Intel Reportedly Wins Slice Of NVIDIA's 2028 GPU Chip Packaging In Strategy Shift

Intel may take on a limited role in manufacturing and packaging NVIDIA's next-next-generation GPUs, according to a report from DigiTimes Asia. Citing supply chain sources, DigiTimes reports that NVIDIA plans to partially involve Intel in production of its post-Rubin "Feynman" GPU architecture, expected in late 2028. Under the reported arrangement, the primary GPU compute die would remain with TSMC, while portions of the I/O die and up to 25% of final advanced packaging could shift to Intel.

While this makes sense—especially given the excellent efficiency of the just-released Panther Lake processors partially fabricated on Intel's 18A process—it's not for reasons of superior process technology. Instead, it's really about politics and the appearance of diversification.

The report says Intel's contribution would rely on either its cutting-edge 18A process or its upcoming 14A node, with advanced packaging handled via Intel's EMIB technology. TSMC would reportedly retain responsibility for the remaining 75% of packaging. Any use of Intel's leading-edge fabrication would supposedly be contingent on yield improvements and limited to non-core silicon; NVIDIA and Intel have not commented on the report.

Assuming DigiTimes' sources are legit, this move would represent a strategic diversification rather than a real shift away from TSMC. DigiTimes frames the reported plan as part of a broader dual-foundry strategy driven by political pressure, supply-chain resilience goals, and constraints in advanced packaging capacity.


This builds directly on the landmark collaboration NVIDIA and Intel announced back in September, where the companies committed to jointly developing multiple generations of custom products. For datacenters, Intel is designing and manufacturing custom x86 CPUs tailored for NVIDIA's AI infrastructure platforms (with tight NVLink integration). For personal computing, Intel will produce x86 SoCs that integrate NVIDIA RTX GPU chiplets, creating "Intel x86 RTX SoCs" for compact PCs with high-performance NVIDIA graphics.

Of course, that deal also included a $5 billion USD investment in Intel, slotting the House of GeForce in as a major shareholder. Intel has increasingly positioned itself as a domestic manufacturing option aligned with U.S. industrial policy—particularly following the U.S. federal government's acquisition of a 10% stake in the company last year.

Using Intel for limited, lower-risk portions of a future GPU design (like I/O silicon and partial packaging) could help NVIDIA demonstrate supply-chain diversification to the U.S. government while keeping its most performance-critical silicon with TSMC, as the Taiwan foundry simply has a more proven track record. Along those same lines, restricting Intel's role to I/O silicon and partial packaging limits exposure to execution risk—a key concern given Intel's limited history as a contract foundry.

DigiTimes Asia also reports that Intel is working with Apple on a deal where Intel might fabricate a new entry-level processor in the M series for price-conscious MacBooks. This, too, seems more motivated by policy and an attempt to make a show of diversification than a true shift in strategy, but it's fascinating considering the reportedly bitter nature of the split between the two companies back in 2020.

As with most supply-chain reporting, the details remain speculative; while DigiTimes has a strong history with early semiconductor leaks, stories like this are always best read as indicators of direction rather than finalized plans.