Intel is joining Elon Musk's $25 billion AI chip manufacturing effort, adding semiconductor fabrication experience to a project that has been short on specifics since it was announced.
The partnership brings Intel's foundry operations into the picture - the division that manufactures chips for outside customers using Intel's own production facilities. For Musk's AI company xAI, which has been aggressively building compute infrastructure to compete with OpenAI and Anthropic, getting chips made fast enough has been a real constraint. The company's Colossus supercomputer in Memphis already runs over 100,000 Nvidia H100 GPUs, and building custom silicon would reduce that Nvidia dependency over time.
Intel Needs This as Much as xAI Does
Intel has been under pressure on two fronts: losing ground to TSMC on advanced process nodes (the manufacturing technology that determines chip speed and energy efficiency), and struggling to attract major customers to its U.S. foundry business. Being named in a $25B AI infrastructure project is exactly the kind of high-visibility deal that helps Intel position its foundry services as a serious alternative - particularly for U.S.-based AI companies looking to diversify away from overseas manufacturing.
The CHIPS Act has funneled federal funding into Intel's domestic fab expansion, but government money alone doesn't build a customer base. Landing xAI as a partner does more for Intel's foundry credibility than almost any other announcement it could make right now.
The $25B figure covers the full scope of the manufacturing ambition, not just Intel's piece. What chips they're actually designing, which process node they'll use, and when production could start have not been disclosed publicly. For xAI, the goal is clear - most AI companies at the frontier eventually want what Google has with its TPUs: control over your own chip supply, your own cost curve, and your own roadmap instead of waiting on Nvidia's next release cycle.