OpenAI to invest billions in CoreWeave as tensions with Microsoft continue growing

CoreWeave recently published a press release confirming it has struck a timely deal with OpenAI, which will benefit both companies greatly. On the one hand, the investment, which contemplates up to $11.9 billion spread across five years, will boost investor confidence as CoreWeave prepares for a much-anticipated IPO. On the other, OpenAI will become one of CoreWeave's biggest customers, if not the largest. As part of the agreement, OpenAI will now have a stake in CoreWeave worth $350 million in shares.

Interestingly enough, before the investment and accompanying negotiations took place, Microsoft was CoreWeave's main customer, accounting for more than two-thirds of the latter's $1.9 billion-worth revenue in 2024. Investors tend to be cautious when a company seems over-reliant on a single major customer. Thus, signing on another, even more profitable customer such as OpenAI is bound to favor CoreWeave on its journey towards becoming a publicly listed company.

As for OpenAI, the company has long claimed that its product timelines are being delayed due to a lack of compute. This issue reportedly became a persistent source of tension between the startup and Microsoft, an important investor, and until recently, exclusive provider of OpenAI's compute infrastructure. Since the Stargate Project was announced, it became clear that Microsoft would no longer be OpenAI's exclusive compute provider. The current deal with CoreWeave, is a bold message from OpenAI's part, as it gives the startup direct access to at least part of the cloud resources it used to get from Microsoft.

But more importantly, some consider that the relationship between Microsoft and OpenAI started eroding when Microsoft hired Sam Altman's rival, Mustafa Suleyman, to lead its AI initiative. Reportedly, Microsoft is developing proprietary foundation models to compete with recent launches from rival companies such as OpenAI or Anthropic. The models, known as MAI, have been tested on their performance powering the Microsoft Copilot assistants, and may at some point replace the models currently used for the product: a mix of models, including some by OpenAI. Microsoft is also said to be working on its own 'reasoning' models, the latest trend in the industry to deliver scalable performance.