,000 Nvidia H100 GPUs For AI

A nonprofit funded by a cryptocurrency billionaire said it has bought up $500 million (£411m) in advanced Nvidia GPU chips and plans to deploy them in data centres for lease to third parties wishing to train artificial intelligence systems.

Cloud computing organisation Voltage Park said it believes the market for machine learning compute resources is “broken” and plans to offer more flexibility.

The firm is a wholly owned subsidiary of the nonprofit Navigation Fund, with any profit it earns going to Navigation.

Navigation was was founded by cryptocurrency billionaire Jed McCaleb, who earned billions from creating three of the best-known crypto companies: the Mt. Gox trading platform, payment protocol developer Ripple Labs and blockchain firm Stellar.

Accessibility

Voltage Park says it seeks to make in-demand AI training resources more easily available to organisations across the board, from large enterprises to research universities, seed-stage start-ups and nonprofits.

Chief executive Eric Park said the shortage of processing power brought on by intense demand for GPUs made by Nvidia and others has created an artificial damper on AI innovation.

“ML teams and AI founders have to wait months or pay exorbitant sums to access the latest hardware to train their models,” Park said. “We hope to redress this imbalance and accelerate cutting-edge work in AI.”

He said the firm plans to offer more flexibility than some providers, suitable for smaller firms, while shortening wait times and offering lower GPU rental rates.

Nvidia H100s

The firm says it has bought up some 24,000 Nvidia H100 GPUs, making it “one of the most powerful collections of cutting-edge ML compute in the world” once those resources come online by about February.

It plans to offer short-term leases and hourly billing and is currently calling for feedback on how to engineer its clusters, which are made up of 80GB H100 SXM5 GPUs interconnected with 3.2T InfiniBand.

Voltage Park is planning to set up its clusters in Texas, Virginia and Washington.

The success of OpenAI’s ChatGPT over the past year has led to high demand for Nvidia’s GPUs in particular, leading to a surge in profits at the company and pushing its market value over $1tn in May.