Nvidia's Next-Gen GPU Could be Coming to Intel Foundry

Core Viewpoint - Intel is potentially on the verge of securing a significant foundry contract with Nvidia, which could greatly enhance its foundry business, particularly in the context of AI chip manufacturing [1][5]. Group 1: Nvidia's Interest in Intel Foundry - Nvidia is reportedly considering Intel's foundry for its upcoming Feynman generation of GPUs, expected to launch in 2028, due to supply constraints at TSMC [2]. - Intel will not manufacture the GPU compute die, which will remain with TSMC, but will use its 18A or possibly 14A process for parts of the I/O die, contingent on yield improvements [3]. - The I/O die, while less performance-critical than the GPU compute die, still requires advanced processes, and Intel will also manage a portion of the advanced packaging for the Feynman GPUs [4]. Group 2: Advanced Packaging Opportunities - Intel's CFO indicated that advanced packaging deals could exceed $1 billion, significantly higher than previous expectations, aligning with the potential Nvidia contract [6][7]. - If Nvidia utilizes Intel for 25% of its advanced packaging needs, it could represent a deal worth over $1 billion for Intel, with multiple customers potentially engaging in similar opportunities [7]. - TSMC's advanced packaging services are reportedly unable to meet demand, suggesting that Intel's advanced packaging could attract foundry customers, leading to expanded manufacturing relationships [8]. Group 3: Market Position and Future Prospects - Analysts are optimistic that Apple may also use Intel's 18A process for some M-series chips and potentially the 14A process for future iPhone chips, indicating positive momentum for Intel's foundry ambitions [9]. - Intel requires significant external customers for its foundry to justify the substantial investments needed to compete with TSMC, and recent hiring at its Ohio fab complex suggests growing confidence in securing orders [10].