Big Chip Players Line Up For TSMC N2 Process But Intel Sits Out, Suggesting 18A Confidence

TSMC's next-generation N2 process has officially booked its first customers: Apple, MediaTek, and AMD, with NVIDIA reportedly paying a hefty premium to get pole position for A14. Notably absent from the confirmed list is Intel, which is striking precisely because Intel has relied on TSMC's 6nm, 5nm, and 3nm-class nodes for GPU, SoC, and even Compute tiles in its Arc graphics and Core Ultra CPU products. So why not hedge with N2? One possible answer boils down to confidence in its in-house 18A process.

If Intel doubted 18A's readiness, it would almost certainly have booked N2 capacity years in advance as insurance. Reports indicate that AMD and NVIDIA did just that, ensuring wafer allocation for 2026–2027 products. Intel has not yet been confirmed for early N2 allocation, which suggests that its leadership genuinely believes 18A will carry the company's crown jewels. The strategy is risky, but it signals a kind of all-in confidence.


Whereas TSMC N2 is a more evolutionary nanosheet design, Intel's 18A leap introduces both RibbonFET and PowerVia technologies at the same time. RibbonFET is Intel's take on gate-all-around transistors, while PowerVia is the world's first implementation of a backside power delivery network. Together, these could yield stronger perf/W gains than a straight density race. By focusing on 18A, Intel avoids fighting AMD, NVIDIA, and Apple on identical process footing and instead markets a unique technical narrative: "only Intel has both GAAFETs and backside power delivery."

There's also a broader geopolitical calculus to consider. Intel is the poster child for U.S. semiconductor independence—especially after the United States federal government acquired a 10% stake in the chipmaker. Leaning heavily on TSMC N2 would undermine the specific Intel Foundry Services messaging that has won Intel billions in subsidies. Betting on 18A instead strengthens Intel's image as the Western counterweight to TSMC.


For fabless firms like AMD, Apple, and NVIDIA, failing to secure N2 allocation is existential. These companies must have cutting-edge wafer supply to sustain their high-margin products. Intel, by contrast, has more flexibility. If 18A falters, it can still shift some products to TSMC later—as it already does with Arrow Lake tiles. However, if 18A hits its performance and yield targets, Intel gets both process bragging rights and structural independence. The upside outweighs the downside.

Intel CFO David Zinsner recently acknowledged that 18A yields dipped due to tweaks aimed at boosting Panther Lake performance, but reports indicate recovery is already underway. Intel is expected to debut both Panther Lake laptop chips and Diamond Rapids server CPUs on 18A. Further out, Nova Lake—slated for 2027—will be Intel's attempt to reassert desktop performance leadership. Leaks point to as many as 52 cores enabled by dual compute tiles (each 8 P-cores + 16 E-cores), plus four low-power E-cores on the SoC tile. This aligns against AMD's rumored Zen 6 chips on TSMC's N2P/N2X processes, potentially pushing core count up to 24 cores and frequencies up to 7 GHz.

Intel not being on the confirmed N2 launch roster is probably a double-down bet. If RibbonFET + PowerVia deliver on their promises, Intel could regain process credibility and restore competitive balance in both client and server markets. If execution slips, however, AMD and NVIDIA will enjoy the volume and maturity advantages of TSMC N2 while Intel once again struggles with credibility. At this stage, Intel's apparent absence from early N2 allocation looks like a high-stakes declaration that 18A isn't just alive, but central to Intel's comeback story.