NVIDIA Tests Intel's 18A Process, Opts Against Mass Production

NVIDIA Explores Intel's 18A Process, Ultimately Declines for Mass Production

In a move that sent ripples through the semiconductor industry, NVIDIA, the undisputed titan of AI and graphics processing, reportedly tested Intel's cutting-edge 18A manufacturing process. However, after initial evaluations, the graphics giant ultimately decided against adopting the Intel Foundry service for large-scale chip production. This revelation, brought to light by Reuters and confirmed by sources familiar with the matter, paints a complex picture of the evolving foundry landscape.

A Standard Industry Practice, Not Necessarily a Red Flag for Intel

It's crucial to understand that such exploratory steps are not uncommon in the high-stakes world of semiconductor manufacturing. For any major chip designer, like NVIDIA, meticulously evaluating alternative fabrication technologies is standard operating procedure. This process is akin to a chef tasting various ingredients before committing to a new signature dish. Before locking in massive production capacities, companies like NVIDIA must benchmark potential partners against established giants like TSMC, its current primary fabrication partner. Therefore, NVIDIA's dip into Intel's 18A pool doesn't automatically signal a fundamental flaw in Intel Foundry's capabilities. It's a pragmatic assessment of options in a fiercely competitive market.

Intel 18A: A Process with Internal Focus and Future Ambitions

Currently, the 18A process is largely perceived as a technology tailored for Intel's own internal product lines. Interestingly, there were even periods when Intel itself contemplated shelving this particular process. However, Intel Foundry's strategic vision extends further. Future iterations, namely the 18A-P and 18A-PT variants, are specifically designed to court external clients and represent Intel's long-term commitment to its contract manufacturing business. These advanced nodes are geared towards more specialized applications, showcasing Intel's ambition to diversify its foundry offerings.

The Emerging Promise of Intel 14A

While the 18A process garnered NVIDIA's attention, the real buzz within Intel Foundry is increasingly centered around the upcoming 14A process, slated for a 2027 debut. This node is emerging as a cornerstone for Intel's contract manufacturing ambitions. Early engagement with potential customers is already underway, with companies actively involved in the design phase to assess how the 14A technology aligns with their specific requirements. Initial feedback from these early adopters has been remarkably positive, with the 14A node being lauded as genuinely competitive. This suggests Intel is making significant strides in developing a compelling offering for the broader market.

The NVIDIA-Intel Partnership: A Different Avenue

Parallel to these foundry developments, a separate, yet significant, partnership between Intel and NVIDIA is set to materialize in 2026. This collaboration will focus on integrating NVIDIA's RTX graphics cards into Intel's x86 system-on-chips (SoCs). The exact scope of this partnership, specifically whether it will encompass foundry services, remains unclear. For now, the agreement is strictly about the supply of finished products, not about NVIDIA outsourcing its chip fabrication to Intel. As Intel rolls out the 14A 0.5 PDK (Process Design Kit), the industry eagerly awaits further insights and potential commitments from prospective Intel Foundry clients, with a keen eye on the 2027 launch of the 14A and the specialized 18A variants.