Elon Musk Transforms xAI into a Computing Power Supplier: Cursor is the First Client

xAI, the artificial intelligence company founded by Elon Musk, plans to make part of its enormous computational capacity available to the startup Cursor.

According to various sources, Cursor will use tens of thousands of xAI's GPUs to train its new AI model dedicated to programming, called Composer 2.5. GPUs are essential for training AI models and represent one of the most sought-after and expensive resources in the current technology sector.

This collaboration marks a significant shift in xAI's strategy, as it moves increasingly toward the business model of major cloud providers like Amazon, Microsoft, and Google. These companies own millions of chips and generate enormous revenues by renting computing power to developers and businesses. Even new players like CoreWeave and Lambda have made their mark by providing GPUs for AI.

Opportunities for xAI and Elon Musk

For xAI, this move represents an opportunity to monetize the enormous infrastructure built over the past few years, reducing the operating costs of its data centers while strengthening relationships with a rapidly growing startup. Cursor, in fact, is valued at about $50 billion and is in direct competition with entities like OpenAI and Anthropic in the development of programming assistants.

At the heart of this strategy lies the Colossus project, through which xAI has rapidly expanded its infrastructure. The company already has around 200,000 NVIDIA GPUs and aims to reach one million.

However, challenges abound. Internally, xAI is seeking to improve the efficiency of its resources: the GPU utilization rate (MFU) is currently around 11%, well below the industry standard of between 35% and 45%. Improving this figure will be crucial for maximizing the return on infrastructure investments. In this context, allowing third parties access to computing power for a fee could be highly beneficial.