Artificial intelligence cloud-computing infrastructure startup CoreWeave Inc. today said it has raised $7.5 billion in debt financing.

The funding was led by Blackstone Inc. with participation from Carlyle Group and BlackRock Inc. in one of the largest private debt funding rounds to date. It follows hot on the heels earlier this month of the company’s Series C equity financing round of $1.1 billion, which valued the company at $19 billion.

As a supplier of graphics processing units for training and deploying AI models, CoreWeave has set itself up as a central figure in the AI revolution. It is a primary provider of Nvidia Corp.’s high-powered AI processing chips as demand for computing power soars. GPUs are essential for the preparation of AI models for production and running them in production, such as OpenAI’s ChatGPT, which launched in 2022 and set the entire internet alight with its ability to hold human-like conversations.

“It is a monumental financing in nominal terms, but it’s also a monumental financing in terms of the extent to which it drives the company forward,” CoreWeave Chief Executive Michael Intrator told Wall Street Journal, speaking about the funding.

The company operated 14 data centers at the end of 2022 and has plans to double that number by the end of this year. Many of the GPUs that the company offers in its centers are Nvidia’s H100, which are built from the ground up for processing AI workloads – such as expensive training and running in production. The company also offers access to A40 GPUs, which, unlike the AI-optimized cards, provide access to RT-cores for graphical processing capabilities and ray tracing, which is necessary for rendering 3D special effects.

Intrator said that much of the new funding would go toward building out the company’s footprint in new data centers and purchasing the expensive GPUs to fill them to keep up with the increasing demand for both AI and graphics processing.

Recent developments in artificial intelligence have pushed models even further than they started in 2022. ChatGPT began as a text-to-text-only model, but a recent announcement from OpenAI now gives it powerful multimodal capabilities giving it the ability to hold conversations in voice and understand video. Within a day, Google LLC announced a similar set of features for its Gemini AI model with Project Astra.

As these advancements continue to expand the complexity of AI models, so will the demands on the infrastructure that will run them. As a result, investors are flocking to the cloud compute companies that will power the next boom in AI capabilities.

“I don’t think anyone would have expected another hyperscaler was going to come into the market this quickly,” Intrator added.

Image: CoreWeave