Intel previously said that the upcoming Core Ultra chip, Lunar Lake, won’t be manufactured internally. The company is now saying the same for its next-gen desktop chip, Arrow Lake, as well.
Will the decision affect Intel’s ability to release the chip on time, its features, or its clock speeds? Almost certainly not. But working with “external partners” to produce Arrow Lake will entail higher manufacturing costs than using Intel’s own factories — costs which may or may not be passed along to consumers.
It’s also a matter of pride as Intel has been manufacturing its own chips for decades. But Intel chief executive Pat Gelsinger said that Lunar Lake would break that streak and make all of its high-profile chiplets at TSMC. Now Arrow Lake is doing the same.
Good news or bad news?
Intel’s decision can be seen in one of two ways.
First, Intel said earlier this year that it intended Arrow Lake to be designed on Intel’s 20A process, among the first of Intel’s “angstrom” generation of manufacturing technologies. But that’s no longer the case. Intel later said it will skip 20A and move on to 18A. As a result, Intel sees this as a positive move, an endorsement that its “five nodes in four years” manufacturing strategy is actually moving faster than expected:
“One of the benefits of our early success on Intel 18A is that it enables us to shift engineering resources from Intel 20A earlier than expected as we near completion of our five-nodes-in-four-years plan. With this decision, the Arrow Lake processor family will be built primarily using external partners and packaged by Intel Foundry.”
That’s one way of looking at it.
But it can also be viewed against a negative backdrop, too: after investing billions of its own money into creating Intel Foundry — a foundry business to potentially rival TSMC and others — Intel received billions more from the US government, but still doesn’t have any major public customers to show for it. Then, a month ago, Intel announced layoffs for thousands of employees, with billions of cost cuts also in the offing.
Intel said previously that it’s betting big on 18A, which is designed to debut with Intel’s next-gen Panther Lake chip due in 2025. Tom’s Hardware reported that 18A should tape out in early 2025, and Intel said that its defect density (the number of errors per square centimeter) was less than 0.4. That number tends to decline sharply as production approaches, as a graph from TSMC points out. In short, Intel’s progress on manufacturing quality is already good and should get better.
But Intel’s decision to use an outside foundry for Arrow Lake means that Intel will have to pay more for the privilege of manufacturing the chip and suffer yet another blow to its pride. Meanwhile, customers may still wonder if the decision is a sign of weakness or if Intel’s fabs are now back on track. Arrow Lake and Lunar Lake are just on the cusp of shipping, but 18A and Panther Lake may be the real focus for the company.