CoreWeave Inc., a cloud provider of graphics cards for artificial intelligence workloads, said today that it has added a $650 million credit line led by major investment firms to scale its global operations.

JPMorgan Chase, Goldman Sachs and Morgan Stanley led the credit facility, with participation from Barclays, CitiGroup, Deutsche Bank, Jefferies, Mizuho, MUFG and Wells Fargo.

Before raising the credit facility, the company had raised $12.7 billion in equity and debt over the past 18 months, including a $1.1 billion round in May valuing the company at $19 billion. “This credit facility provides additional liquidity to accelerate our growth strategy and capitalize on new opportunities in the rapidly evolving AI space,” said Mike Intrator, co-founder and chief executive of CoreWeave. 

CoreWeave operates a public cloud that provides access to Nvidia Corp. graphics processing units, targeting two main use cases: artificial intelligence and graphics rendering. Some of its processors include the powerful Nvidia H100 and H200 GPUs, which are built to provide accelerated computation power for AI workloads.

Training and deploying large language model AIs require a tremendous amount of compute. Larger and more complex models need to be spread over multiple high-performance GPUs. Big models are commonly used for cutting-edge tasks such as content creation, video production, question answering, document analysis, intelligent AI agents and other applications demanding advanced reasoning. CoreWeave says its infrastructure gives companies a software cloud scaling advantage for hosting their AI models.

Investments in AI companies are thriving as demand for AI and compute continues to grow. OpenAI raised a $4 billion revolving credit line earlier this month alongside a $6.6 billion funding round. Many of the same banks involved in CoreWeave’s credit facility also joined that line of credit.

The company recently opened a European headquarters in London and committed to invest $3.5 billion into expanding into Europe. This included opening two U.K. data centers and plans to build three new data centers in Norway, Sweden and Spain. CoreWeave currently operates 14 data centers, but with these new funds, the company expects to double that capacity to 28 operations globally by the end of 2024, with an additional 10 data centers planned for 2025.

The company also announced in August it was the first cloud provider to bring Nvidia H200 GPUs to its customers. The H200 is Nvidia’s next-gen Tensor Core GPU with double the performance compared to the H100 when handling LLMs such as Meta Platform Inc.’s Llama2.

Image: CoreWeave