Despite a previously announced $5 billion strategic collaboration, the high-profile foundry partnership between Intel and NVIDIA has reportedly hit a roadblock. Sources indicate that NVIDIA has completed its sampling and testing of Intel's 18A process technology but will not proceed further with it for manufacturing. This development underscores the complex and evolving dynamics in the competition for advanced semiconductor nodes.
Intel’s 18A process (roughly 1.8nm-equivalent) is a cornerstone of its IDM 2.0 strategy. It incorporates next-generation technologies like RibbonFET transistors and PowerVia backside power delivery, optimized for high power efficiency, particularly for Intel's own upcoming mobile processors. According to industry analysis, the main reason for the halt is a mismatch in technical focus: NVIDIA's primary need for its high-performance computing (HPC) and AI chips is peak performance, whereas Intel's 18A is tuned for power efficiency.
It is standard industry practice for major chip designers like NVIDIA to test process design kits (PDKs) from multiple foundries. NVIDIA has reportedly made TSMC's N2 process its preferred choice for the 2nm generation. Intel had previously indicated that widespread external adoption of its foundry services would likely begin with the subsequent generation, 14A process, which is more targeted toward high-performance applications. The original $5 billion agreement between the companies was reportedly centered on collaboration in the x86 ecosystem, not a binding foundry commitment.
ICgoodFind's Insight
The pause in NVIDIA's 18A testing highlights the significant execution challenges and market skepticism facing Intel Foundry. While not a terminal blow to their partnership—with future collaboration on 14A still possible—it reinforces TSMC's entrenched position as the de facto standard for cutting-edge, performance-first chip manufacturing.
