GMI Cloud, a Mountain View company that rents out Nvidia GPU servers, has raised $668m in new financing, the company said on Wednesday. That is $223m of Series B equity and a $445m credit facility. The Taiwanese bank CTBC led the credit facility.

ARCHIV, a new San Francisco firm that invests in AI and robotics, led the equity round, according to the press release. Nvidia took part, alongside investors from Asia-Pacific including DSC Investment, Trend Micro, KB Investment, Kyobo Life and KT Corporation.

“AI is driving a new renaissance, and reliable compute is its foundation,” said Alex Yeh, GMI Cloud’s founder and chief executive.

Contracted revenue above $600m

GMI Cloud’s contracted annual recurring revenue tops $600m, the company said. That is more than nine times its level at the end of 2025. The revenue already live in production has grown more than 4.5 times over the same period.

Its inference platform processes about 4 trillion tokens a week. Customers include Fireworks, Higgsfield, Nous Research, OpenRouter, Reflection, Cartesia, Trend Micro and Utopai Studios.

The money will add GPU capacity in the US, Taiwan and the rest of Asia-Pacific. GMI Cloud opened its first Asian facility, in Taiwan, in 2025. It announced a sovereign AI project in Japan earlier this year. Most of the world’s AI servers are made in Taiwan, and the company says its ties there help it deliver on time.

“In AI infrastructure, a delivery date is a promise,” Yeh said.

One of Nvidia’s seven

GMI Cloud builds its facilities entirely on Nvidia’s recommendations, Phoebe Liu reported for The Information. Only seven cloud providers do so, according to the report. The Information said the round shows Nvidia is still helping smaller cloud providers.

Earlier this month, Nvidia held $99bn in shares of companies that buy its chips. On Tuesday, the Financial Times reported that it had talked to insurers about loans backed by chips. In August, Nvidia paused parts of a revenue-sharing programme it launched in July.