$INTC doubled estimates and jumped after hours. By the next morning it was down 8%. Management framed a record server quarter and yields ahead of plan as proof the turnaround has arrived. Our post-earnings calls with five former Intel operators suggest two pieces are still missing.
First, foundry. All five sources engaged the external foundry claim, and none characterized $293 million of quarterly revenue as commensurate with the ambition. The proof they want is committed volume, not pilot work and PDK milestones.
A three-decade corporate veteran who worked closely with Intel Foundry: "And if the customers believed that they could really execute, that they would be able to get the products they need, then they would be sold out, right? And they're not sold out." Back in Q1 (April), a former product executive said the same thing differently: Intel needs to "prove that they can manufacture for somebody that isn't them." Three quarters of calls, same ask.
Second, 18A yields. Management says output exceeded targets by 25%. Sources did not dispute the number. They disputed what it means at scale. A former VP in Intel's high-performance computing and AI group named the mechanism: "A fab process can be completely stable and have amazing yields for tiny die, but when you start getting to die that are 750 mil per side or something like that, the yields will plummet." A former product director recalled prior nodes where qualification results did not hold in production. Process stability and large-die yield are different problems, and Intel's largest, highest-priced chips are the most exposed.
The October 14A PDK release is the next checkpoint. Management framed it as the moment customers get excited. Sources framed it as the window where a whale either commits volume or stays in pilot.
That gap is what separates a foundry business from a foundry promise.
Source views are those of former employees interviewed by Heron. Not investment advice.