Manufacturers increasingly choose Intel Foundry as a selective second source
- Intel Foundry drew fresh attention this week after TrendForce said Apple is evaluating Intel 18A-P and Google is exploring EMIB packaging for TPUv8e. - The telling detail is selective use: Apple is tied to 18A-P process tech, while Google is linked to EMIB packaging rather than a full-node switch. - That matters because Intel now looks like a credible second source, not just a long-shot TSMC replacement.
Semiconductor manufacturing is usually discussed like a winner-take-all race. But this week’s Intel Foundry story is more interesting than that. The real shift is not that Apple or Google are about to dump TSMC overnight. It’s that Intel is starting to look useful as a selective second source — one process here, one packaging service there, one backup path for a future product. That is a much smaller claim, but it is also a much more believable one. (trendforce.com) ### What actually changed? TrendForce said on April 29 that Apple is evaluating Intel’s 18A-P node for future M-series chips, while Google is exploring Intel’s EMIB advanced packaging for a future TPUv8e program. Those are still reported evaluations, not announced production awards. But the pairing matters because it shows interest in two different parts of Intel’s offer — leading-edge wafers and advanced packaging. (trendforce.com) ### Why is “second source” the key idea? Because almost nobody serious expects a clean break from TSMC. The foundry market does not work like changing cloud vendors with a few clicks. Design teams build around a process, a packaging stack, a software flow, yield as(trendforce.com)e, or package. That is the psychology change here. (trendforce.com) ### Why would customers even want that? Two reasons. Capacity and leverage. AI demand has kept advanced manufacturing and packaging tight, and Intel has been arguing that inference-heavy systems are reviving CPU demand alongside accelerators. If you are Apple, Goog(trendforce.com)el to replace TSMC. You just need Intel to be good enough to matter. (trendforce.com) ### Why is packaging part of the story? Because packaging is where “some business” can become easier than “all the business.” Google’s reported interest is in EMIB — Intel’s bridge-based advanced packaging — not necessarily in moving the whole TPU design onto an In(trendforce.com)ing-led adoption can be the on-ramp. (trendforce.com) ### Is Intel actually more credible now? More than before, yes. Intel said 18A is in risk production and expected to reach volume manufacturing in 2025, and its April 23, 2026 results said foundry operating loss improved to $2.4 billion, helped by better yields acr(trendforce.com)st chips. That does not prove customer wins — but it does show the plumbing is getting less hypothetical. (intc.com) ### What is 18A-P, in plain English? Basically, it is an enhanced version of Intel’s 18A family. Reports around this week’s story describe 18A-P as the variant Apple has been watching, which fits the usual pattern: customers often wait for a process and its design kit to mature before committing real(intc.com)e settle down. (techpowerup.com) ### What would customers have to do next? A lot of boring but load-bearing work. Design-flow portability. PDK validation. IP remapping. Package-level architecture tradeoffs. Reliability qualification. If a company wants one product family to have both a primary and secondary manufacturing (techpowerup.com)umor can become a shipment. (intc.com) ### So what is the bottom line? Intel does not need to become “the new TSMC” for this story to matter. It just needs to become the foundry that big customers trust with selected jobs. That is a lower bar — but also a strategically important one. If Apple or Google gives Intel even partial business, the bigger signal is that multi-sourcing at the leading edge is starting to feel practical again. (trendforce.com)
