Intel chasing foundry relevance
Intel says it is moving from roadmap promises to production with an aggressive push for its 18A node and is courting external foundry customers as part of a turnaround strategy. Market coverage suggests commercial production progress and advanced packaging plays are changing sentiment about Intel’s ability to compete beyond its historical CPU niche ( ).
Intel is trying to prove it can do two jobs at once again: design its own chips and manufacture chips for other companies. After years of delays, it is now saying its 18A process is ready for customer projects, which is the point where a roadmap turns into something sales teams can actually pitch. (intel.com) A foundry is a chip factory for hire. Taiwan Semiconductor Manufacturing Company became the default choice because companies like Apple, Nvidia, and Advanced Micro Devices could design chips without also owning the multibillion-dollar factories that print them. (britannica.com, tsmc.com) Intel used to be the opposite model. It built most of its own central processing units in its own fabs, but manufacturing slips let Taiwan Semiconductor pull ahead, and Intel ended up outsourcing some flagship products to the rival it once tried to outrun. (reuters.com, msn.com) The 18A name is Intel’s label for its next big manufacturing step, and the company says it brings up to 15 percent better performance per watt and 30 percent higher chip density than Intel 3. Intel also says 18A uses backside power delivery, which is like moving a building’s plumbing into the walls so the rooms above can be arranged more efficiently. (intel.com) The first real proof point is not an outside customer. It is Panther Lake, Intel’s next personal computer chip family, which Intel says is already in production and will enter high-volume manufacturing in Arizona later in 2025. (intel.com, newsroom.intel.com) That matters because foundry customers do not buy promises. If Intel can manufacture its own high-volume laptop chip on 18A first, it gives outsiders a working reference instead of a slide deck. (intel.com, intel.com) Intel is also selling something less glamorous and possibly faster to monetize: advanced packaging. Packaging is the step that connects several chip pieces into one working product, and modern artificial intelligence chips increasingly depend on that because memory, compute, and networking parts are often built separately and then stitched together. (arstechnica.com, intel.com) Intel’s two main packaging brands are Embedded Multi-die Interconnect Bridge and Foveros. Embedded Multi-die Interconnect Bridge acts like a short high-speed bridge between neighboring chip pieces, while Foveros stacks chip pieces vertically, more like adding floors to a building than spreading everything across one block. (intel.com, arstechnica.com) Chief Financial Officer David Zinsner said in March 2026 that Intel was close to packaging deals worth billions of dollars a year. That is important because packaging revenue can arrive before Intel wins the harder prize of full wafer production for a big outside chip designer. (theregister.com) The customer chase is now broad enough that Intel packed more than 1,000 customers and ecosystem partners into its Foundry Direct Connect event in San Jose on April 29, 2025. A foundry business lives or dies on software tools, design libraries, and trust, so Intel is trying to show it has an ecosystem, not just factories. (intel.com, businesswire.com) The bet investors are making is simple: if 18A works in volume and packaging wins start landing, Intel stops being only a central processing unit company with a comeback story and starts looking like a second Western alternative to Taiwan Semiconductor. Intel still has to prove yields, margins, and outside customer volume, but the conversation has shifted from “can it build this” to “who will sign first.” (intel.com, theregister.com, arstechnica.com)
