Key Takeaways
1. Samsung Foundry struggles with stable yields for SF2, leading companies to prefer TSMC for advanced chip production.
2. Apple and Nvidia are considering Intel’s 14A node for future products, with mass production expected in 2028.
3. TSMC continues to meet Apple’s current chip needs, with upcoming products transitioning between N3P and N2 nodes.
4. Nvidia plans to use a custom TSMC 3 nm variant for its next-gen AI chips, while Intel’s 14A may be reserved for specific lower-end GPUs.
5. TSMC and Intel are the main contenders for 1.4 nm production, as Samsung may withdraw from this competition, and Rapidus could emerge as a 2 nm alternative.
With Samsung Foundry facing challenges in achieving stable yields for SF2, TSMC remains the primary choice for many companies needing advanced chips. However, rising wafer costs may push some firms to consider other options for their chip production. Major players like Apple and Nvidia are now eyeing a lesser-known contender in the semiconductor foundry scene: Intel.
Potential Shift to Intel
According to a recent research note from industry analysts, as reported by Reuters, Nvidia and Apple may shift to Intel’s 14A node for upcoming products. Intel’s timeline indicates that 14A is not expected to begin risk production until 2027, with mass production slated for 2028, assuming there are no additional delays. These speculations are not entirely fresh, as previous reports suggested Apple might utilize Intel’s 18A for certain chips.
Apple and TSMC
Currently, TSMC is fulfilling Apple’s immediate chip requirements, with the A19 and M5 series being produced on TSMC’s N3P node. In 2026, the A20 and M6 will move to N2, likely staying there for about a year alongside the A21/M7. However, it seems that only the Apple M8 could transition to 14A, while the A22 is expected to remain with TSMC.
Nvidia’s Future Plans
As for Nvidia, its upcoming Rubin architecture is projected to use a tailored TSMC 3 nm variant (probably 3N) for its AI chips and RTX 60 series desktop GPUs, anticipated to launch around 2027 if Nvidia’s plans hold. The successor to Rubin, the Feynman architecture, set for 2028, could potentially utilize Intel’s 14A. Yet, Nvidia isn’t entirely abandoning TSMC, as 14A will only be designated for certain ‘low-end’ GPUs.
TSMC’s next-gen N2 successor, A14, is not expected to debut until 2028, making it premature to predict which original equipment manufacturers (OEMs) will adopt it. Samsung might withdraw from the 1.4 nm competition due to its renewed efforts to make SF2 more appealing to OEMs. This leaves TSMC and Intel as the two foundries capable of 1.4 nm production, although the Japanese company Rapidus may also present itself as a possible 2 nm alternative by that time.
Source:
Link
Leave a Reply