Intel Targets 8% Monthly Yield Gain in Foundry Turnaround

Intel Drives 8% Monthly Yield Gains on 18A Process Node

Intel board member Lip-Bu Tan confirmed on February 3, 2026, that the company is executing a complex turnaround of its foundry business, marked by significant manufacturing improvements. After describing initial production yields for the crucial 18A process node as "quite poor," Tan stated that a focused effort, including help from industry partners like PDF Solutions and KLA, is now delivering 7% to 8% yield improvement per month. This progress is a critical proof point in Intel's strategy to regain its technological edge.

Tan noted that the tangible improvements are already attracting external interest, with "a couple of customers" showing interest in the 18A node. The ability to demonstrate consistent, high-volume manufacturing performance is central to Intel's plan to operate a general-purpose foundry capable of competing with industry giants for external contracts, moving beyond serving only its own product lines.

14A Node Set for 2028 as Intel Courts Foundry Clients

Looking beyond the current generation, Intel is "laser-focused" on its 14A process node, its most advanced planned technology. The company has established a clear roadmap, scheduling risk production for 2028 and full volume production for 2029. To accelerate customer adoption, Intel is releasing a preliminary process design kit (PDK) this month, allowing potential clients to begin working with the technology on test chips.

This customer engagement is pivotal to the foundry's success. Tan explained that Intel hopes to see volume commitments from new customers in the second half of the year. The strategy involves earning trust incrementally by encouraging clients to start with small production allocations—such as 5% or 10% of a key product—before scaling up. Tan also emphasized that observers can track Intel's success by watching for capital investments in materials like glass substrate, which would signal that "real customers" have committed to the platform.

AI Memory Bottleneck Looms Until 2028

Tan provided a broader view of the technology landscape, identifying high-bandwidth memory as the single biggest constraint to the growth of artificial intelligence, stating there will be "no relief until 2028." He pointed to surging compute demand from customers and other bottlenecks, including processor thermal issues, data interconnects, and the software stack required to manage large AI clusters. In response, Intel is embracing architectural flexibility beyond its traditional x86 dominance, including RISC-V and Arm, and hiring top talent to build out its future GPU capabilities.

He also issued a warning about global competition, describing China's progress as a "wake-up call." Citing the deep talent pool at companies like Huawei, which he claimed has "100 CPU architects" who are "top-notch," Tan cautioned that China is "shortly behind" the U.S. and could "leapfrog" if American firms are not careful. This competitive pressure underscores the urgency of Intel's mission to re-establish its manufacturing and design leadership.