xAI to Provide Massive GPU Computing Power to Cursor for AI Code Model Training

Gate News message, April 17 — Elon Musk’s AI company xAI plans to provide large-scale computing power to programming tool startup Cursor, with Cursor expected to use thousands of GPUs acquired by xAI from suppliers like Nvidia to train its next-generation code model Composer 2.5.

This collaboration marks xAI’s transition from a single-focus model research company to a role encompassing computing power services, positioning itself as a potential alternative to major cloud service providers including Microsoft, Google, CoreWeave, and Nebius. For xAI, renting out computing capacity can alleviate the high costs of data center construction and operations while strengthening connections with developer ecosystems. Cursor, which possesses extensive developer data, holds strategic value as AI coding competition intensifies.

The partnership is expected to increase xAI’s revenue streams and address investor concerns regarding SpaceX’s ongoing IPO process. xAI, which completed its acquisition by SpaceX before the IPO, remains in a high-intensity R&D investment phase with monthly operating losses exceeding $300 million and net losses surpassing $4 billion in 2025. Improving resource utilization represents a critical priority—xAI’s current GPU utilization rate in model training stands at approximately 11%, significantly below the industry standard of 35% to 45%.

XAI has deployed approximately 200,000 GPUs as part of its “Colossus” data center project and plans to expand to 1 million units. Cursor, which generated $2 billion in annual recurring revenue last month with a run rate double that of three months prior, is valued at approximately $50 billion in ongoing financing discussions. The startup has grown into one of the fastest-growing companies in history since its founding less than five years ago, with approximately 60% of revenue derived from enterprise clients.