NVIDIA plans to utilize Intel Foundry nodes for its upcoming "Feynman" AI GPU, starting in 2028, according to a report. This move follows NVIDIA's recent investment in Intel, signaling a strategic shift in its manufacturing partnerships. Specifically, NVIDIA is interested in Intel's 18A and 14A nodes.

The "Feynman" GPU will primarily rely on TSMC for its core IP, using the A16 (1.6 nm) node. However, Intel will handle the chiplet interconnectivity using its EMIB technology, with final packaging taking place in the U.S. This collaboration aims to leverage Intel's advanced packaging capabilities.

This partnership could allow NVIDIA to potentially circumvent import tariffs. The "Feynman" is expected to incorporate newer memory standards like HBM4e or HBM5. The move underscores the growing competition in the semiconductor foundry market, with Intel aiming to regain its manufacturing prowess.

With AI GPUs in high demand, this deal is a win-win for both companies. NVIDIA gains access to advanced manufacturing processes, and Intel secures a major customer, showcasing its ability to compete with TSMC. What remains to be seen is how this will affect the overall GPU market.