Converting Intel 18A Capacity to Intel 3: Waste or Smart?

One of the quieter decisions on the October 29 earnings call of Intel Corporation (NASDAQ: INTC) may matter more than the headlines. Intel will take some of its Oregon 18A capacity, wafers that would mostly have gone into Panther Lake laptop chips, and use the space to produce Intel 3 instead, starting in the second half of 2027. On the surface it looks like a company stepping back from its most advanced process. Look closer and it is a trade-off between the business Intel has today and the one it hopes to have tomorrow.

What Intel announced

Intel’s data center business is short of supply. Granite Rapids, the Xeon 6 server chip that Lip-Bu Tan called Intel’s largest server revenue source, is made on Intel 3, and Intel says it is still constrained on that process. CFO Dave Zinsner explained on the call that the capacity being taken offline in Oregon is 18A wafers that would generally have built Panther Lake, and that freeing it up allows Intel to produce more data center wafers. Tan said the plan is to have Intel 3 in production there in the second half of next year.

Zinsner also admitted the other side of the trade: 18A will not be as fully loaded as Intel would like, and moving Intel 3 into Oregon only partly fixes that. Underused 18A capacity will remain a headwind for gross margin.

The case that it is a waste

The objection is easy to state. Intel spent years and billions of dollars to bring up 18A, the first process in the world to ship gate-all-around transistors with backside power for revenue. A fab line that is already tooled and running 18A is a valuable asset. Converting part of it to an older node means re-tooling, a long gap before the first new wafers come out, and less 18A capacity ready for the day an external foundry customer signs. For a business trying to prove it is a credible alternative to TSMC, shrinking the leading-edge line sends the wrong signal.

There is also a timing risk. The Intel 3 output arrives in the second half of 2027. By then the server market may have moved toward Diamond Rapids on 18A-P, and Granite Rapids demand may have peaked.

The case that it is smart

The stronger argument is that Intel is following the money. The PC market is weak, with memory shortages pushing OEMs to cut inventory, so Panther Lake wafers are the least urgent use of 18A right now. Server demand is the opposite: DCAI revenue reached $7.3 billion in the third quarter, up nearly 80% from a year earlier, and Intel says demand is running ahead of what it can supply. Moving capacity from where demand is soft to where it is strong, and where products sell at higher prices, is what a capital-constrained company should do.

It also fits a longer plan. Intel has said Oregon will remain the pilot line for every new process, that it wants to push 18A volume to Arizona, including Fab 62, its second high-volume 18A fab, and that Oregon will run volume on 14A. 18A is mostly an internal node; outside customers are being offered 18A-P and 14A. The 18A capacity in Oregon was never going to be where external customers landed.

The AMD problem Intel is trying to solve

There is a competitive reason for urgency. AMD outsources its manufacturing to TSMC, so when server demand surges it can buy more wafers without building or converting a fab. Intel has to make its own supply, and every quarter it cannot ship enough Xeon is a quarter in which customers can try AMD instead. Converting existing space is faster and cheaper than building new shells. The conversion is Intel’s way of competing for server share now, rather than waiting for the next node.

Even then, wafers are not the only limit. Intel says it is also short of substrates, and management expects wafer and substrate constraints to cap growth in the first half of 2027. More Intel 3 wafers only help if there is packaging material to put them on.

Our view on Intel Foundry

On balance, the conversion looks less like a retreat from 18A and more like triage. Intel is putting scarce capacity where the margin is today, while keeping its leading-edge story focused on 18A-P and 14A, where external customers will actually decide. The risk is execution: a late Oregon ramp would leave Intel with neither the extra Xeon supply nor the 18A capacity. The milestone to watch is Intel 3 production in Oregon in the second half of 2027, and whether the customer announcements Tan promised for the next two quarters land on 18A-P or 14A.

Sources: Intel Q3 2026 earnings call (October 29, 2026); Intel management remarks at the Deutsche Bank Technology Conference (August 2026) and the UBS Global Technology Conference (December 2025); Intel Q2 2026 earnings call. Blue Team analysis.