NVIDIA is reportedly considering outsourcing semiconductor production to Intel in 2028, as well as a partial cooperation between the companies on its next-generation Feynman GPU architecture.
In a new report from DigiTimes picked up by analyst @Jukan on X, we're hearing that thanks to TSMC pretty much dominating advanced processes and advanced packaging, it is a new target for the United States.
Chip companies are having to deal with the 'heavy responsibility' of American manufacturing, with cost issues and capacity shortages. The current model of highly concentrated semiconductor production with TSMC over the last few years has moved towards a new strategy: 'multi-sourcing and risk diversification'.
- Read more: NVIDIA unveils next-gen Feynman GPU in GTC 2025 roadmap, HBM5 in 2028
- Read more: NVIDIA GTC 2026 on March 16-19: expect Blackwell, Rubin, Feynman GPUs details
New rumors floating through the semiconductor supply chain suggest that NVIDIA's next-gen Feynman architecture platform debuts in 2028, and Intel will reportedly partially cooperate with NVIDIA on Feynman. NVIDIA and Intel's mindset is reportedly "low-volume, low-end, and non-core" being the primary strategy for their purported semiconductor partnership.
NVIDIA and Intel reportedly hope to meet the aims of the Trump administration, without hurting their cooperative relationship with TSMC... however, they're adopting a dual-foundry model, with the aim being to minimize mass-production risks.
The new cooperation between NVIDIA and Intel on Feynman is not a total surprise, as NVIDIA recently announced a $5 billion investment into Intel in September 2025, with the plan being to work with Intel on the Feynman architecture chip, which succeeds the just-arriving Rubin architecture.
- Read more: NVIDIA inks $5B deal with Intel for x86 SoCs with RTX GPUs for consumers
- Read more: NVIDIA CEO: 'Intel was on a mission to kill us for 33 years'
- Read more: Acer CEO insight on 'operational headaches' from NVIDIA + Intel x86 CPU deal
- Read more: Intel wants Microsoft + Tesla + NVIDIA as advanced packaging customers
- Read more: NVIDIA N1X + N1 gaming laptop chips rumored for debut soon, will fight x86 processors
The reports suggest the GPU die for Feynman will be outsourced to TSMC, while the new I/O die will "partially" adopt Intel 18A or Intel 14A, with 14A coming online for mass production in 2028... just in time for Feynman. The chip will be handed over to Intel's EMIB for advanced packaging, with the report stating that based on the final advanced packaging proportion, Intel would take care of about 25% while TSMC still eats most of it up with around 75%.



