Intel has quietly shifted strategy on one of its most important manufacturing initiatives. The company's 18A chip node, once reserved exclusively for Intel's own processors, is now being aggressively marketed to external customers as a foundry service. The change signals both progress and persistent uncertainty about whether the business model can work at scale.
Intel CFO David Zinsner told analysts at the Morgan Stanley Technology Conference that CEO Lip-Bu Tan has reversed an earlier decision to focus foundry efforts on 14A and keep 18A internal, recognising the node is "actually a good node to offer to external customers as well." Zinsner said Intel has been "getting some kind of inbound interest in 18A-P as a foundry node," describing it as "pretty positive."
This represents a significant vote of confidence. When Tan joined Intel in early 2025, both the functional and parametric yields of chips made using 18A were low and unpredictable. The fact that yields have apparently improved enough to justify external customer pitches is material progress. The 18A node delivers transistor density and performance roughly equivalent to TSMC's N2 process and positions Intel to potentially recapture manufacturing leadership for the first time since 2016.
Yet there is a notable elephant in the room: Nvidia. Intel and Nvidia announced broad collaborations combining Intel CPUs with Nvidia GPUs in both AI servers and consumer PCs, raising the question of whether manufacturing would be part of the deal. The short answer appears to be no, at least for now. Nvidia recently tested whether it would manufacture its chips using Intel's 18A process but stopped moving forward, according to two people familiar with the matter.
Zinsner emphasised that the Intel-Nvidia deals are about products, not manufacturing, saying "It was a product-driven engagement between the two CEOs," focused on customers "looking for an x86 solution." This distinction matters. Without Nvidia as an anchor customer, Intel's broader foundry ambitions lack a marquee name.
The strategic positioning reflects a genuine industry shift. Unlike TSMC, which only makes chips for outside clients, Intel also makes devices powered by its own chips, positioning it as a competitor to potential customers. As one Harvard Business School professor noted: "If I'm an Nvidia or AMD or Qualcomm or Broadcom, do you really want to put your secret sauce into a manufacturing operation where you're giving Intel access to that secret sauce?"
This structural conflict has not disappeared. But Intel is moving forward regardless. The world's two largest cloud-service providers have announced products using Intel 18A technology, part of nine total announced Intel 18A awards. While no major fabless chip designers have made publicly-announced commitments to the node, it is possible that some American chip developers may outsource production to Intel in coming years to reduce geopolitical risks associated with Taiwan and tariffs on offshore chips.
The pragmatic view is that Intel faces genuine constraints. The 18A process is real, yields are improving, and demand for non-TSMC manufacturing capacity is genuine, particularly among US government-backed projects. Nvidia's decision to test but not commit is not a failure; PDK sampling is routine in foundry relationships. The real test will come when other major customers move beyond talks to actual tape-outs.
For Intel, the foundry business remains a multi-year bet. The company has the manufacturing assets, the technology appears competitive, and the geopolitical tailwind is real. But without major household-name customers committing volume, the foundry division remains a strategic aspiration rather than a thriving business.