NVIDIA’s “Feynman” AI GPU to Leverage Both TSMC and Intel Foundry Technologies
NVIDIA is preparing to launch its next-generation “Feynman” AI GPU in 2028, and according to a recent DigiTimes report, the company is set to diversify its manufacturing strategy by incorporating both TSMC and Intel Foundry technologies. This move marks a significant shift in NVIDIA’s approach to chip production, aiming to optimize performance and supply chain resilience for its future AI hardware.
Strategic Partnership with Intel Foundry
NVIDIA has recently acquired a 5% stake in Intel, signaling a deepening collaboration between the two tech giants. The report highlights that NVIDIA is exploring the use of Intel’s advanced foundry nodes—specifically Intel 18A and 14A—for select components of the “Feynman” AI GPU. This partnership is not only about manufacturing; NVIDIA also plans to utilize Intel’s EMIB (Embedded Multi-die Interconnect Bridge) technology to enhance chiplet interconnectivity, a critical factor for high-performance AI accelerators.
TSMC to Remain the Primary Manufacturing Partner
Despite the new collaboration with Intel, TSMC will continue to play a central role in the production of the “Feynman” GPU. The core intellectual property of the chip will be fabricated using TSMC’s advanced A16 (1.6 nm) process node, which is expected to account for approximately 75% of the chip’s overall value. This ensures that NVIDIA’s flagship AI GPU will benefit from TSMC’s leading-edge semiconductor manufacturing capabilities.
Advanced Packaging and Memory Technologies
The remaining silicon content of the “Feynman” GPU will be produced on Intel’s 18A or 14A nodes, with final chip packaging taking place at Intel Foundry facilities in the United States. Intel will be responsible for the EMIB chiplet bonding process, which is essential for integrating multiple dies into a single, high-performance package.
In addition to manufacturing innovations, the “Feynman” AI GPU is expected to adopt next-generation memory standards such as HBM4e or potentially HBM5. This upgrade will significantly increase the memory capacity per package, enabling each GPU to efficiently handle trillion-parameter scale AI models—a key requirement for future advancements in artificial intelligence and machine learning.
Looking Ahead
NVIDIA’s decision to leverage both TSMC and Intel Foundry technologies for its upcoming “Feynman” AI GPU reflects a strategic effort to combine the strengths of multiple industry leaders. By integrating advanced process nodes, innovative packaging solutions, and cutting-edge memory standards, NVIDIA aims to deliver a powerful platform capable of supporting the next wave of AI breakthroughs.
