Intel's 14A chip is improving faster than expected, but its foundry business lost $2 billion last quarter

Intel's CFO told investors this week that the company's next chipmaking process, called 14A, is improving faster than its own engineers had planned. The same week, Intel's foundry business, the division that manufactures chips designed by other companies, reported a $2.09 billion operating loss on $5.77 billion in revenue, and Intel shares fell 2.85% on Friday. The technology story and the business story are now two different stories running in parallel.

14A is Intel's next-generation manufacturing process, the recipe for etching transistors onto silicon wafers. The metric CFO David Zinsner pointed to is defect density, the number of imperfections on each wafer. Lower defect density means more usable chips per wafer, which in turn means lower cost per chip and better fab utilization. Zinsner said 14A's defect density is tracking above Intel's own target curve and characterized the trajectory as the strongest Intel has seen since its 22nm process, more than a decade ago.

The catch is that defect density is a leading indicator, not a result. The number that matters for Intel's bottom line is whether 14A reaches profitable high-volume manufacturing on the schedule the company has set: risk production in the second half of 2027, with high-volume manufacturing in 2028. Risk production is the first batch of chips built with a new process, used to validate equipment and yields before full-scale manufacturing begins. That is at minimum a two-year gap between an improving metric and the revenue that metric is supposed to unlock.

The foundry division's quarterly results make the gap concrete. Intel Foundry booked $5.77 billion in second-quarter revenue against a $2.09 billion operating loss, per Hindustan Times reporting that consolidates the company's filings. The unit is scaling, with revenue up, but it is not yet breaking even, and the 14A process is the bet that closes the gap. Until 14A reaches high volume, the foundry is essentially paying to learn.

One piece of evidence supports the bet without overstating it. According to Zinsner, potential foundry customers have shifted from asking Intel for technology data to asking about 14A manufacturing capacity. That change in the conversation, from "show us your data" to "how much capacity can you give us," is how a fab transitions from selling promises to selling output. It is not a signed contract, and Zinsner described the talks as ongoing. It is, however, the first signal in several years that the foundry is being taken seriously as a manufacturing option, not just an Intel-internal cost center.

The market read the same week as a verdict on the gap, not on either signal alone. Defect density improving faster than planned is genuinely good news for a process whose economics depend on yield. A $2.09 billion quarterly loss is genuinely bad news for a division that has to show it can stand on its own. The stock's 2.85% drop suggests investors are pricing the distance between the two rather than rewarding one and punishing the other.

What to watch through 2027: whether 14A actually enters risk production in the second half of the year, whether Intel Foundry's quarterly loss narrows as existing nodes fill out, and whether any of the "capacity" conversations convert into named foundry customers. If risk production slips past the second half of 2027, the defect-density story stops being a story.