In a landmark shift for the global semiconductor landscape, recent reports from late January and early February 2026 indicate that NVIDIA (alongside Apple) is planning to move portions of its 2028 chip production to Intel Foundry.
The move is seen as a “dual-foundry” strategy designed to diversify supply chain risks and align with U.S. manufacturing mandates, while still maintaining a primary relationship with TSMC.
1. The “Feynman” Architecture Partnership
NVIDIA’s next-generation Feynman GPU architecture, slated for a 2028 debut, is at the heart of this rumored collaboration.
- Strategic Split: While the highly complex GPU compute dies will reportedly stay with TSMC, NVIDIA is looking to outsource the I/O (Input/Output) dies to Intel.
- The Process Nodes: NVIDIA is evaluating Intel’s 18A (1.8nm) and upcoming 14A (1.4nm) processes. The 14A node, which recently saw the installation of the industry’s first commercial High-NA EUV tools at Intel, is considered the primary target for 2028 mass production.
- Investment Backdrop: This partnership follows NVIDIA’s massive $5 billion equity investment in Intel in late 2025, which signaled a thawing of the long-standing rivalry between the two tech giants.
2. Advanced Packaging: The $1 Billion Opportunity
A critical component of the deal involves advanced packaging, a current bottleneck in the AI supply chain.
- EMIB Technology: NVIDIA reportedly plans to allocate up to 25% of its advanced packaging needs for the Feynman series to Intel, utilizing Intel’s EMIB (Embedded Multi-die Interconnect Bridge) technology.
- Financial Impact: Intel’s CFO, David Zinsner, recently noted that individual advanced packaging engagements are now trending “well north of $1 billion,” a figure that aligns with the scale of NVIDIA’s projected 2028 needs.
- Geographic Shift: This work would likely be handled at Intel’s sophisticated packaging facilities in New Mexico and Arizona.
3. Why Now? Geopolitics and Yields
The shift is driven by a mix of political pressure and manufacturing reality:
- U.S. Manufacturing Mandates: Both NVIDIA and Apple are facing pressure to secure domestic, “all-American” supply chains to avoid potential tariffs and geopolitical disruptions in the Taiwan Strait.
- Capacity Constraints: TSMC’s CoWoS packaging capacity is currently stretched to its limit by the global AI boom, leaving NVIDIA with little choice but to find a secondary high-volume partner.
- Intel’s 18A Validation: The reports surfaced just as Intel’s 18A process reached High-Volume Manufacturing (HVM) status in January 2026, boosting market confidence in Intel’s “five nodes in four years” roadmap.
4. Market Reaction & TSMC Response
- Stock Surge: Intel shares jumped over 11% following the reports, as the “Feynman” deal provides the high-profile validation the company’s foundry business has desperately needed.
- TSMC Strategy: Industry observers suggest TSMC views these “non-core” order diversions as a benefit, as it reduces the company’s monopoly risk and regulatory scrutiny without threatening its lead in high-end compute silicon.
Conclusion: A Pragmatic Rebalancing
NVIDIA’s 2028 shift to Intel is not a total departure from TSMC, but rather a pragmatic rebalancing of the silicon ecosystem. By offloading “non-core” I/O dies and a portion of its packaging to Intel, NVIDIA secures a domestic “pressure valve” for its most critical products. For Intel, securing a titan like NVIDIA marks its transition from a struggling legacy manufacturer to a credible, world-class foundry.
